Dataquest

Dataquest

Dataquest

A Gartner Group Company 251 River Oaks Parkway, San Jose, CA 95134-1913

(408) 468-8000 Fax (408) 954-1 780

February 1996

Dear Dataquest Client,

By now, you shotild have received the 1996 Dataquest research program binders for filing the newsletters and reports that will be sent to you throughout the year.

To let you know what documents you can expect to receive as part of your subscription, enclosed please find the 1996 datasheet for the research program to which you subscribe. The back side of the datasheet lists the Dataquest

Perspectives, Market Trends reports. Market Statistics books, reports, and electronic newsletters that are included in this year's research portfolio.

If you have any questioris about the research schedule, please contact your

Dataquest research analyst or client services representative.

Sincerely,

Jeffrey A. Byrne

Vice President

Worldwide Marketing

#

December 1995

Dear Dataquest Client:

In 1996, Dataquest will celebrate its 25th year as the leading global supplier of market intelligence to the IT vendor and financial conununities. I would like to thank you, on behalf of all Dataquest associates worldwide, for your support. We are proud to be your information partner by providing the IT market insight and analysis you need to make crucial business and planning decisions.

The enclosed binder is for fiBng and storing the printed market research newsletters and reports that you will receive on an ongoing basis throughout 1996 as part of your subscription to Dataquest. You may notice that we've streamlined the binder tab and document filing structure this year. We hope that this 5-tab scheme increases your efficiency in filing and locating documents.

You probably know that in addition to paper-based delivery, Dataquest is also committed to delivering our market statistics and analysis electronically. We expect that our electronic products, known collectively as Dataquest on the Desktop, will play an increasing role in our ability to deliver information to you in a timely, efficient way. For your information, our electronic tools include:

Dataquest on Demand—Our monthly CD-ROM containing a roUing 13 months of Dataquest's printed documents

MarketView — A data analysis tool containing many of Dataquest's market statistics databases

Electronic NewsTakes and Dataquest Alerts — Weekly/event-driven summary and analysis of top IT new^s, published via e-mail or fax by most Dataquest research groups

Dataquest Interactive — Our Internet-based electronic delivery system that you are invited to preview at this URL: http://www.dataquest.com

One last note: an optional binder called Electronic News is available on request for clients who wish to file their electronic newsletters and Dataquest Alerts. To order your copy, please fiU out the

FaxBack form found in the binder pocket and fax it back to us.

We look forward to working with you in our continuing process to improve the content, quality, and timeUness of our products and services. I encourage you to share with us your comments about our publications and electronic deUvery tools.

Sincerely,

Jeffrey A. Byrne

Vice President, Worldwide Marketing

F:.'':i;,.'Mf .

'17 ; 1 ' ; ' . ; : ^

Computer

Systems and

Peripherals

Online,

Multimedia, and Software

Services

1996 RESEARCH PROGRAMS

From semiconductors to systems, software to services, telecommunications to document management, Dataquest's scope of expertise provides clients with a clear view of the relationships among information technology segments — relationships that can have a profound impact on making strategic business decisions.

Computer Systems

Client/Server Computing Worldwide

Computer and Client/Server Systems Europe

Servers Europe

UNIX a n d Open Systems Europe

Workstations

Advanced Desktop and Workstation

Computing Worldwide

Workstations Europe

Computer Storage

Removable Storage Worldwide

Optical Disk Drives Worldwide

Optical Disk Drives Europe

Rigid Disk Drives Worldwide

• RAID Storage Systems Worldwide

Rigid Disk Drives Europe

Tape Drives Worldwide

Tape Drives Europe

Graphics

Graphics and Displays Worldwide

Personal Computing

Personal Computers Worldwide

Personal Computers Strategic Service Europe

Personal Computers Asia/Pacific

Mobile Computing Worldwide

PC Distribution Channels Worldwide

PC Distribution Channels Europe

Desktop PC Technology Directions Worldwide

Mobile PC Technology Directions Worldwide

Personal Computers Central and Eastern Europe

Quarterly Statistics

Advanced Desktop and Workstation Quarterly Statistics

Worldwide

Workstation Quarterly Statistics Europe

Server Quarterly Statistics North America

Server Quarterly Statistics Europe

PC Quarterly Statistics United States

PC Quarterly Statistics Europe

PC Quarterly Statistics Japan

PC Quarterly Statistics Asia/Pacific

PC Quarterly Statistics Worldwide by Region

Emerging Technologies

Multimedia Worldwide

Multimedia Europe (Module)

Online Strategies Worldwide

Online Strategies Europe (Module)

Productivity/Development Tools

Client/Server Software Worldwide

Workgroup Computing Worldwide

Workgroup Computing Europe (Module)

Persona] Computing Software Worldwide

Personeil Computing Software Europe (Module)

Technical Applications

AEC and CIS Applications Worldwide

Electronic Design Automation (EDA) Worldwide

Mechanical C A D / C A M / C A E Worldwide

C A D / C A M / C A E / G I S Europe (Module)

C A D / C A M / C A E Asia/Pacific (Module)

Customer Services

Customer ServiceTrends North America

Customer Services and Management Trends

Europe

Professional Services

Professional Service Trends North America

• Systems Integration and Applications

Development

• Consulting and Education

• Systems Management

Vertical Market Opportunities North America

Professional Services Europe

• Systems Integration

• Consulting a n d Education

• Systems Management

Professional Services Vertical Market Opportvmities

Europe

Professional Service Trends Asia/Pacific

Sector Programs

System Services North America

• Desktop Services

• Notebook Services

• Server Services

User Computing Services Europe

Network Integration and Support Services North America

Network Integration and Support Services Europe

Software Services North America

Strategic Service Partnering North America

Dataquest

1 9 9 6 R E S E A R C H P R O G R A M S

Document

Management

Copiers

Copiers North America

Copiers Europe

Facsimile

Facsimile North America

Printers Europe

Colour Products Europe (Module)

Printer Quarterly Statistics Europe

Printer Distribution Channels Europe

Printers Asia/Pacific

Printer Quarterly Statistics Asia/Pacific

Printers

Printers North America

Semiconductors Regional Mari<ets

Semiconductors Worldwide

Semiconductors Europe

Semiconductors Japan

Semiconductors Asia/Pacific

• China/Hong Kong

• Taiwan

• Korea

• Singapore

Devices

ASICs Worldwide

ASIC Applications Europe

Memories Worldwide

Memory Applications Europe

Memory IC Quarterly Statistics Worldwide

Embedded Microcomponents Worldwide

Microcomponent Applications Europe

DRAM Quarterly Supply/Demand Report

User Issues

Semiconductor Supply and Pridnfi Worldwide

Application Markets

Semiconductor Application Markets Worldwide

Semiconductor Application Markets Europe

Semiconductor Application Markets Asia/Pacific

Communications Semiconductors & Apphcations WW

Consumer Multimedia Semiconductors & Applications

Worldwide

Semiconductor Directions in PCs & PC Multimedia WW

PC Teardown Analysis

PC Watch Europe

Electronic Equipment Production Monitor Europe

Electronic Application Markets Europe — Automotive

Electronic Apphcation Markets Europe —Communications

Electionic Application Markets Europe — Consimier

Electronic Application Markets Europe — EDP

Manufacturing

Semiconductor Equipment, Manufacturing, & Materials

Worldwide

LCD Industry Worldwide

Semiconductor Contract Manufacturing Worldwide

Telecom- Networking

munlcations Networking Nortt America

• Local Area Networks North America

• Wide Area Networks North America

• Modems North America

Networking Europe

• Asynchronous Transfer Mode Europe

• ISDNEwrope

• Modems Europe

• Local Area Networks Europe

• WANs Europe

Quarterly Market Watch North America

Intelligent Hubs & Switches

• Network Interface Cards

Network Distribution Channels Europe

Voice

Voice Comm.unications North America

• Voice Processing North America

• Computer-Integrated Telephony &

Automatic Call Distributors North America

• Premise Switching Systems North America

Voice Commimications Europe

• Voice Processing Europe

• Call Centres Europe

• Telephones Europe

• PBX/KTS Systems Europe

Public

Public Network Equipment & Services North America

• Public Network Eqmpment North America

• PubUc Network Services Norih America

Public Network Equipment & Services Europe

• Public Network Equipment Europe

• Public Network Services Europe

Personal

Cellular Telephony Worldwide

Personal Communications Norih America

Personal Communications Europe

• Infrastructure and Services Europe

• Terminals Europe

Personal Communications Distribution Europe

Cross-

Technology

Programs

Technology Insights for:

Financial Services

Government Agencies

Publishing, Media, and Consulting Firms

IT Business Development for Financial Organizations

IS and Purchasing Organizations

IT Supporting Industries

Emerging IT

Markets

Central and Eastern Europe

Personal Computers

Telecommimications

Latin America

Personal Computers

Printers

Asia/Pacific

IT Market Insight Asia/Pacific

Personal Computers Asia/Pacific & Quarterly Statistics

Printers Asia/Pacific & Quarterly Statistics

Professional Service Trends Asia/Pacific

• Country-level reports on Asia/Pacific IT markets

Dataquest

A Gartner Group Company

Corporate Headquarters

251 River Oaks Partairay

San Jose, CA 95134-1913

United States

Phone: 1-40S-I168-8000

Fax: 1-108-954-1780

Fax-on-Demand: Dial 1-800328-2954 and press 4 (Limited to North America)

©1996 Dataquest Incorporated

Boston Area

Nine Tednology Drive

P.O. Box 5093

Wfeslborough, MA 01581-5093

United States

Phone: 1-508-871-5555

Fax: 1-508^71-6262

United Kingdom

Holmets Farm Way

High V\4^x>mbe, Buckinghamshire

HP124XH

United Kingdom

Phone: 4441494422 722

Fax: +441494422 742

Tokyo

Shinkawa Sanko Building

6th Floor

1-3-17, Shii*a«ra

Chuo*u, Tokyo 104

Japan

Phone: 81-3-5566-0411

Fax: 81-3-5566-0425

Dataquest is a registered trademark of the A.C. Nielsen Company

Program Code: NONE-WW MKTG 1/96 (PSB)

,1 r ( - « :'

North America

Europe

Japan

Dataquest

Invitational

Computer

Conferences

DATAQUEST 1996 CONFERENCES

Dataquest sponsors an on-going series of conferences and invitational events focusing on trends and issues in information technology and IT services. These conferences are the preeminent source of insight and analysis of global IT market dynamics.

January 24

January 30

February 20

March 7

April 1-2

April 1 *

May 6-7

May 13-14

June 26-27

July 1 *

September 25-26 *

October 24-25

December 1 *

Capitalizing on the Wireless Phenomenon

Dataquest Predicts

Dataquest Predicts

Channel Trends Conference

ServiceTrends Conference

Mining the Internet

Personal Computer Conference

Copier Conference

Storage Track Conference

SEMICON/West

Multimedia

Semiconductors '96

Mining the Internet

San Jose, California

Boston, Massachusetts

San Jose, California

San Jose, California

Orlando, Florida

Boston, Massachusetts

San Jose, California

Boston, Massachusetts

Monterey, California

San Francisco, California

San Jose, California

Palm Desert, California

San Jose, California

January 24

May 22-23

September 10

Computer Storage

Semiconductors '96

Computer Storage

May 13-14

September 10-12

December 6

December 1 *

December 1 *

December 1 *

December 1 *

December 1 *

December 1 *

March 5

April 10

April 24

September 24

April 1

May 21

October 30

November 6

Semiconductors '96

Computers and Peripherals

Telecommunications

Asia/Pacific Series

Asia/Pacific Series

Asia/Pacific Series

Asia/Pacific Series

Asia/Pacific Series

Asia/Pacific Series

Dataquest Storage Solutions Series •USA

Mediterranean Series

Mediterranean Series

Mediterranean Series

Mediterranean Series

Munich, Germany

Frankfurt, Germany

London, England

Tokyo, Japan

Tokyo, Japan

Tokyo, Japan

Tokyo, Japan

Seoul, Korea

Beijing, PRC

Shanghai, PRC

Xi'an, PRC

Guangzhou, PRC

San Jose, California

Irvine, California

Nashua, N e w Hampshire

Newton, Massachusetts

Dubai, UAE

Athens, Greece

Tel Aviv, Israel

Istanbul, Turkey

* Date tentative/tnay change

Dataquest

D A T A Q U E S T 1 9 9 6 C O N F E R E N C E S

Dataquest

Invitational

Computer

Conferences

(continued)

January 17

January 23

January 30

February 1

June 10

June 12

June 21

Jime25

J u l y l

September 1

September 5

September 11

September 19

Dataquest Storage

Solutions Series-Europe Paris, France

Dataquest Storage

Solutions Series-Europe Munich, Germany

Dataquest Storage

Solutions Series-Europe Milan, Italy

Dataquest Storage

Solutions Series-Europe Rome, Italy

Dataquest Storage

Solutions Series-Europe Budapest, Hungary

Dataquest Storage

Solutions Series-Europe Prague, Czech Republic

Dataquest Storage

Solutions Series-Europe St. Petersburg, Russia

Dataquest Storage

Solutions Series-Europe Moscow, Russia

Dataquest Storage

Solutions Series-Europe Warsaw^, Poland

Dataquest Storage

Solutions Series-Eurojje Amsterdam, Holland

Dataquest Storage

Solutions Series-Europe Stockholm, Sweden

Dataquest Storage

Solutions Series-Europe London, England

Dataquest Storage

Solutions Series-Europe Frankfurt, Germany

Want more information about

Dataquest?

Place your request by calling our

Fax-on-Demand system at

1-800-328-2954

October 1 *

October 1 *

October 1 *

October 1 *

October 1 *

October 1 *

October 1 *

Latin America Series

Latin America Series

Latin America Series

Latin America Series

Latin America Series

Latin America Series

Latin America Series

February 19

February 22

South Africa Series

South Africa Series

Caracas, Venezuela

Mexico City, Mexico

Sao Paulo, Brazil

Buenos Aires, Argentina

Santiago, Chile

Bogota, Columbia

Lima, Peru

Capetown, South Africa

Johannesburg, South Africa

April 11

April 30

May 1

May 9

LINK Series North America

LINK Series North America

LINK Series North America

LINK SeriesNorth America

May 14

May 21

November 1 *

LINKSeriesNorth America

LINK Series -

North America

LINK Series -

North America

November 1 *

November 1 *

LINK Series-

North America

LINKSeriesNorth America

November 1 * LINK Series • North America

November 1 * LINK Series North America

Orlando, Florida

Austin, Texas

Philadelphia, Permsylvania

Charlotte, North Carolina

Denver, Colorado

Portland, Oregon

Montreal, Quebec

Ottawa, Ontario

Calgary, Alberta

Vancouver, BC

Toronto, Ontario

*Date tentative/may change

Dataquest

A Gartner Group Company

Corporate Headquarters

251 River Oaks Parkway

San Jose, CA 95134-1913

United States

Phone: 1JHD8-468-8000

Fax: 1-4Ce-954-1780

FaxKsn-Demand: Dial 1-80(W28-2954 and press 4 (Lin:iited to North America)

©1996 Dataquest Incorporated

United Kingdom

Holmers Farm Way

HighWV»nnl)e,

Buckinghamshire

HP124XH

United Kingdom

Phone: +441494422 722

Fax: +441494422 742

Fiance

Immeuble Defense Bergeres

345, avenue Georges

Ciemenceau

TSA 40002

92882 - Nanterte CTC Cedex 9

France

Phone: +33141351300

Fax: +331 41 3613 13

Boston Area

Nine Technology Drive

P.O Box 5093

WtestlxxDugh, H/IA 01581-5093

United States

Phone: 1-508-871-5555

Fax: 1^608-871-6262

Germany

Kronstadter Strasse 9

81677 Munchen

Deutschland

Phone: +49899309090

Fax: +49899303277

Asiaypacific

7/F China Underwriters

Centre

88 Gbucester Road

V t ^ C h a i

Hong Kong

Phone: 85228246168

Fax: 85228246138

Israel

Phone: +97 2 9 926 111

Fax: +97 29 925 791

Italy

Phone: +39 2 24 40 539

Fax: +3922624400

Japan

Shinkawa Sanko BuiMing

6thFk)or

1-3-17, Shinkawa

Chuo-ku, Tokyo 104

Japan

Phone: 81^3-55660411

Fax 81-3-5566-0425

South Africa

Phone: +27 11 468 1084/7'

Fax: +27114681241

Spain

Phone: +34 1 57 13 804

Fax: +3415714266

Dataquest is a registered trademark of the A C Nielsen Company

Program Code: DQGE-VWV MKTG 1/96 (PSB)

4

fytt-;;nt i^fJ

^illt'i^^ih

4

4

SEMICONDUCTOR EQUIPMENT,

MANUFACTURING &

MATERIALS WORLDWIDE

Dataquest's Semiconductor Equipment, Manufacturing & Materials Worldwide program is a comprehensive market research service emphasizing an integrated perspective of the interdependence of wafer fabrication equipment technology, semiconductor materials applications, and IC process technology as they relate to the broader issues of semiconductor manufacturing.

Market Coverage

Semiconductor

Manufacturing Issues

Dataquest provides executives with strategic insight into issues relating to the global nature of the semiconductor manufacturing envirorunent. Topics include:

• Global fab facilities

• Semiconductor production

• Capital spending

• Capacity trends

• Factory productivity

• Factory automation

• R&D trends

• IC process technology

Wafer Fabrication

Equipment

Detailed market estimates and forecasts are provided for products in the following broad categories:

• Lithography

• Etch and clean

" Deposition

• Diffusion/RTF

• Implantation

• Wafer inspection

• Chemical mechcinical polishing

Semiconductor Materials

The following represents some of the areas covered in our materials research:

• Silicon and epitaxial wafers

• Gases

• Photoresist

• Other materials

Partnering to

Provide Solutions

As a client, you have direct access to experienced analysts w h o can provide insights and advice on market dynamics, industry events, and competitive issues.

Want more information about

Dataquest?

Place your request by calling our

Fax-on-Demand system at

1-800-328-2954

Inquiry Support

Personalized inquiry support is a primary component of your

Dataquest annual subscription program.

Through an interactive approach, Dataquest analysts w o r k with y o u to tailor the progrEim to meet the unique needs of your orgEinization.

Information Resource

Centers

Clients have unlimited access to Dataquest's extensive print and online resource libraries worldwide.

Electronic Delivery

Dataquest offers a variety of tools, k n o w n collectively as Dataquest

On The Desktop, that have

the power to deliver

Dataquest insights directly to you, whether you are on the road or in the office.

Optional Custom Research

Should your needs exceed the scope of this program,

Dataquest offers extensive primary research and consulting services.

Dataquest

W H A T Y O U W I L L R E C E I V E A S A C L I E N T

SEMICONDUCTOR EQUIPMENT, MANUFACTURING & MATERIALS

WORLDWIDE

Perspectives Dataquest Perspectives present analysis and commentary on key technologies, companies, market opportunities, trends, and issues in the market. A minimum of six

Perspectives will be published on an event-driven basis throughout the year, as well as two Dataquest Predicts.

Dataquest Predicts: In these hard-hitting documents, Dataquest takes a bold, opinionated, often controversial look at key issues, products, and trends shaping the

"semiconductor equipment and materials markets. The reports make predictions about why, when, and how events will happen.

Marl<et Trends

Marltet Statistics

Reports

Semiconductor Equipment, Manufacturing, and Materials Forecast: This report, updated twice a year, contains forecasts for semiconductor consumption and production by region, semiconductor capital spending by region, silicon wafer consumption by region, and wafer fabrication equipment consumption by region and by major equipment category.

Available in July and December 1996

Semiconductor Consumption and Shipment Forecast: Five-year revenue forecasts for the global semiconductor market by region.

Available in the Second and Fourth Quarters 1996

Wafer Fabrication Facilities—North America, Japan, Europe, and Asia/Pacific:

Pilot and production fab lines by company, location, fab name, products produced, process technology, minimum linewidth, wafer diameter, estimated wafer capacity, square feet of cleanroom, and cleanroom class. (Foitr separate publications)

Available in November 1996

Silicon Market Share: Market share by company and by region for merchant suppliers, as well as trends in the merchant silicon and epitaxial wafer market, captive silicon production, wafer prices, and silicon square-inch consumption.

Available in June 1996

Wafer Fabrication Equipment Market Share: Market share by company by region for 150 companies that participate in 40 different segments of front-end wafer fab equipment.

Available in May 1996

Focus Report - Planned Fabs: This report will deliver a forward-looking assessment of planned fabs emd related market issues.

Available in the First Quarter 1996

Electronic News

Dataquest Alerts : Event-driven news and analysis, delivered by fax, giving a concise overview of significant announcements in the semiconductor industry.

Event-driven

The DQ Monday Report - OPTIONAL: Weekly news and commentary on semiconductor industry events and issues with a monthly snapshot of semiconductor pricing for 25 key semiconductors in 6 regions. Clients may wish to subscribe to this valuable newsletter for a broad view of this dynamic industry.

Available Weekly via Electronic Mail

Conferences Dataquest hosts the industry's preeminent semiconductor conferences in the United

States, Europe, Japan, Taiwan, emd Korea. Clients receive a discount on conference fees.

#

4

DataQuest

A Gartner Group Company

Corporate Headquarters

251 River Oaks Paikway

San Jose, CA 95134-1913

United Slates

Phone: 1-408-468-8000

Fax 1-408-954-1780

Fax-Bacic 1-800^28-2954 and press 4

<£>1996 Dataquest

Bosun Area

Nine Technology Drive

P.O. Box 5093

Westborough, MA 01581-5093

United States

Phone: 1-508671-5555

Fax 1-508671-6262

United Kingdom

Holmers Famn Way

High Wycombe

Buckinghamshire

HP124XH

United Kingdom

Phone: +441494422722

Fax +441494 422 742

Asia/Pacific

7/F China Undeiwriters Centre

88 Gloucester Road

Wan Chai

Hong Kong

Phone: 85228246168

Fax 85228246138

Japan

Shinkawa Sanko BuikJing ethFkxir

1-3-17, Shinkawa

Chuo-ku, Tokyo 104

Japan

Phone: 81-3-5566-0411

Fax 81-3-556&0425

Dataquest is a registered trademark of the A C . Nielsen Company

Program Code: SEMM-WW MKTG 1/96 (JCG)

#

#

Dataquest Fax Back—408>954'1780

To:

City:,

Leticia Martinez

San Jose, California

Co.

Country:.

Dataquest Incorporated

U.S.A.

_Total Pages 1 of 1

Here's How to Order Your Electronic News Binder

Dataquest provides a separate binder called

Electronic News to help you organize your printouts of the electronic newsletters and

Dataquest Alerts that will be sent to you by your

Dataquest North America research programs throughout the year.

Although not all clients will print out electronic news bulletins or file faxes, the Electronic News binder is available by request for those who do.

To order your Electronic News binder, just fill out the form below and fax it back to us. We will mail your binder to you immediately.

Note: If you subscribe to more than one

Dataquest North America research program, then indicate how many binders you need in the space provided below (plan on one binder per research program), and we'll send them to you in one shipment.

Thank you for helping us serve you better.

Customer Name

Title

Company

Street Address

City

Country

Telephone

Fax

Total number of Electronic News binders needed:

State/Province

Postal/Zip Code

Dataquest

251 River Oaks Parkway, San Jose, CA 95134-1913

SEMICONDUCTOR EQUIPMENT, MANUFACTURING

AND MATERIALS WORLDWIDE 1996

VOLUME 1

TABLE OF CONTENTS

PERSPECTIVES

9604

9605

9606

9607

9608

7/22/96

9/9/96

9/16/96

9/30/96

10/28/96

Tight Longer Term

Polysilicon Supply/Demand Update: The Industry

Responds and ReUef is in Sight, but only after 1997

The Dynamic Etch Equipment Market: 1995 Market

Review and Projection

How Long and Deep[ WiU the Slowdown in Wafer Fab

Equipment Be? What WiU Be the Shape of the Industry

Recovery?

CVD Market Analysis

The 1995 Worldwide Wafer Fab Equipment market: A

Record Year-Competitive Analysis of Regional Dynamics,

Industry Segments, and Company Market Shares

MARKET TRENDS

9601

9602

5/13/96

8/26/96

9603 3/17/97

Semiconductor Five-Year Forecast Trends-spring 1996

Midyear 1996 Forecast: Capital spending, Wafer Fab

Equipment and Sdicon

Year-End 1996 Forecast: Capital Spending, Wafer Fab

Equipment, and Sihcon Markets

GUIDES

9601

2/19/96

Market Definitions ~ Semiconductor Equipment

Dataquestyt T p ]? HT

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

With Applied's Acquisition of Opal and Orbot, Has tlie n/lerger and Acquisition Era Begun in tlie Semiconductor

Capital Equipment {Market?

With the acquisition of Opal Inc. and Orbot Instiuments Ltd., Applied Materials Inc. enters the process contiol market. AppHed purchased Opal for $175 million and Orbot for $110 million.

Based on the financial community's estimates of the revenue of Opal and Orbot, the acquisition took place for tinder 3 times and 2.5 times each company's revenue, respectively.

Implications

We believe there are two aspects to this event:

• The impact of Applied's entiy in the process contiol arena

• The capital equipment industiy's continuing tiend of market concentiation

Applied's Entry into Process Control

Applied has acquired two very strong companies. We estimated 15 percent market share for

Opal in the 1995 CD-SEM market. It foUow^s only the market leader Hitachi, with its estimated

7b percent share of the $313 market in 1995. In the U.S. market, Opal's market share is almost equal to that of Hitachi. Its weakness is in Japan and Asia/ Pacific, where the market is, on average, 1.5 times that of the United States, and Hitachi dominates both regions with over 90 percent market share.

Orbot participates in the mask and wafer inspection market, the latter generating half of the revenue of the first. Orbot has recently entered the wafer inspection market, and in 1995, it did not hold significant market share compared with the top three players — KLA Instruments (66 percent), Tencor (14 percent), and Hitachi (10 percent). In mask inspection, with about onequarter market share, Orbot is a stiong second to the industry leader, KLA Instiuments. This is basically a two-player market. Maskmaking equipment is a significant market, and it should experience strong growth in the next several years as semiconductor manufacturing faces capacity shortage and mask shops (captive and merchant) build capacity fast.

December 2,1996 ©1996 Dataquest

SEMM-WW-DA-9607 Semiconductor Equipment, Manufacturing, and Materials Worldwide

It is important to note that Orbot and Opal's product offering is directly related to lithography exposure systems, or better steppers. The stepper market is a highly cyclical market that follows closely the DRAM technology and price-per-bit transitions and trends. Process control as a whole has gained more prominence in the past several years, with w^ell over 50 percent average growth, compared with the overall front-end equipment market growth of under 35 percent. However, we believe, given the overall market conditions and the DRAM cycles, that process control will grow at slightly below the overall equipment market's growth (estimated at 16 percent) over the next five to six years.

Therefore, Applied's acquisitions may not be for the sole reason of increasing revenue but also to enable Applied to incorporate process control into its vast and dominating product portfolio and perhaps offer its clientele a turnkey fab in the future. Does this imply that Applied will be looking at the only two major equipment segments that are not included in its products — lithography and clean process? Well, we will have to wait and see, but this brings us to the second issue concerning the overall trend of market concentration in the equipment market.

Market Concentration

We have focused and quantified this rather overwhelming trend for some time now. Dataquest has in publication a report that analyzes the implications of this tiend for the participants in the capital equipment market and the overall semiconductor manufacturing arena. Of the 10 major equipment categories, we cover seven that are segments in which the top two players have at least two-thirds of their market. We do not see any reason for this trend to turn around or not to continue to grow.

It is speculative to look at the market and anticipate mergers and acquisitions. However, the season is pregnant with expectation, with good reason. The cash reserves of the equipment companies have grown tremendously in the past several years. Applied, for instance, has led and grow^n its market share from the overall market in every one of these years of hypergrowth. But Applied is not alone. The market is currently characterized as one that is experiencing a downturn, but with a silver lining, relatively soon, of recovery because of grow^ing applications across the semiconductor product spectrum.

We are seeing an upsurge in advanced equipment orders as semiconductor manufacturers are gearing for the next generation of devices at below 0.3 micron. Our forecast calls for technology-oriented spending throughout 1997, with an eventual recovery in 1998. In this environment, market positioning becomes important, and technical portfolio plays an important role.

Many companies are "catching their breath" in these slower times. After three years of strong growth where the focus has been reactive in trying to meet customer demand and manage backlog, w e see the focus shifting to building the infrastructure to handle the next u p t u r n and trying to differentiate products and capabilities from competitors.

There are two major ways of acquiring technology and gaining advantage over competition.

First is obviously internal R&D. But with adequate cash reserves, a more direct alternative is external—or merging with or acquiring companies that fit a marketing stiategy. In Applied's

4

December 2,1996 ©1996 Dataquest

• » »

SEMM-WW-DA-9607 Semiconductor Equipment, Manufacturing, and Materials Worldwide case, w^e see an expanding product portfolio—whereas we may also see mergers in the same equipment segments between players, even the top ones.

We also see a trend in acquiring companies that are OEMs to larger e n d users. This is very strong a m o n g the software companies that support the equipment companies; however, it does not have be limited to software. These acquisitions also aUow companies to block key technology that may be available to their competitors otherwise.

In conclusion, w e see the Applied Materials acquisition as a continuing trend, for Applied and also for the overall industry. Applied is on a path to grow beyond the segments it now dominates while expanding its technology base. The industry, loaded with cash reserves, n o w needs to build for the next upsurge and to be able to compete and gain market share. Mergers and acquisitions are a very "efficient" way of doing so.

By Nader Pakdaman

4

December 2,1996 ©1996 Dataquest

Dataquest^ j^ ^ R T

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

With Applied's Acquisition of Opal and Orbot, Has tlie iyierger and Acquisition Era Begun in tlie Semiconductor Capital

Equipment ii/laricet?

With the acquisition of Opal Inc. and Orbot Instruments Ltd., Applied Materials Inc. enters the process control market. Applied purchased Opal for $175 million and Orbot for $110 million.

Based on the finandal community's estimates of the revenue of Opal and Orbot, the acquisition took place for under 3 times and 2.5 times each company's revenue, respectively.

Implications

We believe there are two aspects to this event:

• The impact of Applied's entry in the process control arena

• The capital equipment industry's continuing trend of market concentration

Applied's Entry Into Process Control

Applied has acquired two very strong companies. We estimated 15 percent market share for

Opal in the 1995 CD-SEM market. It follows only the market leader Hitachi, with its estimated

75 percent share of the $313 market in 1995. In the U.S. market. Opal's market share is almost equal to that of Hitachi. Its weakness is in Japan and Asia/Pacific, where the market is, on average, 1.5 times that of the United States, and Hitachi donunates both regions with over 90 percent market share.

Orbot participates in the mask and wafer inspection market, the latter generating half of the revenue of the first. Orbot has recently entered the wafer inspection market, and in 1995, it did not hold significant market share compared with the top three players—KLA Instruments (66 percent), Tencor (14 percent), and Hitachi (10 percent). In mask inspection, with about onequarter market share, Orbot is a strong second to the industry leader, KLA Instruments. This is basically a two-player market. Maskmaking equipment is a significant market, and it should experience strong growth in the next several years as semiconductor manufacturing faces capacity shortage and mask shops (captive and merchant) build capadty fast.

It is important to note that Orbot and Opal's product offering is directly related to lithography exposure systems, or better steppers. The stepper market is a highly cyclical market that

November 27,1996 ©1996 Dataquest

SEMM-WW-DA-9607 Semiconductor Equipment, Manufacturing, and Materials Worldwide follows closely the DRAM technology and price-per-bit transitions and trends. Process control as a whole has gained more prominence in the past several years, with well over 50 percent average growth, compared with the overall front-end equipment market growth of under 35 percent. However, we believe, given the overall market conditions and the DRAM cycles, that process control will grow at slightiy below the overall equipment market's growth (estimated at 16 percent) over the next five to six years.

Therefore, Applied's acquisitions may not be for the sole reason of increasing revenue but also to enable Applied to incorporate process control into its vast and dominating product portfolio and perhaps offer its clientele a turnkey fab in the future. Does this imply that Applied will be looking at the only two major equipment segments that are not included in its products— lithography and clean process? Well, we will have to wait and see, but this brings us to the second issue concerning the overall trend of market concentration in the equipment market.

Market Concentration

We have focused and quantified this rather overwhelming trend for some time now. Dataquest has in publication a report that analyzes the implications of this trend for the participants in the capital equipment market and the overall semiconductor manufacturing arena. Of the 10 major equipment categories, we cover seven that are segments in which the top two players have at least two-thirds of their market. We do not see any reason for this trend to turn arovmd or not to continue to grow.

It is speculative to look at the market and anticipate mergers and acquisitions. However, the season is pregnant with expectation, with good reason. The cash reserves of the equipment companies have grown tremendously in the past several years. Applied, for instance, has led and grown its market share from the overall market in every one of these years of hypergrowth. But Applied is not alone. The market is currentiy characterized as one that is experiencing a downturn, but with a silver lining, relatively soon, of recovery because of growing applications across the semiconductor product spectrum.

We are seeing an upsurge in advanced equipment orders as semiconductor manufacturers are gearing for the next generation of devices at below 0.3 micron. Our forecast calls for technology-oriented spending throughout 1997, with an eventual recovery in 1998. In this environment, market positioning becomes important, and technical portfolio plays an important role.

Many companies are "catching their breath" in these slower times. After three years of strong growth where the focus has been reactive in trying to meet customer demand and manage backlog, we see the focus shifting to building the infrastructure to handle the next upturn and trying to differentiate products and capabilities from competitors.

There are two major ways of acquiring technology and gaining advantage over competition.

First is obviously internal R&D. But with adequate cash reserves, a more direct alternative is external—or merging with or acquiring companies that fit a marketing strategy. In Applied's case, we see an expanding product portfolio—^whereas we may also see mergers in the same equipment segments between players, even the top ones.

November 27,1996 ©1996 Dataquest

SEMM-WW-DA-9607 Semiconductor Equipment, Manufacturing, and Materials Worldwide

We also see a trend in acquiring companies that are OEMs to larger end users. This is very strong among the software companies that support the equipment companies; however, it does not have be limited to software. These acquisitions also allow companies to block key technology that may be available to their competitors otherwise.

In conclusion, we see the Applied Materials acquisition as a continuing trend, for Applied and also for the overall industry. Applied is on a path to grow beyond the segments it now dominates while expanding its technology base. The industry, loaded with cash reserves, now needs to build for the next upstirge and to be able to compete and gain market share. Mergers and acquisitions are a very "efficient" way of doing so.

By Nader Pakdaman

November 27,1996 ©1996 Dataquest

Dataquest A T p /? "T"

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

* •

Wafer Fab Equipment Forecast-Putting Recent Events into Perspective: A "Square U" or a "V"?

Dataquest is in the process of updating the wafer fab equipment forecast, to be released in the first week of January. Recent positive events in the market include the semiconductor book-to-biU's exceeding 1.0, Intel's current stronger business outlook, Applied Materials' optimistic earnings conference call, and its news release of a $117 miUion order from Hyundai. Also released this past week was the equipment book-to-biil ratio at 0.75, with monthly bookings continuing a downw^ard trend.

The popular question is—have w^e hit bottom in the equipment market, or when will we? But w^e believe the most pertinent question is—what is the nature and speed of the recovery? We feel it is important to share with our cUents our current thinking in this uncertain market.

One thing is true: Current sentiment is polarized. Depending on w h o m you talk with, either w^e have bottomed and a recovery is in progress in equipment, or the bottom is ahead of us and the sluggish market conditions could extend into 1998. So now w e wiU analyze what the "shape" of the recovery will look like.

Summary of Event Details

First, the recent semiconductor book-to-bill exceeded unity in October for the first time this year as the market for PCs remains strong and as suppliers gear u p for the holiday season. Spot prices of

DRAM ticked u p slightly but have since settled back dow^n as suppliers quickly responded with inventory releases of packaged products and die inventory in wafer form.

Second, Intel recently m a d e positive comments about fourth quarter demand for MRUs and has recently increased spending plans for 1997. In mid-year, we were estimating Intel would spend

$3.9 billion for production capacity in 1996. This figure has since been reduced to $3.4 billion, and we now estimate that the company will spend over $4 billion in 1997.

Third, Applied Materials released results for the fiscal quarter ending October 1996 that were in line with expectations. Orders were $683 million, down significantly from the $1.3 biUion level experienced in the first part of the year. Sales were $861 million in the fourth quarter, down from

$1.1 billion in the prior quarter. However, this was not the significant news. Coirmients m a d e in the conference caU after market close on November 21 were much more positive than negative.

Although members of Applied Materials management did indicate that they expected the

November 22,1996 ©1996 Dataquest SEIVIM-WW-DA-9606

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide slowdown in the equipment market to continue into 1997, they did give guidance that orders in the next quarter would be in the range of $725 to $750 million and sales would be about $800 ixullion. The primary focus being placed on the call by the investment community is the fact that

orders are up sequentially from the previous queirter: a rally cry has ensued. Fourth, Applied

announced a $117 million dollar order from Hyundai for its DRAM fab in Oregon. This represents the first order over $100 million Applied has received in over six months, and it has been interpreted by some that the rebound is under way.

Fifth, Semiconductor Equipment and Materials International (SEMI) released the equipment bookto-bill for October at 0.75. While this ratio is u p modestly from the 0.70 reported for September, the bookings continued to decline in October.

Initial Thoughts and Some Background Facts on the Events

We are viewing the semiconductor book-to-bill improvement as one primarily driven by the

Christmas seasonal buying of components for the PC market. It remains to be seen w h a t the enduse d e m a n d wiU actually be and the effect on the channel flow^ of orders for components, come

January. Our PC group believes that there is less "overbuying" occurring this year than last, yet we remain cautious, thinking that order rates wHl soften again early next year as DRAM prices continue to decline.

The positive comments and recent investment increase made by Intel reflect its unique position in the market and the strong unit growth of PCs in the fourth quarter. While its increase in spending will stabilize the market in the United States in 1997, capital spending in MPU represents only perhaps 10 percent to 12 percent of the market, on average. There is no fundamental reason w h y this short-term trend would transfer to other areas of spending in present conditions.

Applied Materials is a very well-rvm company, with seasoned management that has seen many downturns. Management took quick action to position the company to w^eather the storm. For this reason. Applied should be considered to perform slightly better than the market. It should also be noted that its backlog level at the beginning of the year w^as about six months and is currently five months and declining. We would expect the company to consume and manage its backlog to about the four-month level, at the minimum. Not all equipment companies will be so weU positioned.

However, an order rate increase from $683 million to $725 or $750 million is noise-level movement in this industry, and caution should be exercised about reading too much into this increase.

Further, the $117 million order from Hyundai does not reflect a return to capacity buys, in our opinion. In a recent set of visits to companies, the consensus was that any orders received today are "strategic" buys. In the case of Hyundai, Samsung, and TSMC, orders for equipment for U.S. production are seen as strategic for the company's long-term success in the market. This kind of sustained investment is normal in the earlier stages of a downturn, and likely represents only a portion of the planned investment to position. There was a recent announcement by Samsung that capital spending will decline 17 percent in 1997 overall, meeming that spending in Korea is being cut deeper than that.

November 22,1996 ©1996 Dataquest SEMM-WW-DA-9606

Dataquest Alert

Semiconductor Equipment, Manufacturing, and Materials Worldwide

The SEMI equipment book-to-bill reflects continued order weakness. Although the ratio wiU likely strengthen in the coming months, being very heavily influenced by Applied Materials, w e would expect order rates to remain somewhat flat, with some fluctuations throughout most of 1997. This is a very volatile indicator, as the equipment business is much more "liunpy" than the chip

market. Consensus from most companies w e speak with sees order rates bottoming in the second quarter of next year.

Some Fundamental Analysis—Numbers Do Not Sum to Recover Now

The capital spending market for 1996 splits about: 30 percent foundry/logic, 50 percent memory,

10 percent MPU, a n d 10 percent others (mixed signal, analog, a n d discrete, among others). Let u s look at each capacity area separately.

Foundry/Logic

Foundry factory utilization rates are falling, and wafer prices have collapsed, according to a recent pricing survey completed. Investment will continue in the foimdry area because many projects have joint-venture and partnership commitments, but initial r a m p rates are Ukely to be cut back.

Current projections are that excess capacity may last through all of 1998 in this area, as a result.

With cheap foundry capacity available, this will Ukely subdue large capacity outlays for nonfoundry logic. While there is a stability in the investment picture here, there is also not a big driver for growth tmtil 1999.

Memory/DRAM

DRAM capacity is in oversupply today. As w e have noted several times, because the industry is in a transition from the 4Mb to the 16Mb DRAM and since die sizes continue to shrink and yields to r a m p u p , high bit demand does not necessarily translate to a similar higher requirement for silicon capacity. Table 1 shows the result of fundamental capacity analysis.

Table 1

F u n d a m e n t a l Silicon Capacity Analysis for DRAMs in 1996 a n d 1997

1996 1997 Comments

Bit Demand Growth (%)

73

70 to 75 1997 growth is industry consensus,

Dataquest forecast is about 60 percent

Growth Silicon Area Required to

Meet Bit D e m a n d (%)

m

12 to 15 Average die size, yields, and product mix for 4Mb a n d 16Mb DRAM included

Growth in DRAM Silicon Area

Capacity (%)

23

21

1997 growth factors in capital spending decline, as currently forecast, and slow r a m p in new DRAM fabs

Oversupply Percentage in Capacity

^'

14 to 17 Assumes baseline is a balanced market at the end of 1995

Source: Dataquest (November 1996)

November 22,1996 ©1996 Dataquest SEMM-WW-DA-9606

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

This fundamental analysis indicates that the "strategic investments" being made by DRAM suppliers wiU actually delay the timing of when demand catches supply. Current factory utilization rates for DRAM fabs are ruiuiing around 70 percent, and w e would expect utilization to continue falling in 1997, perhaps a percentage in the low 60s. According to this fundamental

analysis, capacity spending in DRAM is not expected to return tmtil late in 1998. An obvious question from this analysis is: What does bit demand have to be in 1997 to create a balanced market at the end of 1997? The answer is about 100 to 110 percent, meaning the average PC would have to ship with more than 32MB of memory in the next year. While higher bit demand than that shown is possible, creating a balanced market in 1997 is highly unHkely, so we would expect

DRAM prices to remain under pressure through all of 1997.

MPU and Other Areas

We do expect these areas to be the first to recover and to be tightly coupled with unit demand from the various product areas. But because this represents a smaller percentage of spending, the result w^ill only create some stability in the market short term and not be a fuel for growth.

Where Are We Likely to Take Our Forecast?

We would not expect large changes in our thesis and forecast outlook. Here is a short summary of any adjustments being considered:

• Our July forecast placed growth in wafer fab equipment for 1996 at about 17 percent. The third quarter results were significantly weaker than anticipated and, when carried forward, should lower the actual growth for 1996 in the range of 11 to 14 percent.

• Our outlook for 1997 should not change appreciably. We are getting indications that the absolute levels will be adjusted down sUghtly but, with a weaker 1996, the market should be close to our 16 percent decline forecast of July, or down in the range of 15 to 20 percent.

• Our biggest challenge is to forecast the 1998 market. In July w e were cautious on growth prospects, as we believed the fundamentals for s u p p l y / demand dynamics would continue to be weak at the beginning of the year. Given the analysis in this Alert, we would continue to be cautious on 1998 and will likely be in the range of 3 to 7 percent growth.

• At this point, w e would maintain the stronger growth forecasts for 1999 and beyond, as our outlook for semiconductor growth in the longer term remains good, but this will come under review as we progress through our forecast process currently under way.

There are some v^rho would like to believe that this recovery wiU be shaped like a "V" and that w e are at the bottom now. Although we are likely close to a bottom, we believe that the recovery in equipment spending growth will depend on fundamentals of supply and demand dynamics.

Large capacity spending growth will likely not return until well into 1998, and the industry is likely to experience several quarters of "skirting along the bottom," making the likely shape of the recovery a "square U."

By Clark Fuhs, Calvin Chang, Ronald Dornseif, Takashi Ogawa, and Nader Pakdaman

November 22,1996 ©1996 Dataquest SEMM-WW-DA-9606

Dataquest^ Z ^ i ? ^

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

MEMC Releases Weaker Second-Half Outlook—What

Does It Mean Compared with Our Forecast?

MEMC announced late on Tuesday, September 10, that the third quarter, ending September 30, would be below analysts' expectations and that the fourth quarter would be significantly below their earlier estimates and likely below the third quarter. Specific figures provided w^ere limited to the third quarter and were limited in scope.

Comments m a d e in a conference call on the morning of September 11 indicate a volume-driven slowdown, most notably in Asia, as the company's Korean joint venture, Posco-Hiils, is experiencing lower sales. It v/as mentioned that the lower volumes result from semiconductor manufacturers conserving cash by low^ering wafer inventories on hand, as well as from the weakness in the memory market. MEMC noted that semiconductor manufacturers are likely to take advantage of the current situation in the soon-to-begin contract negotiations for 1997 in terms of pricing. No specifics were given.

Summary of Details—Calibrating MEMC to the Market

MEMC specifically released the information that it expected third quarter sales to be in the neighborhood of $290 million, dow^n from second quarter revenue of $324 million. Expectations for the fourth quarter are down from the third but were not specified. Gross margin and profit will be under pressure not only from reduced sales but also from start-up costs associated with the company's joint venture, Taisil, in Taiwan. Although not specifically mentioned, we expect that the start-up of MEMC Southwest in Texas and pricing issues in 1997 will subdue reported profits through early 1997.

Given the u p d a t e d third quarter sales expectations, the first nine months of 1996 would see

MEMC sales at about $900 million, u p 44 percent from the comparable period in 1995. Making a reasonable guess at the range for the fourth quarter, Dataquest estimates a revenue growth in the range of 31 percent to 33 percent for the year 1996 over 1995. This is slightly higher growth than our general market revenue forecast grovi^th of 25 percent (see Table 1). This is also consistent w^ith our belief that, through more aggressive joint-venture expansion recently,

MEMC is likely to gain market share slightly in 1996.

Before reviewing additional issues, we think it important to put MEMC in context with the general market in order to understand some specific issues that were addressed in its press release and subsequent conference call.

September 11, 1996 ©1996 Dataquest SEMM-WW-DA-9605

Dataquest Alert Semiconductor Equipment, IVIanufacturing, and IVlaterials Worldwide

Table 1

Silicon W o r l d w i d e MSI and Revenue Forecast, 1995-2001 (Published in July 1996)

Millions of Square Inches of Silicon

Percentage Growth

1995

3,524

21.2

1996

4,102

16.4

1997

4,545

10.8

CAGR (%)

1998 1999 2000

2001 1995-2001

4,944 5,450 6,221

6,924 11.9

8.8 10.2 14.2 11.3

Silicon Wafer Revenue ($M)

Percentage Growth

Source: Dataquest (July 1996)

6,298

37.4

7,867

24.9

8,729

11.0

9,544 11,137 13,359 15,120 15.7

9.3 16.7 20.0 13.2

MEMC has focused its business on supplying advanced logic and DRAM fabs, primarily outside of Japan, with silicon wafers. A significant percentage of the general market is in the power and discrete market, a market on w^hich MEMC has chosen not to focus. The power/discrete wafer market primarily involves wafer sizes less than 150mm, and, as a result,

MEMC's business is generally less concentrated in smaller wafer sizes.

Samsung is a joint-venture partner to Posco-Hiils, and a large part of the production from

Posco-Htils is shipped to Samsung. Further, MEMC is known to be a major supplier of IBM, whose DRAM capacity is run on epitaxial silicon. Because of the IBM connection, Dataquest suspects that the company also supplies Siemens w^ith some epitaxial wafers for DRAM. These are the two companies that currently use epitaxial silicon for DRAM; the rest of the manufacturers use polished wafers. Thus, the inventory correction that MEMC is likely to see would be more balanced between polished and epitaxial wafers than the general market, which would tend to see less weakness in epitaxial than polished.

Does This Announcement Change Our Outlook? ... A Little

For 1996, we expected a marked slowdown in the second half, and that is already factored into our millions of square inches (MSI) growth of 16.4 percent and revenue growth of 25 percent.

MEMC's torrid pace of growth being moderated is essentially in line with our near-term expectations, and our 1996 forecast remains unchanged.

What worries us a little are the assumptions in our forecast for 1997 and 1998. The forecast in

Table 1 assumes that there would be a moderate transition to the 16Mb DRAM, extending into

1998. However, the sooner-than-expected wafer weakness in South Korea, coupled with the fact that both Samsung and Hyundai are no longer producing 4Mb DRAMs appreciably, makes us believe that the transition is happening more quickly than we expected. Instead of forecasting the majority of the transition to occur by mid-1998, we now expect it to be substantially complete by the end of 1997. As a result, we would guide our clients to the following changes in our forecast:

• MSI growth in 1997 is likely to be weaker than our 10.8 percent forecast. It is more likely to be in the range of 4 percent to 7 percent growth, and most of the reduction in growth would come in Asia/Pacific and Japan.

September 11, 1996 ©1996 Dataquest SEMM-WW-DA-9605

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

M MSI growth in 1998 will Kkely be higher than the 8.8 percent growth but likely below the gross value of 4,944 MSI. We would guide an MSI growth range of 11 percent to

12 percent.

• Our current silicon wafer revenue per square inch forecast is flat for 1997 and 1998, with the belief that pricing pressures would be generally offset by the change in mix to

200mm wafers, which sell at a higher price per square inch. This outlook remains unchanged, and, as a result, we would maintain a revenue forecast in line with MSI growth for 1997 and 1998.

• At this point, we would maintain the MSI and revenue forecasts for 1999 through 2001, as our outlook for semiconductor unit growth in the longer term remains good, but this will come under review foUov/ing our updated semiconductor forecast due out in

October.

Why Can Wafer Demand Decline When DRAM Bit Demand Is Up?

An obvious question in today's environment is, "With bit and unit demand u p strongly and weakness in the DRAM market almost purely revenue driven, how can wafer demand decline?" The answer lies in the fact that the DRAM industry migrates product densities, in today's case moving from 4Mb to 16Mb units. The industry's average die size (not the smallest) for the 4Mb DRAM is about 45 to 50 square millimeters, and the 16Mb DRAM is about 80 to 85 square millimeters today. When shipping the equivalent number of bits, there is much less silicon used if the demand is met by 16Mb chips. The industry dynamic in a product transition includes fabs that convert from the older generation to the new, and, in the short term, the actual number of wafer starts can decline. This is the key driving force in our demand forecast model presented in Table 1.

The short-term situation is also driven by the inventories currently held at the DRAM suppliers, w^hich we understand are in the range of two to three months. The inventory correction has not ended, just shifted.

How Do the Shortages Look Now?

Raw Polysilicon Material

In a recent Dataquest Perspective (SEMM-WW-DP-9603, July 22, 1996), we detailed our outlook that polysilicon will still be short later this year into 1997. Recent input on industry inventories have verified continued depletion, and a recent visit to Hemlock Semiconductor verified an intermediate-term very tight market. Silicon manufacturers have efficiency programs, in some cases recycling material, and there has been a supply restarted from Ukraine this year. This pushed back the "depletion to zero" point into the fourth quarter of this year. Will this softening d e m a n d improve the situation? Perhaps, so the industry can barely squeak by, but true capacity relief will not occur until later in 1997, in our opinion. MEMC shared this view in the conference call. Further, if there are any production hiccups in the polysilicon supply, the food chain is w^ired so tight that spot shortages in wafers could reemerge. With luck, that wHl not happen.

September 11, 1996 ©1996 Dataquest SEMM-WW-DA-9605

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

200mm Wafers

In a recent Dataquest Perspective (SEMM-WW-DP-9603, July 22,1996), w e concluded that the

200mm market was entering a short-term (two-year) period of easier supply. This manifested itself in a calculation that the highest 200mm wafer production plant utilization rates are behind us, and the trend in utilization would be down for the next 12 to 18 months. In the longer term, w e expect tightness to reemerge in late 1998 and into 1999. In the shorter term, we would expect pricing for 200mm wafers to ease by 6 percent to 8 percent in 1997, essentially coming back to 1995 pricing. Also, w^e would expect that while the basic programs for capacity expansion continue, the timing of equipment installation for wafer production would be pushed out from 1997 into 1998 or 1999, bringing the market into s u p p l y / d e m a n d balance more quickly. MEMC verified that plant utilization rates were declining but also stated that although no major programs would be delayed, and no immediate capital spending plans would be curtailed, they would likely review the timing of some equipment installation after

1996.

Our 1997 Outlook for Wafer Fab Equipment May Be Optimistic

In July, w e released a forecast for a decline in the wafer fab equipment market of 16 percent.

Given the increased speed at w^hich the w^afer manufacturers are feeling the downturn, our

1997 forecast could be optimistic. We w^ould not be surprised to see a decline w^ell in excess of

20 percent, w h e n all is said and done, as the significance of capital spending cuts from Korean companies is factored.

By Clark Fuhs

September 11, 1996 ©1996 Dataquest SEMM-WW-DA-9605

Dataquest^ ^j^ R T

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

Silicon Market Forecast and Supply/Demand Update—

Growth Picture Intact, Industry Responses Bringing

Balance

The silicon market demand picture is only a little changed from our outlook six months ago.

Dataquest has predicted a slower growth picture for square-inch consumption over the next few years as the industry migrates to the 16Mb DRAM, which uses silicon area more efficiently per bit than the 4Mb. The dynamics of our forecast have been modified slightly to account for the quicker than expected price decline and migration timing. Responses from the silicon suppliers and semiconductor companies to s u p p l y / d e m a n d issues have brought most aspects of the silicon shortage closer to a balanced picture for later in the decade.

This document will be followed by publication of the Semiconductor Equipment,

Manufacturing, and Materials Worldwide 1996 Midyear Forecast, as well as two documents providing more detail of the silicon market outlook. Semiconductor Equipment,

Manufacturing, and Materials Worldwide clients should receive these within the next few weeks.

The Silicon Forecast

Table 1 shows the millions of square inches (MSI) and revenue forecasts for the silicon wafer industry. The forecast for 1996 has been raised slightly because lower DRAM prices are resulting in more memory installed per PC. Because the industry has only started to migrate to the 16Mb DRAM, the short-term dynamic will be an increase in silicon area consumption.

T a b l e 1

S i l i c o n W o r l d w i d e M S I a n d R e v e n u e Forecast, 1995-2001

Millions of Square Inches of Silicon

Percentage Growth

Silicon Wafer Revenue ($M)

Percentage Growth

Source: Dataquest (July 1996)

1995

3,524

21.2

6,298

37.4

1996

4,102

16.4

7,867

24.9

1997

4,545

10.8

8,729

11.0

1998

4,944

8.8

1999

5,450

10.2

2000

6,221

14.2

2001

6,924

11.3

9,544

9.3

11,137

16.7

13,359

20.0

15,120

13.2

CAGR (%)

1995-2001

11.9

15.7

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9604

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

The accelerated schedule of the product migration has caused our growth outlook to be slightly more cautious for 1997 and 1998, but stronger into 1999. Still, the slowest growth year of 1998 is expected to be about 9 percent, primarily because PC vinit demand is expected to remain robust through the decade. Silicon consumption tends to depend more on unit demand than on revenue of the semiconductor industry.

How can the silicon industry grow in the face of a slowing semiconductor market? The answer

Ues in the cyclical dynamic of the semiconductor industry revenue per square inch. Figure 1 shows the historical and forecast semiconductor industry revenue per square inch. What is evident is that it is a cyclical pattern, and Dataquest has concluded that the cycles are driven primarily by the price fluctuations of the DRAM market.

Figure 1

S e m i c o n d u c t o r R e v e n u e per Square Inch of Silicon^ Test Excluded

Dollars

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Source: Dataquest (July 1996)

How Do The Shortages Look?

The silicon shortage has many dimensions, as outlined in Dataquest's November 1995 Focus

Report. Each of them will be briefly reviewed here as an update. The various dimensions include:

• Raw^ polysiHcon material

• 200mm supply and demand

• Smaller wafers, 125mm and below

• Potential for a 150mm shortage later in the decade

July 11, 1996 ©1996 Dataquest

SEMM-WW-DA-9604

"Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

Raw Polysilicon Material

The industry has been consuming more polysilicon than it has produced since 1994, and w^e expect that to continue through 1997, depleting industry reserves later this year. However, several new plant commitments have been made, and polysilicon production will again exceed supply starting in 1998. The polysilicon market is expected to remain tight through 1998, but the threat of shortage after 1997 seems to have disappeared through our forecast horizon. More details w^iU foUow^ in a Dataquest Perspective to foUow in a few weeks.

200mm Wafers

Future demand for 200mm wafers has been affected positively by the semiconductor industry's excess capacity. Why? Semiconductor companies now have some tinne to convert ISOmm fabs to the more cost-effective 200mm wafers. This has added incremental product wafer demand to the industry at 200mm wafers. However, IC manufacturers are being more aggressive in the reduction of test wafer consumption, which has the impact of reducing demand. Suppliers have increased 200mm wafer production commitments because their good cash flow has given them the ability to make plant investments. The net of all this response activity is that 200mm wafers wiU be essentially closely in balance through the decade. Although the supply base has increased commitment, w e do not see evidence of overspending in capacity as yet. More details will follow in a Dataquest Perspective to follow in a few weeks.

Small Wafer Sizes

The silicon industry began "squeezing" the semiconductor industry supply at 125mm and

100mm wafers in order to address ISOmm wafer demand. Semiconductor companies generally did not accept this, and silicon manufacturers have responded by incrementally increasing capacity at ISOmm wafers. This segment of the industry has n o w reinitiated a natural decline in demand, driven by programs of conversion to larger wafers. Although w^e expect the d e m a n d decline to be slow and long, it appears that the shortage at the smaller wafers is starting to dissipate.

Possible 150mm Shortage?

Demand for 150mm wafers is dependent on several factors: how many new ISOmm fabs are being built (there are still several), how fast fabs are converting from smaller wafer sizes to

ISOmm (slowly occurring), and how fast supply and fab conversions ramp to 200mm wafers

(accelerating rapidly). In our November 1995 report, Dataquest outlined a demand point in the year 2000 at which, based on various scenarios, the demand could be in the range of 85 million to 120 million wafers, with our last forecast at 91 million wafers. The current capacity of the industry w^as estimated at 68 miUion per year. As a result primarily of the 200mm r a m p acceleration, the demand picture has eased for ISOmm wafers slightly, and suppliers have also increased capacity incrementally. We no longer see the need to increase ISOmm capacity beyond 1996 for this cycle, and it is unlikely that any new capacity is required for the next.

By Clark Fuhs and Takashi Ogawa

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9604

Dataquest Alert

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Please join us at our SEMICON/West Seminar

Thursday, July 18,1996

Sheraton Palace Hotel, San Francisco

8:00 to 11:30 am

Call 1-800-457-8233 (in U.S.) or 1-805-298-3262 to register

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9604

Dataquest^ ZBI^T

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

Wafer Fab Equipment Market Today—Setting Up for

Resumption of Growtii in 1998

Overcapacity in the DRAM market, created by the massive spending from 1994 through the first part of this year and the normal DRAM product migration, is finally taking its toU, resulting in a faster than expected contraction in the wafer fab equipment market. We expect this contraction to be sharp and relatively deep but slightly shorter than historical norms, around 18 to 24 months. The key reason for the belief that the slowdown will be shorter than normal is the continued robust forecast for PC unit shipment growth through this decade.

This document will be followed by publication of the Semiconductor Equipment,

Manufacturing, and Materials Worldwide 1996 Midyear Forecast, which Semiconductor

Equipment, Manufacturing, and Materials Worldwide clients should receive within the next few weeks.

What Happened?

Although Dataquest has been forecasting a DRAM oversupply-driven capital spending slowdown over the last year to occur in 1997, the swift nature of its arrival still has shock value.

We should point out w h y w e were looking for a slowdown in capital spending.

The industry migration from 4Mb to 16 Mb DRAMs would cause overcapacity even in the face a high bit d e m a n d growth. Why? Die size relationships mean that the average 16Mb DRAM has tw^o to three times more bits per square inch than the 4Mb generation. Because capacity in the industry is built in wafers, this event causes a step-function increase in bit capacity of the industry. This usually happens in a hurry and is triggered by the new generation hitting a critical yield level of about 65 percent, estimated to have been hit by the end of 1995.

The complicating factor, and the reason for the severity of the downturn today, was the artificial d e m a n d created in 1995 by the anticipation of Windows 95, which, taken with the tight supply of memory last year, caused inventories to inflate to enormous levels. This corrected in the final quarter of last year, slightly in advance but essentially on top of the conversion timing.

How Are Semiconductor Companies Responding?

It is very normal in this type of a downturn to get a pocket of companies that will stay and continue to invest in the infrastructure. These companies today are IBM, Texas Instruments, and, on a moderate level, the Korean companies. Japanese companies, which were the

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9603

Dataquest Alert Semiconductor Equipment, Manfacturing, and Materials Worldwide companies that held on in the last cycle for a year or more, have already shut off spending, and these companies as a group will spend slightly less in dollar terms in 1996 than in 1995. Except for Taiwan Semiconductor Manufacturing Co., Chartered Semiconductor, and Macronix, Asian companies are cutting back dramatically, starting in the last couple of months. As a group, the

DRAM manufacturers are responding more rapidly to the situation, cutting spending faster than in previous cycles.

Table I s h o w s the latest Dataquest forecast for w^afer fab equipment by region.

Table 1

Wafer Fab E q u i p m e n t Forecast, 1995-2001 (Millions of U.S. Dollars)

Total Wafer Fab Equipment

Percentage Growth

Americas

Percentage Growth

Japan

Percentage G r o w t h

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Source: Dataquest (July 1996)

1995

19,054

76.6

5,179

62.3

6,352

73.5

2,316

69.1

5,208

102.9

1996

22,309

17.1

6,217

20.1

6,423

1.1

2,642

14.1

7,026

34.9

1997

18,760

-15.9

5,453

-12.3

5,272

-17.9

2,428

-8.1

5,607

-20.2

1998

19,900

6.1

6,163

13.0

5,635

6.9

2,486

2.4

5,616

0.2

1999

25,380

27.5

7,797

26.5

7,011

24.4

3,124

25.7

7,448

32.6

2000

36,219

42.7

10,530

35.1

9,679

38.1

4,253

36.1

11,757

57.8

2001

46,889

29.5

13,105

24.5

11,694

20.8

5,520

29.8

16,570

29.5

CAGR (%)

1995-2001

16.2

16.7

10.7

15.6

21.3

The Concept of "Capacity Trickle" and the Dynamics for Recovery

The industry is experiencing primarily a DRAM oversupply coupled with a product transition.

In order to determine how^ capital spending may recover, it is important to understand how this excess capacity may migrate or "trickle" to other areas in semiconductor capacity.

There are two general blocks of capacity now available. The first comprises old 4Mb DRAM fabs that cannot r u n 16Mb chips. These are limited to two-level metal and are 0.6-micron to 0.8micron technology fabs. Microcontrollers, telecommunications chips, mixed signal, and analog

ICs are quite h a p p y being processed in these fabs. It is likely that most of these fabs in Japan and some in Korea will migrate into this area. The power and discrete chips require specialized processes not found in old DRAM fabs, so these segments are relatively isolated from extraneous supply impacts. The second block of capacity comprises idle or vmderused advanced 16Mb capacity, which is limited today to two-level metal but at 0.4 micron to 0.5 micron. Because these fabs generally lack the process sequences of self-aligned siUcide and three-level metal or more, they cannot be redirected effectively to advanced logic or fast SRAM.

Therefore, they are limited to commodity SRAM, flash, other nonvolatile memory, or a limited span of logic products.

Therefore, Dataquest believes that the first areas of spending recovery wiU be in the advanced logic area, and equipment companies positioned for these markets will have more moderate

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9603

Dataq uest Alert Semiconductor Equipment, Manfacturing, and l\/lateriais Woridwide slowdowns, starting to recover as early as mid-1997. The microcontroller, analog, mixed signal, and telecom chip capacity wiU be next to recover, but probably not tmtil the end of 1997 or early 1998. The DRAM segment, the root cause of the problem, is not expected to resume robust spending until mid-to-late 1998. A quarterly revenue forecast, seasonally adjusted (see

Figure 1) shows that the first sequential quarter of growth wiU by the fourth quarter of 1997.

This scenario again assumes continued growth in the PC vmit and other core semiconductor businesses.

Figure 1

Quarterly S e q u e n t i a l G r o w t h Forecast,. Wafer Fab E q u i p m e n t

Q1/95 Q2'95 Q3'95 CM/95 Q1/96 02/96 Q3'96 QV96 01/97 02*97 03*97 QV97 01/98 02*98 03*98 QV98

Source: Dataquest (July 1996)

By Clark Fuhs, Nader Pakdaman, Calvin Chang, Takashi Ogawa, and Yoshihiro Shimada

Please join us at our SEMICON/West Seminar

Thursday, July 18,1996

Sheraton Palace Hotel, San Francisco

8:00 to 11:30 am

Call 1-800-457-8233 (in U.S.) or 1-805-298-3262 to register

July 11, 1996 ©1996 Dataquest SEMM-WW-DA-9603

Pateguest^ j^ ^ R T

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

Do We Still Expect a Silicon Shortage in the Face of the

Lower Semiconductor Forecast? In a Word ... Yes.

Dataquest's outlook for silicon consumption remains basically unchanged from the forecast released in January 1996, even in the face of our lower semiconductor forecast of 7.6 percent growth for 1996. This Alert will review the reasons why our outlook remains firm. The next formal update to the silicon forecast is planned for release in mid-July, but we do not expect adjustment beyond a "fine tuning." Further, the basic analysis contained in the November 1995 report on the silicon shortage remains vaUd, and fine tuning updates will be made over the next couple of months through the publication of Dataquest Perspectives.

Lower Semiconductor Forecast Driven by Revenue Events, Not Units

The new semiconductor forecast made primarily revenue revisions, as opposed to unit revisions. Table 1 may illustrate this niore clearly.

Table 1

Review of Changes Made to Semiconductor Revenue Growth in 1996 (Billions of

Dollars)

Event or Characteristic

October 1995 Forecast

DRAM Pricing

Yen-to-DoUar Exchange Rate

Other Demand Factors

April 1996 Forecast

Source: Dataquest (May 1996)

1996 Growth Rate (%)

+ 22.1

+ 7.6

Growth Impact (%)

-8.5

-3.0

- 3 . 0

Year 2000 Forecast

331

310

Two points must be made regarding Table 1. First, DRAM prices and exchange rate issues are

revenue effects only and account for nearly 80 percent of the 1996 revision. Unit demand, which is the primary driver for silicon demand, will be relatively imchanged. (SUghtly lower tmit growth in general is expected to be offset by higher DRAM units in 1996 because of demand elasticity.) Second, the longer-term view can be almost entirely explained by the yen/dollar exchange rate change (again, no significant unit forecast decrease).

May 23,1996 ©1996 Dataquest SEMM-WW-DA-9602

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

DRAM Pricing Assumptions Are in Silicon Forecast IVIetliodology

Dataquest's silicon forecast has two things factored into the methodology that have anticipated and modulated the effects of the revision in the semiconductor forecast. The changes were largely anticipated, and therefore we do not expect the silicon forecast to change appreciably.

The DRAM price assumption is part of the silicon forecast methodology. Dataquest actually attempts to forecast semiconductor industry revenue per square inch of silicon. The price of

DRAM is the single most important factor in projecting this industry number, and exchange rates are also taken into account. The new semiconductor forecast wiU simply adjust these assumptions, effectively canceling the effect of the lower revenue inputs. Our silicon squareinch growth forecast for 1996 is 15.6 percent, and w e do not expect this to change appreciably.

Further, the silicon forecast was compiled in December, a couple of months after the previous semiconductor forecast and in the middle of the memory price sHde. Therefore, w e attempted to incorporated additional information in the January silicon forecast that had not been available in the October semiconductor forecast. So, in essence, we anticipated the adjustments a n d built the silicon forecast model on an assumption of a $310 biUion to $315 billion model for semiconductors in the year 2000, with weakness near-term loaded, factoring in a more aggressive price decline in memory.

Polysiiicon Demand Will Still Cause Shortage

With imit demand and forecast models intact, the inevitable shortage in polysiiicon wiU happen. The fact remains that industry inventories of polysiiicon raw material have decreased in 1994 and 1995 because consumption exceeded production. Industry inventories in polysiiicon wiU be exhausted during 1996. PC imit growth would have to decline to about 5 percent in 1996 to avoid this situation (which, by the way, would tend to drive the semiconductor market to an actual decline of 4 percent to 7 percent or more in 1996). Dataquest just does not see this happening.

Since the report on the s u p p l y / d e m a n d situation in November, there have been several announcements for increased polysiUcon capacity. This wiU drive production to exceed consumption starting in late 1997 through 2000. However, the industry operates efficiently with some inventory in the channel, and it wiU take until 1999 for the industry to get back to the two-month inventory level considered to be the minimum for comfortable operation. So we would expect tight (although not short) conditions to exist in the market through 1999. The details of this d e m a n d up da te will be in a forthcoming document.

200mm Wafer Demand Continues to Ramp Faster Than Supply

Regardless of unit demand for semiconductors, the 200mm wafer market is expected to continue to be supply constrained for the rest of the decade. Suppliers are simply behind the power curve in ramping u p production. In the worst case of a PC unit slowdown, more semiconductor companies wiQ migrate their 150mm fabs to 200mm fabs, resulting in a

May 23,1996 ©1996 Dataquest SEMM-WW-DA-9602

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide continued tight 2QDmm market. In the worst case (which we are not forecasting), a PC unit slowdown would mean that the 150mm wafer supply situation only would be eased.

Bottom Line: PC Unit Demand Drives tlie Silicon Shortage

The bottom Hne to the silicon shortage resides with PC imit demeind. PCs drive the need for about one-third of all semiconductor revenue and about one-quarter of silicon in the world, and they continue to be one of the fastest growth segments for semiconductors. As long as this growth trend is intact at the level of a 17 percent compound annual growth rate (and

Dataquest's 1996 PC iinit growth was recently confirmed at 19 percent), silicon demand will remain high and continue to drive a tight market for wafers.

By Clark J. Fuhs

May 23,1996 ©1996 Dataquest SEMM-WW-DA-9602

-4

Dataquest^ Z ^ ^ T

251 River Oaks Parkway • San Jose • CA • 95134-1913 • Phone 408-468-8000 • Fax 408-954-1780

Industry Backlogs Make 1996 Wafer Fab Equipment

Growth of 36 Percent a "Slam Dunk"-What About 1997?

The industry almost hopes that 1995 will not be repeated for a while because the infrastructure of supply is being strained. Backlogs in the equipment industry exploded in 1995, with reports that stepper lead times are over one year. Companies are having trouble filling aU their open positions, and silicon supplies are extremely tight. Yet the engines of demand for semiconductors —PCs and commimications/networking equipment—continue.

This document wiU be followed by publication of the Semiconductor Equipment,

Manufacturing, and Materials Worldwide 1995 Year-End Forecast, which SEMM clients should receive within the next few weeks.

What Are the Trends?

On top of booming growth 1994, global semiconductor capital spending grew another 72 percent during 1995 to $38 biUion, with the wafer fab equipment market growing 67 percent to

$17.9 billion. Anticipated tight capacity and a strong semiconductor market in 1995 mean continued growth into 1996.

North America is showing consistent strength in 1995. North American capital spending is expected to remain strong in 1996 and moderate in 1997 as investment is absorbed, but we expect the North America region to grow at faster-than-market rates as foreign multinationals and foiindry companies invest in capacity in the United States.

Japanese companies are continuing to invest in semiconductor capacity to preserve their market share position in memories, although the strength of the yen in the middle of the year temporarily put a lid on spending enthusiasm. Japan as a region kept pace with the world in investment in 1994, but lagged the market in 1995 as Japanese companies invested more outside

Japan. Lagging investment within Japan is expected to continue throughout the decade.

Japanese companies, however, increased spending during 1995 about 61 percent worldwide to a total of $12.2 billion, second only to Nortiti American companies, with spending of $13.6 billion, and well ahead of Asian companies, at $9.1 billion.

Dataquest has been bullish on the prospects in Europe, and remained so in 1995, although the region came in slightly under expected spending. European companies are a large part of this expansion, aided by strong domestic economies, and major projects by the multinational

January 2,1996 ©1995 Dataquest

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide manufacturers are also contributing. We still see Europe as a significant growth region for spending through the decade.

Following very strong capital investment growth in 1994, spending in the Asia/Pacific region doubled in 1995 as Korean DRAM expansion accelerated (further), foxmdry expansion in

Taiwan, Singapore, and others continued to grow, and new DRAM players entered the scene in

Taiwan. The new projects started in 1995 will continue to consume capital funds in 1996.

Asia/Pacific and the rest of world region (ROW) will continue to be one of the fastest-growing regions through this decade.

Performance of Equipment Segments

Segment growth in 1994 and 1995 was led by DRAM or capital spending-sensitive equipment

(those segments that tend to be bottlenecks or directly influence incremental yield), with steppers, implant, wafer inspection, and factory automation exhibiting significantly strongerthan-market growth. New technology segments such as chemical mechanical polishing (CMP), high-voltage implant, and rapid thermal processing (RTP) were the fastest-growing segments.

We expect no major segment declines in 1996, as capacity additions are broad-based and worldwide, and we expect those segments that lagged in 1995 to be among the leaders in 1996

(etch, tube chemical vapor deposition, and sputtering). Companies wiU continue to invest in • advanced technology in 1996, with the CMP and RTP segments remaining strong.

Is This Cycle Different?

Dataquest believes that the relatively large capacity expansion of 1993 to 1995 (three-year growth of 265 percent) has now exceeded the three-year growth recorded in the 1987-to-1989 expansion. It should be noted, however, that the two periods are different in two key respects.

First, the current period is experiencing accelerated long-term growth for the underlying semiconductor industry, driven by a productivity related PC boom. The PC boom is expected to continue, so we are not overly alarmed about the magnitude of this cycle.

Second, the manufacturing infrastructure is more efficient today, and there is a diminishing return in productivity and yield improvements. This has led to a higher natural capital investment ratio required today than in the 1987-to-1989 period, closer to 22 percent of revenue, on average, being a standard (versus 18 percent in the late 1980s).

However, we also believe that, in 1996, spending will decelerate, starting in the second half of the year, causing a relatively flat spending pattern through 1997 and 1998. Although we continue to beHeve that the cyclical nature of investment in semiconductor capacity will diminish, the PC boom must continue to drive the underlying semiconductor growth strongly enough to dampen the memory component of the cycle. After a flat two-year period, investments should pick up again in 1999.

January 2,1996 ©1995 Dataquest

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

What about 1997? Some Fundamental Capacity Analysis

Point 1: Fab Announcements

Over 130 fabs planned fabs have been announced for the forecast horizon (1995 and after). The density of new fabs, about 35 per year, has not changed significantly, but the announcement horizon has lengthened. A year ago, new fabs were being announced 18 months to 21 months before they were to come on hne. Today, with huge backlogs in equipment companies, annoimcement are made for fabs 24 months to 30 months in the future. Are these fabs real? We beHeve so, because to achieve a $330 billion semiconductor market by 2000, the industry needs to maintain a run rate of 30 to 35 fabs per year. Our concern is that the 1996 and 1997 n m rates are slightly ahead of this figure, and some of the late-1997 and 1998 fabs may be delayed by six to 12 months, softening the quality of backlogs for 1997.

Point 2: DRAIVI Capacity

Although we think the recent talk on DRAM pricing has been overblown (we see continued shortages through 1996), we do believe we are past the "pinch point" in the supply/demand imbalance, and availability of DRAM is beginning improve. We are now viewing 1997 as a year for a slight oversupply of DRAM (based on silicon area capacity), so capital investment in 1997 is likely to cool in this area.

Point 3: Foundry

Does this speU doom for 1997? No, the foimdry industry supply wiU get tighter in 1996, leading to increased spending for logic-oriented capacity over DRAM starting late in 1996 and continuing into 1997. The foundry market will not see relief in capacity untQ the middle of 1998, by our silicon area-based capacity analysis.

Point 4: Discrete and Analog Capacity Shortages

It may come as a surprise to many people that significant capacity is being added at the 1micron level cind above. The boom in electronic equipment and semiconductors has created a demand for power/discrete and analog devices. Normally, companies woidd rely on migration of older capacity toward these areas to meet demand, but the boom in microcontrollers has prevented this. Although the demand for new eqtiipment may be limited to some lithography and older lines of other types of equipment, because semiconductor companies in these segments tend to rely on the used equipnient market, it does mean new fabs. Mainland China is expected to benefit specifically.

Result: Crosscurrents for 1997

DRAM capacity investment wiU be weaker, but very strong logic capacity growth will continue. As DRAM becomes oversupplied, investment is likely to dry up quickly. But the continued tight supply of foundry and logic capacity is expected to produce continued

January 2,1996 ©1995 Dataquest

Dataquest Alert

Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide investment in the United States and Asia in 1997. We expect DRAM investment to decline some

10 percent to 20 percent, while logic investment increases 20 percent to 30 percent.

The Forecast

Tables 1 and 2 shows the regional top Une capital spending and wafer fab equipment forecast through the year 2000. Further details wUl be reviewed in the report mentioned previously.

Table 1

1994

Total Capital Spending

Growth (%)

Percent of

Semiconductors

North America

Growth (%)

Japan

Growth (%)

Europe

Growth (%)

22,037

53.9

19.6

7,182

45.3

6,654

50.8

2,481

43.5

Asia/Pacific

Growth (%)

5,720

76.7

Source: Dataquest (December 1995)

1995

37,994

72.4

25.1

12,169

69.4

9,777

46.9

4,384

76.7

11,665

103.9

1996

49,281

29.7

26.7

1997

50,765

3.0

23.9

16,579

36.2

12,102

23.8

5,293

20.7

15,308

31.2

18,021

8.7

11,766

-2.8

5,521

4.3

15,458

1.0

1998

49,441

-2.6

20.6

17,954

-0.4

11,320

-3.8

5,240

-5.1

14,927

-3.4

1999

54,778

10.8

19.7

20,645

15.0

12,542

10.8

5,968

13.9

15,623

4.7

2000

69,472

26.8

20.9

25,722

24.6

15,986

27.5

7,402

24.0

20,361

30.3

CAGR (%)

1994-2000

21.1

Average =

= 22.1

23.7

15.7

20.0

23.6

Table 2

Wafer Fab Equipment Forecast, 1994-2000 . Dollars)

1994

Total Wafer Fab

Equipment

Growth (%)

10,755

North America

Growth (%)

Japan

Growth (%)

56.4

3,141

47.5

3,668

Europe

Growth (%)

Asia/Pacific

Growth (%)

49.1

1,385

41.6

2,562

95.7

Source: Dataquest (December 1995)

1995

17,911

1996

24,338

66.5

5,262

67.5

5,901

60.9

2,130

53.8

4,618

80.2

35.9

7,512

42.8

7,094

20.2

2,733

28.3

6,998

51.5

1997

24,709

1.5

7,969

6.1

6,445

-9.1

2,590

-5.3

7,705

10.1

1998

23,293

1999

26,542

2000

34,365

-5.7

7,855

-1.4

6,082

-5.6

2,503

-3.3

6,852

-11.1

14.0

9,281

18.2

7,073

16.3

2,927

16.9

7,261

6.0

29.5

11,723

26.3

9,055

28.0

3,655

24.9

9,931

36.8

CAGR (%)

1994-2000

21.4

24.5

16.3

17.6

25.3

January 2,1996 ©1995 Dataquest

Dataquest Alert Semiconductor Equipment, Manufacturing, and Materials Worldwide

By Clark Fuhs, Nader Pakdaman, Calvin Chang, Yoshihiro Shimizu, and Yoshihiro Shimada

January 2,1996 ©1995 Dataquest

DataQuest

^ ^ .

Perspective

S e m i c o n d u c t o r Equipment, Manufacturing, a n d Materials W o r l d w i d e

Market Analysis

3

m

UJ

> a.

O

<

o oc

LU

- J <

IL S

The 1995 Worldwide Wafer Fab Equipment Market: A Record

Year—Competitive Analysis of Regional Dynamics, Industry

Segments, and Company Market Shares

Abstract: The wafer fah equipment experienced a banner year in 1995. How did the industry's

regional markets, product segments, and equipment suppliers weather the tidal wave of industry activity? This document offers a market performance index to better analyze the industry's competitive dynamics and trends.

By George Shiffler and Nader Pakdaman

Introduction

In 1995, estimated front-end equipment manufacturers' total revenue grew a dizzying 77 percent, from $10.8 billion in 1994 to $17.8 billion. Dataquest had anticipated a very strong year in capital spending in 1995, following

57 percent growth in 1994. However, 1995 growth exceeded even the most aggressive predictions we made at the beginning of the year. Our initial forecast was based on healthy growth in ICs driven by continued broadening in semiconductor applications, led by a strong PC market.

The PC market continued to sustain and grow demand for DRAM bits.

With low 16Mb yields and availability, prices for 4Mb not only remained strong but actually inched upward. We had estimated that 16Mb shipments would begin to dominate the market by midsummer. This prediction did not materialize until the end of 1995. As a result, the DRAM market not only experienced rising production but climbing prices, as well. This scenario led to a DRAM market in 1995 that grew by 66 percent over 1994,

Dataquest

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-DP-9608

Publication Date: October 28, 1996

Filing: Perspective

(For Cross-Technology, file in the Semiconductor Regional Markets and Manufacturing binder.)

Semiconductor Equipment, Manufacturing, and Materials Worldwide from an estimated $23 billion to over $42 billion. Excluding the hypergrowth experienced by the DRAM market in 1995, the market for other semiconductors grew by 24 percent. Including the hypergrowth experienced by DRAMs, the market for all semiconductors grew by 37 percent.

Capital spending and equipment investment are driven by semiconductor profitability. Throughout 1995, margins for DRAM products, dominated by

4Mb devices, were at historical highs. In a soon-to-be published Dataquest

Perspective, we will investigate the impact of DRAM pricing and technology transitions on the capital equipment market. In this document, we will focus on a competitive and regional analysis of the 1995 capital equipment market.

1995 Regional Wafer Fab Equipment Markets

Vigorous equipment spending by DRAM producers spurred strong equipment revenue growth in the world's dominant DRAM-producing regions and countries, Asia/Pacific, Korea, and Japan. The regional distribution of wafer fab equipment revenue in 1994 and 1995 is shown in Figure 1. Beginning this year, Dataquest separated the Korean equipment market from the

Rest of Asia/Pacific market. As indicated in the figure, we have developed separate data for Korea and Rest of Asia/Pacific for 1994 and 1995. We are in the process of publishing an extensive Focus Report on the wafer fab equipment market, capital spending, and semiconductor production in

Korea. Figure 2 shows the historical regional distribution of equipment revenue from 1985 to 1995. (In the figure, Asia/Pacific includes both Korea a n d

Rest of Asia/Pacific.)

Asia/Pacific's rising share of equipment consumption accelerated in 1995.

Equipment consumption in this region grew an astounding 103 percent over 1994, raising the region's contribution to worldwide figures from under 24 percent in 1994 to over 27 percent in 1995. For the first time, this translated into an equipment market equivalent in size to the one in

Americas. The surprise is, perhaps, that Korean growth fell below the cumulative average for the region! Whereas Korean equipment sales grew a n impressive 84.5 percent, more strongly than those of any region or country outside Asia/Pacific, Rest of Asia/Pacific grew even more strongly, with its 140 percent expansion. The contribution of Rest of Asia/Pacific to world revenue grew from 8 percent in 1994 to nearly 11 percent in 1995. This growth was not only fueled by capital spending on the part of the foundryrelated manufacturing companies in Taiwan, Singapore, and Thailand, b u t also b y the aggressive entry of Taiwan into the advanced DRAM market. As

Figure 2 shows, 1995 marks the fifth consecutive year in which the Jfeia/

Pacific region's share of worldwide equipment consumption increaled.

Since 1990, Asia/Pacific's share of the worldwide market has a b o u t w p l e d , rising from under 9 percent in 1990 to over 27 percent in 1995. W

1**1

Elsewhere in the world, equipment revenue also realized robust gro^i^th.

The Japanese market grew 73.5 percent in 1995. However, because t h ^ r a t e of growth was just under the average worldwide rate of 76.6 percent^|*jtfipment sales in Japan as a percentage of worldwide revenue declined s l ^ h t i ^ sliding from under 34 percent to an estimated 33.3 percent. As shown'in ^

Figure 2, Japan's contribution to world revenue has now declined for five'!,J consecutive years, falling from a historic high of over 50 percent in 1990 tp^ t

/ ^ • < -

SEMM-WW-DP-9608 ©1996 Dataquest October 28,1996

.#V

Semiconductor Equipment, Manufacturing, and IVlaterials Worldwide

Figure 1

Worldwide Wafer Fab Equipment Market by Region, 1994 and 1995

1994

1995

AsiaJPadfic

Total = 23.8%

Asia/Pacific

Total = 27.3%

Total = $10.8 Billion

Total = $19.1 Billion

Source: Dataquest (October 1996)

Figure 2

Regional Distribution of Worldwide Wafer Fab Equipment Market, 1985-1995

Percentage of Worldwide Market

5 0 -

4 0 -

^

^

^

^

\

\

Americas

Japan

Europe, Middle

East, and Africa

Asia/Pacific

3 0 -

2 0 -

1 0 -

0 -

1

1 1 1 1 1 1 1 1

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

967173

Source: Dataquest (October 1996)

SEIVl!\^-WW-DP-9608

©1996 Dataquest October 28,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide the current 33.3 percent. This decline is not only because of the relative strengthening in spending within Asia/Pacific but also because of

Japanese companies' strong investment outside of Japan.

For Europe a n d the Americas, we estimated similarly strong spending growth. In both cases, however, growth proved to be below the worldwide average. Equipment revenue in Europe (including consumption in Israel) grew nearly 70 percent between 1994 a n d 1995. Robust growth in the

Americas market lagged every other region at just over 62 percent.

European a n d Americas estimated shares of equipment corisumption fell from 12.7 percent to 12.2 percent and from under 30 percent to slightly over

27 percent, respectively. In short, then, except for Japanese consumption, which almost kept pace with the explosive growth realized in 1995,

European a n d Americas equipment consumption growth fell below average, while Korean and Asia/Pacific growth outstripped every other region.

In a global market, where prices of ICs are relatively uniform by product type across all regions, why are spending trends so uneven? We need to look at capital spending as a fraction of semiconductor production revenue to better answer this question. The ratio of capital spending to estimated semiconductor production revenue is shown in Figure 3. In this figure, we see three distinct periods in the evolution of this spending-to-productionrevenue ratio. In the early to mid-1980s, the semiconductor industry was spending, o n average, $0.27 per dollar of revenue. During this time, Japanese manufacturers were spending nearly $0.39 per dollar of revenue. In contrast, U.S. manufacturers were spending $0.22 per dollar of revenue. In the second period, from the mid-1980s to the early 1990s, the worldwide average spending ratio fell to $0.18 per revenue dollar. Although Korean manufacturers spent $1.65 for every dollar of revenue they generated in this interval, semiconductor manufacturers in larger and more mature regions of the world were improving equipment productivity to generate more revenue per dollar spent. In the third period, from the early to the mid-1990s, we see the average at around $0.22 per revenue dollar. In this regime,

Japanese and U.S. manufacturers are estimated to have spent around

$0.20 per revenue dollar, while manufacturers in Korea and the rest of

Asia/Pacific spent over $0.65 for every dollar of revenue. (We should note that in emerging markets government subsidies often play an important role in supporting aggressive manufacturing investment.) In every period, the European spending-to-revenue ratio has fallen below the global average. However, we see European spending ratios rising and approaching the overall average.

We believe that as the IC market and capital spending become more global, every region should more or less approach the approximate spending ratio of the more "mature" regions (that is, Americas and Japan) at about

$0.20 per dollar of revenue. Eventually, higher-than-average spending for

Asia/Pacific should lend itself to smaller ratios than its current aboveaverage figures as this region's advanced manufacturing base becomes more similar to that of its peers in other regions. However these smaller ratios will be for a much larger share of production revenue a n d will thereby translate into larger spending figures. Europe's product mix, led by advanced telecommunications ICs, requires higher spending ratios. This translates into higher spending rates i n Europe, a trend already realized by local and foreign investors in tiiis region.

SEIVIIVl-WW-DP-geOS ©1996 Dataquest October 28,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 3

Capital Spending as a Percentage of Semiconductor Production Revenue

Percentage of Semiconductor Production

"^n

Volume/Unit-Driven Growth

2 8 -

Improving Manufacturing

Productivity

2 6 -

2 4 -

2 2 -

2 0 -

Emerging I

Industry 1

1R_

1 6 -

/ Forecast

^^My' :'' 1 ''

1 4 -

1 i 1 1 1 1 1 1 1 1 1 i 1 1 [ 1

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

967174

Source: Dataquest (October 1996)

1995 Wafer Fab Equipment Segment Growth

In 1995, virtually all major equipment segments posted positive revenue growth. Among the 42 front-end equipment segments that Dataquest tracks in semiconductor manufacturing, we estimate that 39 enjoyed positive growth, 23 posted over 50 percent growth, and nine experienced better than

100 percent growth. Figure 4 illustrates the 1995 revenue growth rates of the equipment segments tracked by Dataquest. As the figure shows, the fastest-growing equipment segment in 1995 was high-voltage implanters.

Incredibly, revenue for this segment grew nearly 350 percent, rising from under $26 billion to over $119 billion.

The only laggards in the 1995 market were rather small segments in lithography (X-ray and projection aligners). The X-ray market is more or less concentrated in advanced R&D centers, while projection aligners are being used more and more in nonsemiconductor production applications.

Equipment Segment Performance Index Defined

There are, of course, many ways to gauge the relative performance of different equipment segments within a year. One could rank equipment segments by revenue growth rate, as in Figure 4, by absolute revenue change, by change in market share, or by some combination of these variables. We believe that relative equipment performance should be based on a measure that evaluates a given segment's performance in terms of that segment's relative impact on tiie overall market. We believe this relative impact is best gauged by looking at how a given segment's growth affects its position in tiie market vis-a-vis other segments. To this end, we have created a marketshare-based performance index defined as follows:

Market performance index = 100 x change in market share x market share weighting

SEMM-WW-DP-9608

©1996 Dataquest

October 28,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 4

Worldwide Wafer Fab Equipment Segment Growth, 1994-1995

High-Voltage Implanter

Chemical Mechanical Polishing

Dedicated LPCVD Reactors

Atmospheric/Subatmospheric Pressure CVD

Post-CMP Clean

CDSEM

Resist Processing (Tracl<)

Spray Processors

Rapid Thermal Processing

Auto Wet Stations

Thin-Film Measurement

Auto Unpatterned Detection

Auto Patterned Detection

Vertical Tube LPCVD

Dedicated PECVD Reactors

Steppers

Other Wafer Fab Equipment

Silicon Epitaxy

High-Density Etch

Low-Density Etch

Dry Strip

Other Clean Process

Factory Automation

Vertical Diffusion

Auto Review and Classification

Medium-Current Implanter

Sputtering

High-Density Plasma CVD

Optical Metrology

Vapor Phase Clean

High-Current Implanter

Manual Detection and Review

Other Process Control

Other Deposition

Contact/Proximity

Direct-Write Lithography

Horizontal Tube PECVD

Horizontal Diffusion

Maskmaking Lithography

X-Ray Aligners

Horizontal Tube LPCVD

Projection Aligners

-50

'^^/M/MMMMMM^^^-^

88888888888^888888

mss^^mm

^msmsmsL

122.0

118.5

03.3

103.2

^msssssssm^. ws^smm,

-^///////////.

102.8

101.4

98.4

V/////M

Ms

96.6

wmmm mssssssm.

i8888888888a

81.2

mssssm^

79.0

78.4

3.5

^88888888^

5.5

9888888888!

66.2

!88888888S

fi2.3

888888888 i8888888Sli

1^88888

46.2

T T T T T

100 150 200 250

300 350 400

1994-1995 Growth (%)

All Wafer Fab

Equipment = 76.6%

967175

Source: Dataquest (October 1996)

October 2 8 , 1 9 9 6

SEMM-WW-DP-9608

© 1 9 9 6 Dataquest

Semiconductor Equipment, Manufacturing, and Materials Worldwide

In this formula, market share weighting equals the arithmetic average of an equipment's market shares for the two consecutive years being studied.

Table 1 presents a ranking of equipment segments in 1995 according to our market-share-based performance index. The index values reported are the mathematical product of the year-to-year difference in each segment's market share (with respect to the overall market) and the arithmetic average of each segment's market share from 1994 and 1995. In effect, the first factor is a measure of annual growth, while the second gauges the importance of a segment in the overall equipment market. Thus, our performance index considers both the market share of each equipment segment and changes in that share that result from revenue growth. As indicated above, we have multiplied the product of the change in market share and market share weighting by 100 to add further resolution to the rankings. By way of example, consider the calculation of our index for steppers:

Index for steppers = 100 x (17.5 -17.0) x (17.5 + 17.0)/2 =

100 x 0.5 X 17.25 = 855.0

(The end result differs from the calculation because of rounding.)

The index will be positive only if a segment experiences growth greater than the industry average. The index heavily penalizes segments with large market shares that fail to keep pace with the industry and thereby lose significant market share. Conversely, the index spotlights segments with small market shares that substantially outpace the industry and so gain significant m a r k e t share. We believe this index eliminates the inherent bias in pure growth-rate-based measures. Those measures tend to single out small, emerging segments that, despite their fast-paced growth, continue to remain small in terms of overall impact on the market. At the same time, our measure eliminates the inherent bias in pure absolute-revenue-change measures. Those measures tend to highlight large established segments that, because of their size, continue to experience large absolute revenue growth b u t that, nonetheless, may be losing market share and hence impact on the market.

Steppers a Top Performer in 1995

According to Dataquest's performance index, steppers proved to be the topperforming equipment segment in 1995. Stepper revenue grew nearly

82 percent between 1994 and 1995, rising from $1.8 billion in 1994 to $3.3 billion i n 1995. Stepper revenue increased from 17 percent of total revenue in 1994 to 17.5 percent in 1995. As calculated above, the value of our index for steppers was an impressive 855.0. Table 1 shows that resist processing equipment a n d auto wet stations finished behind steppers as top-performing segments. Both these segments actually enjoyed better growth than steppers in 1995. However, because of their significantly smaller share of the overall market, their growth had a smaller relative impact on the market.

The worst-performing segment according to this index was sputtering.

Although tills segment posted growth, its performance index was estimated at negative 1,106.2. Sputtering experienced below-industry-average growth of 55 percent in 1995. As a result, sputtering's share of equipment

SEMM-WW-DP-9608 ©1996 Dataquest October 28,1996

o o o

CO

C D

CD

CT) o

"D

I

CO

CT>

O

OO

@

k

C D

C D

CJ>

O o>

r-t-

Oi

X3 cr ct>

C/}

Table 1

Top-Performing Equipment Segments in the Worldwide Wafer Fab Equipment Market, 1995 (Milli

1994

Revenue

1994

Market

Share (%)

1995

Revenue

1995

Market

Share (%)

Revenue

Change

1994-1995

Revenue

Growth (%)

Steppers

Resist Processing (Track)

Auto Wet Stations

Dedicated LPCVD Reactors

1,832.7

695.1

467.6

218.1

17.0

6.4

4.3

2.0

3,331.6

1,412.5

927.6

509.3

17.5

7.4

4.9

2.7

1,498.9

717.4

460.0

291.2

81.8

103.2

98.4

133.5

Low-Density Etch

Atmospheric /Subatmospheidk?-

Pressure CVD

1,139.6

191.4

10.6

1.8

2,032.7

425.0

10.7

2.2

893.1

233.6

78.4

122.0

Vertical Tube LPCVD

Dedicated PECVD Reactors

Auto Patterned Detection

Chemical Mechanical Polishing

CDSEM

Total Wafer Fab Equipment

392.3

467.2

281.0

64.2

153.9

10,787.3

3.6

4.3

2.6

0.6

1.4

100.0

733.2

855.9

535.1

196.7

312.9

19,053.5

3.8

4.5

2.8

1.0

1.6

100.0

340.9

388.7

254.0

132.4

159.0

8,266.1

86.9

83.2

90.4

206.1

103.3

76.6

Note: The 1994-to-1995 performance index equals 100 times the change in equipment share times the average of equipment's 1994 and 1995 share.

Source: Dataquest (October 1996)

Semiconductor Equipment, Manufacturing, and Materials Worldwide revenue fell from 9.4 percent in 1994 to 8.2 percent in 1995. Sputtering's significant market size further accentuated its poor performance index. Perhaps sputtering's relatively poor performance should not surprise u s in a year where over 65 percent of the investments were focused on DRAM capacity increases. This equipment segment is heavily dependent on the logic portion of the industry, for which metal deposition requirements are significantly higher than for memory chips. Therefore, in a DRAM spending spree, this segment's relative growth was below overall market performance.

1995 Wafer Fab Equipment Company Shares: Riding the Wave

It is said that a rising tide lifts all ships. In 1995, almost all equipment companies rose on the tidal wave of industry activity. Of the 120 companies with estimated revenue in 1994,109 experienced positive revenue growth.

Of these, 66 experienced revenue growth in excess of 50 percent, and

27 experienced revenue growth in excess of 100 percent. Not surprisingly, the changing fortunes of most companies were strongly linked to the performance of the equipment segments in which they compete. Companies that gained market share in strong equipment segments rode the crest of the industry's tidal wave. There were, however, exceptions. Some companies proved unable to ride the wave and floundered, losing hard-won market share. Others with excellent performances gained market share in what might be seen as unpromising and relatively weaker equipment segments.

North American and Japanese Companies Remain Necic and Necic in Race for iVIarlcet Share

N o r t h American and Japanese equipment companies remained neck and neck in the race for global market share in 1995. North American companies increased their lead slightly in 1995, but the increase cannot be considered significant. Figure 5 illustrates the revenue shares by company base for 1994 and 1995. Figure 6 illustrates historical company shares from 1985 to 1995.

As shown in Figure 5, North American companies as a group increased their estimated share from 45 percent in 1994 to 46.2 percent in 1995. By our estimates, Japanese companies also increased their share from 44.5 percent in 1994 to 45.1 percent in 1995. As Figure 6 shows, however, the race for dominance remains tight since North American companies returned to challenge the Japanese in 1992. In general. North American companies owe their 1995 edge over Japanese companies to their strong market positions in

1995's lead-performing equipment segments. Among tiie 10 segments highlighted as top performers in Table 1, North American companies as a group hold dominant market positions in seven segments.

The worldwide 1995 market share of European companies continued to hold more or less steady in the range of 6 to 7 percent. As Figure 5 shows, the market share of European companies as whole has been trending downw a r d since 1987. Nonetheless, European companies as a group still hold sizable market positions in Europe as well as the Americas. The worldwide market share of joint-venture companies suffered a sharp decline in 1995, falling from 4.2 to 2.3 percent. This sharp decline is largely related to the dissolution of the Tokyo Electron Ltd./Varian joint-venture agreement.

SEMM-WW-DP-9608 ©1996 Dataquest October 28,1996

10

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 5

Worldwide Revenue Market Share by Company Base, 1994 and 1995

19»4

European Companies (6.3%)

- Joint-Venture Companies (4.2%)

1995

European Companies (6.4%)

Joint-Venture Companies (2.3%)

Total = $10 Billion Total = $17.8 Billion

Note: Market shares exclude estimate of unsurveyed market activity by unidentified companies.

Source: Dataquest (October 1996)

Figure 6

Worldwide Revenue Market Share by Company Base, 1985-1995

Percentage of Worldwide Market

60-

5 0 -

4 0 -

3 0 -

2 0 -

North American

Companies

Japanese

Companies

European

Companies

Joint-Venture

Companies

1 0 -

1 1 1 1 1 1 1 1 1

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Source: Dataquest (October 1996)

967177

SEMM-WW-DP-9608

©1996 Dataquest

October 28,1996

Semiconductor Equipment, Manutacturing, and Materials Worldwide 11

Even without the TEL/Varian joint venture, the market share of joint ventures active in both 1994 and 1995 declined slightly from 2.4 percent in

1994 to 2.3 percent in 1995.

Global Competition between Companies Intensifying

The trend toward increased globalization of the wafer fab equipment market continued in 1995. The term "globalization" refers to the situation where suppliers are less dependent on their indigenous markets. Figure 7 illustrates equipment company dependence on local markets. (Note that equipment companies are classified here according to the region of their base.) As the figure shows, reliance on local markets has fallen dramatically for companies since 1986. Most significant is the dramatic decrease of Japanese companies' reliance on Japan. Japanese companies derived nearly 85 percent of their revenue from Japan in 1985. In 1995, they derived just over

50 percent of their revenue from Japan. Not surprisingly, equipment companies throughout the world have become increasingly dependent on the fastgrowing Asia/Pacific region. Asia/Pacific sales now account for over 30 percent of Japanese companies' revenue. North American and European companies now derive about 25 percent and 18 percent of their revenue from Asia/Pacific, respectively. In addition to their su cesses in Asia/Pacific,

Japanese companies have also successfully pursued both revenue and market share in Americas and Europe. Combined, these regioris now account for about 19 percent of Japanese conipanies' revenue. The share of North

American companies' revenue derived from Japan and Europe has remained fairly constant since 1985. However, North American companies have gained some market share in Japan, especially in recent years, while holding share in Europe. Finally, in addition to relying increasingly on

Asia/Pacific, European companies have grown increasingly reliant on the

Americas region but with little gain in market share. European companies have become decreasingly dependent on Japan, losing market share in the process.

Dataquest believes that there are several interdependent factors responsible for the globalization of the equipment market. We note several factors responsible for this trend:

• Global semiconductor manufacturing: Capital spending has grown increasingly international as the semiconductor market has globalized.

Naturally, wafer fab equipment suppliers have worked to satisfy overseas investment demands. Also, globalization of the semiconductor market h a s significantly raised technology requirements and the cost of semiconductor manufacturing. As a consequence, there has been an increase in the number of semiconductor joint ventures. These circumstances virtually force equipment suppliers across the board to compete internationally and penetrate new overseas markets in order to survive.

• Technology spending requirements: Each new generation of semiconductor devices requires technological enhancement and innovation in wafer fab equipment. Because enhancement and innovation require concentrated and long-term investment in equipment technology development from the equipment industry, the cost to new entrants of entering the industry is raised with each new generation of devices. This "raising of the bar" has greatly slowed entry into the industry and has translated into more global presence for those with access to the capital required for introducing enabling technology.

SEMI\/l-WW-DP-9608 ©1996 Dataquest October 28,1996

12 Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide

Figure 7

Regional Companies' Shares of Respective D o m e s t i c R e g i o n s , 1985-1995

Percentage of Worldwide Market

90-

North American

Companies

Japanese

Companies

European

Companies

r T I I I I ! I \

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

Source: Dataquest (October 1996)

967178

Equipment market concentration: We have observed a marked trend toward increased market concentration among many equipment segments in the wafer fab industry. We believe this trend can be traced to many factors. Semiconductor manufacturers are increasingly moving toward replicating processes d o w n to the most minute features from fab to fab. Process variations a n d innovations have become more and more evolutionary and show smaller variances from one device generation to the next. We believe that this "copy-exact" approach on the part of chipmakers is likely to lead to copy-exact equipment suppliers. This, in turn, is likely to create "installed-base momentum," where IC manufacturers enjoy the benefits of shorter training cycles and more intimate familiarity with the next generation of equipment from the same supplier. One other attractive feature of equipment market leaders is their extensive support structure for spare parts a n d services. In sum, rising manufacturing costs and migration to the next fab or device generation promotes concentration in the equipment market. In turn, concentration among equipment suppliers enhances global presence. There are equipment segments that are exceptions to this trend, such as chemical mechanical polishing

(CMP). Primarily "new" enabling technologies would realize a more even distribution of market share a m o n g smaller companies.

SEI\/IM-WW-DP-9608 ©1996 Dataquest October 28,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

13

Increased Concentration: A Historical Market Perspective

In 1995, the trend toward greater market concentration in the wafer fab equipment industry continued. Figure 8 illustrates the market share of the wafer fab equipment industry's top 10, top 20, and top 40 revenue earners from 1985 to 1995. The market share garnered by the industry's 10 largest revenue earners reached an estimated all-time high of just under 64 percent in 1995. There was some decline in the market share of the 20 largest earners, and the share garnered by the 40 largest revenue earners remained about the same. As Figure 8 indicates, market concentration as gauged by the market shares of the wafer fab equipment industry's top 10, 20, and

40 companies has been steadily increasing since the late 1980s. The top

20 companies appear to have gained share largely at the expense of companies outside the top 40. However, the decline in the difference between the shares of the top 20 and top 40 companies suggests that the industry's second-tier 20 have also surrendered market share to their larger counterparts.

The number of companies Dataquest identifies as market participants has also been declining. These changes in the structure of the industry are likely to have important implications for the future performance of both the wafer fab equipment and semiconductor industries.

Dataquest will soon focus further on this trend, its implications, and its impact in a separate Perspective.

Figure 8

Worldwide Market Shares of Top 10, Top 20, and Top 40 Wafer Fab Equipment

Companies, 1985-1995

50

4 0 -

3 0 -

2 0 -

1 0 -

Percentage Share of Worldwide Market

90.

SO-

TO

60-1

TopIO

Companies

Top 20

Companies

Top 40

Companies

1 1 1 1 1 1 1 1 1

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

967179

Source: Dataquest (October 1996)

SEMM-WW-DP-9608

©1996 Dataquest October 28,1996

14 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Top 20 Companies and Top-Performing Companies: The Big Become Bigger

Applied Materials Inc. topped Dataquest's annual ranking of wafer fab equipment companies by individual company market share for the fourth consecutive year. Also, Applied Materials' growth and market share positioned the company in first place according to the performance index described earlier. Table 2 presents our ranking of the industry's top 20 companies based on 1995 revenue. The table also presents these companies' revenue and rankings for 1994. Table 3 presents a list of the top-performing companies for 1995 based on our performance index. As shown in Table 2,

Tokyo Electron Ltd. and Nikon Corporation once again ranked No. 2 and

No. 3 behind Applied Materials. This marked the fourth straight year in which these three companies have ranked in this order since Applied Materials took the top spot from Nikon in 1992. Elsewhere within the top 20, there were some minor changes in rankings but little change in this group of companies. Among the minor changes in rank. Canon Inc. and Lam

Research Corporation exchanged positions. Alcan Technology rejoined the top 20 after a two-year absence, displacing Materials Research Corporation.

In recent years, the industry's top 20 has proven to be a very select club.

Since 1990, only 22 companies have been members of the club for at least one year. During that time, battles for positions at the rear of the club have proved much keener than competition for positions at its head.

Applied Materials' market leadership really comes as no surprise, given its industry-leading performance, as shown in Table 3. Although several smaller companies posted higher revenue growth rates than Applied

Materials in 1995, the company receives the leading mark for its performance on the basis of our performance index. Applied Materials successfully increased its already substantial market share in 1995. Applied

Materials is a good example of a company that rode equipment segments along the industry's breaking growth wave for outstanding performance.

The company holds strong market positions in four of the 10 equipment segments highlighted as top-performing equipment segments in Table 1.

Applied Materials also did an exceptional job of growing in segments that lagged industry growth. Applied Materials' strong market position in sputtering proved not to be the handicap one might have thought given our conclusion that sputtering was 1995's worst-performing segment. Applied

Materials' sputtering revenue grew almost 75 percent, far outpacing revenue growth for the segment as a whole.

Dataquest's list of top performers may appear to contain some surprising members. Actually, few of the members turned out to be real surprises, given the information in Table 1. Virtually all of the companies singled out as top performers in Table 3 hold significant market positions or derive a significant percentage of their revenue from at least one of the segments highlighted in Table 1.

More surprising, perhaps, would be our list of 1995's poorest performers.

That list would be headed by several prominent players in the top 20 revenue list. Nikon, Varian Associates Inc., and Dainippon Screen h a d performance marks of negative 877.1, negative 256.3, and negative 120.8, respectively. Varian's mark is undoubtedly influenced by the dissolution of its joint-venture agreement with Tokyo Electron Ltd. Still, even considering this, Varian's performance would rank fourth worst in the year because of its below-industry-average growth.

SEMI\/i-WW-DP-9608 ©1996 Dataquest October 28,1996

Semiconductor Equipment, iVIanufacturing, and IVlateriais Woridwide 15

Table 2

Top 20 Wafer Fab Equipment Companies, 1994 and 1995 (Millions of Dollars)

Applied Materials

Tokyo Electron Ltd.

Nikon

Canon

Lam Research

Hitachi

Dainippon Screen

Varian

Kokusai Electric (Including

Bruce Technologies)

ASM Lithography

Silicon Valley Group

KLA Instruments

Eaton (Including Sumitomo/

Eaton Nova)

Novellus Systems Inc.

Tencor Instruments

Anelva

ASM International

Watkins-Johnson

Alcan Technology

Ulvac

All Identified Companies

1994

Revenue

1,480.7

1,066.3

1,027.1

499.9

520.6

387.2

342.5

375.3

306.1

272.7

269.9

255.5

227.2

197.4

210.4

170.1

120.1

122.4

93.6

116.0

9,962.2

1994

Market

Share (%)

14.9

10.7

10.3

5.0

5.2

3.9

3.4

3.8

3.1

2.7

2.7

2.6

2.3

2.0

2.1

1.7

1.2

1.2

0.9

1.2

100.0

1994

Ranking

1

2

3

5

4

6

8

7

9

10

11

12

13

15

14

16

19

18

21

20

-

1995

Revenue

2,904.9

2,098.2

1,675.1

984.1

899.2

643.2

545.3

535.4

523.1

498.3

490.1

459.6

430.0

345.2

309.1

256.3

210.4

202.0

188.3

173.6

17,780.9

1995

Market

Share (%)

16.3

11.8

9.4

5.5

5.1

3.6

3.1

3.0

2.9

2.8

2.8

2.6

2.4

1.9

1.7

1.4

1.2

1.1

1.1

1.0

100.0

1995

Ranking

1

2

3

4

Note: For 1994, revenue for Tokyo Electron Ltd. and Varian includes revenue from joint ventures Varian/TEL and TELA/arian.

Source: Dataquest (October 1996)

10

11

12

13

8

9

5

6

7

14

15

16

17

18

19

20

-

In contrast to Applied Materials, Nikon and Dainippon Screen represent poignant examples of companies that were unable to ride the industry's growth wave. Nikon's stepper revenue, which constitutes the great majority of its overall equipment revenue, grew only an estimated 63 percent, while stepper revenue overall grew 82 percent. Noteworthy is Nikon's performance in 1994; in that year the company outpaced the average growth in the stepper market (82 percent growth over 1993) with an estimated 106 percent increase in revenue. Dainippon Screen's track and auto wet station estimated revenue grew only 67 and 43 percent, respectively, compared to overall revenue growth of nearly 100 percent by each segment. Belowaverage growth combined with large market size perialized the performance of both Nikon and Dainippon Screen in our performance index. It is interesting to note that a similar fate befell many top 20 companies when they were evaluated according to our index.

SEMIM-WW-DP-9608

©1996 Dataquest October 28,1996

o o

cr

CD

PO

C D

CT>

CO m

T)

CD

CT>

<Z>

CO

@

C D

CO

<7>

O

Pi

f—I-

w

£J

Table 3

Top Performing Companies in the Worldwide Wafer Fab Equipment Market, 1995 (Millions of Dol

Applied Materials

Tokyo Electron Ltd,

Canon

Eaton (Including Sumitomo/

Eaton Nova)

Sugai

Steag MicroTech

ASM Lithography

IPEC/Planar

Silicon Valley Group

Alcan Technology

1994

Revenue

1,480.7

1,066.3

499.9

227.2

34.1

18.2

272.7

43.0

269.9

93.6

1994 Share

(%)

14.9

10.7

5.0

2.3

0.3

0.2

2.7

0.4

2.7

1995

Revenue

2,904.9

2,098.2

984.1

430.0

123.7

111.6

498.3

118.9

490.1

188.3

1995 Share

(%)

16.3

11.8

5.5

2.4

0.7

0.6

2.8

0.7

2.8

1.1

Revenue

Change

1,424.2

1,031.9

484.2

202.8

89.6

93.4

225.5

75.9

220.2

94.7

1994-1995

Revenue

Growth (%)

96.2

96.8

96.9

89.3

263.2

513.2

82.7

176.5

81.6

101.2

FSI International

Total Identified Companies

74.1

9,962.2

0.9

0.7

100.0

152.3

17,780.9

0.9

100.0

78.2

7,818.6

105.5

78.5

Notes: The l994-to-1995 performance index equals 100 times ttie change in equipment share times the average of equipment's 1994 and 1995 share.

Tokyo EteclfOR Ltd.'s 1994 revenue includes revenue for Varian/TEL joint venture.

Source: Dataquest (October 1996)

Semiconductor Equipment, Manufacturing, and Materials Worldwide 17

Dataquest Perspective

The strength of the DRAM market in 1995 translated into strong growth for wafer fab equipment spending in Asia/Pacific and Japan. Among equipment segments, steppers proved to be the industry's top performer in terms of impact on market growth. Several clean process a n d chemical vapor deposition (CVD) equipment segments were also top performers on this basis. As a consequence, companies with strong positions in these segments performed especially well in 1995.

North American companies as a group held their worldwide market share lead over their Japanese counterparts based on their strong positions in topperforming equipment segments. A m o n g individual companies. Applied

Materials stretched its market share lead over No. 2 rival Tokyo Electron

Ltd., which closed 1995 with a very strong performance.

Finally, 1995 saw a continuation of the trend toward greater market concentration in the wafer fab equipment industry. The combined market share of the industry's 10 largest companies rose for the third consecutive year.

Although the combined shares of the industry's 20 and 40 largest companies declined somewhat, the industry's top 40 companies now account for about seven-eighths of estimated individual manufacturer revenue.

Supporting Tables

Tables 4 through 9 provide the data used to create the figures featured in this Perspective.

Contributing analysts (in alphabetical order): Calvin Chang, Ron Dornseif, Calvin

Chang, Akiko Nakayama, Takashi Ogawa, Yoshie Shima,and Yasumoto Shimizu.

SEMM-WW-DP-9608 ©1996 Dataquest October 28,1996

o o o cr

CD tS3

CD

CD

en

-o

I

CD

0 3

O

@

CO

CD

cy>

(U

Table 4

Worldwide Wafer Fab Equipment Market, Regional Distribution, 1985-1995 (Millions of Dollars)

1985 1986

1987

1988 1989 1990

1991 1992

Americas

Revenue

Percentage of World Market

Japan

Revenue

Percentage of World Market

1,258.6

37.5

1,422.8

42.4

1,077.6

39.7

1,018.2

37.5

1,104.4

35.2

1,276.7

40.7

1,535.0

30.8

2,269.8

45.5

1,656.9

27.6

2,813.3

46.8

1,589.1

27.1

2,995.9

51.0

1,518.5

25.3

3,011.6

50.2

1,575.7

30.9

2,097.9

41.1

2

2

Europe, Middle East, and Africa

Revenue

Percentage of World NTarket

Asia/Pacific

466.9

13.9

454.1

16.7

526.4

16.8

663.2

13.3

721.0

12.0

764.3

13.0

641.1

10.7

641.4

12.6

Revenue

Percentage of World Market

Korea

Revenue

Percentage of World Market

Other Asia/Pacific

208.9

6.2

-

-

163.5

6.0

-

-

229.4

7.3

-

-

518.6

10.4

-

-

820.1

13.6

-

-

521.6

8.9

-

-

831.8

13.9

-

:•'

783.3

15.4

-.

• •

1

Revenue

Percentage of World Market

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

Worldwide Market 3,357.2 2,713.4 3,136.9

4,986.6 6,011.3 5,870.9

6,002.9 5,098.3 6

Note: Asia/Pacific market totals reflect sum of totals for Korea and other Asia/Pacific, when available. No breakout of Korea versus other Asia/Pacific is a

Source: Dataquest (October 1996)

CO

C3

-X3

I

CO

0 3

CZ>

OO

@

<£>

C D

C D

a

°i

o>

. o c=

CD

CO

o

o cj-

OO

CO

CO

(T>

Table 5

Worldwide Semiconductor Capital Spending as a Percentage of Worldwide Semiconductor Produc

1982-2000 (Millions of Dollars)

Capital

Spending

Semiconductor

Production

Percentage of

Capital

Spending

1982

4,530

15,621

29.0

1983

5,666

21,537

26.3

1984

8,107

28,825

28.1

1985

7,299

28,132

25.9

1986

5,129

34,102

15.0

1987

6,505

41,833

15.5

1988

9,748

54,987

17.7

1989

12,331

59,184

20.8

1990

13,230

59,328

22.3

1991

13,145

64,453

20.4

1992

11,599

70,460

16.5

1993

14,333

87,535

16.4

1994

22,085

112,464

19.6

1995

38,308

153,084

25.0

19%

45,344

164,264

27.6

4

18

Notes: Semiconductor production revenue includes merchant plus captive revenue. Numbers for 1996 to 2000 are based on Dataquest's 1996 midyear f

Source: Dataquest (Ociober 1996)

Table 6

Worldwide Wafer Fab Equipment Market, Companies' Market Shares, 1985-1995 (Millions of Dolla

1985 1986 1987 1988 1989 1990

1991 1992

North American Companies

Revenue

Percentage of Identified Market

1,507.6

55.0

1,154.3

51.5

1,221.5

45.7

1,878.5

42.9

2,094.6

39.1

1,977.4

37.4

2,010.7

36.8

1,985.4

44.0

2

JapEinese Companies

Revenue

Percentage of tdentified Market

European Companies

Revenue

Percentage of Identified Market

Joint-Venture Companies

840.2

30.7

240.5

8.8

742.9

33.1

261.2

11.7

1,022.4

38.2

322.4

12.1

1,932.5

44.1

417.9

9.5

2,561.1

47.9

486.9

9.1

2,655.1

50.2

464.7

8.8

2,842.7

52.1

405.2

7.4

2,011.8

44.6

337.5

7.5

2

Revenue

Percentage of Identified Market

Identified Companies' Revenue

Ottier Companies' Revenue

All Companies' Revenue

151.9

5.5

2,740.2

617.0

3,357.2

83.3

3.7

2,241.7

471.7

2,713.4

108.7

4.1

2,675.0

461.9

3,136.9

149.2

3.4

4,378.1

608.5

4,986.6

209.7

3.9

5,352.3

659.0

6,011.3

195.3

3.7

5,292.5

578.4

5,870.9

199.7

3.7

5,458.3

544.6

6,002.9

172.7

3.8

4,507.4

590.9

5,098.3

Note: Totals for identilied companies reflect estimates based on individual company survey data; totals for other companies reflect Dataquest estimates

Source; Dataquest (October 1996)

6

6

O o

r-^

o

ex

CD

—1 ro c»

CD

CD

0 5

Crt

-o

I

C D

cr>

C D

CXJ

@

_^

C D

CO

CJ>

o

tv>

r-t-

a>

X3 c:

CD

CO

Table 7

Regional Companies' Shares of Respective Domestic Regions, 1985-1995

1985

1986

1987 1988

1989

Norlh American Companies

Americas

World

Percentage of Domestic Riqgidn

Japanese Companies

)apan

World

Percentage of Domestic Region

{European Companies

Europe, Middle East, and Africa

World

Percentage of Domestic Regioit

Source: Dataquest (October 1996)

880.6

1,507.6

58.4

706.2

840.2

84.1

104.9

240.5

43.6

719.8

1,154.3

62.4

586.8

742.9

79.0

122.3

261.2

46.8

711.2

1,221.5

58.2

782.6

1,022.4

76.5

152.8

322.4

47.4

1,015.4

1,878.5

54.1

1,472.9

1,932.5

76.2

184.5

417.9

44.1

1,043.1

2,094.6

49.8

1,877.1

2,561.1

73.3

201.0

486.9

41.3

1990

1,010.1

1,977.4

51.1

2,026.5

2,655.1

76.3

193.0

464.7

41.5

1991

927.2

2,010.7

46.1

987.5

1,985.4

49.7

2,022.3

2,842.7

71.1

1,347.2

2,011.8

67.0

139.0

405.2

34.3

1992

119.7

337.5

35.5

Table 8

Worldwide Market Shares of Top 10, Top 20, and Top 40 Wafer Fab Equipment Companies, 1985-19

1985

1986 1987 1988 1989 1990 1991 1992

Top 10 Companies

Revenue

Market Share (%)

Top 20 Companies

Revenue

Market Share (%)

Top 40 Companies

Revenue

Market Share (%)

Identified Companies

Revenue

Number of Companies

Source: Dataquest (October 1996)

1,464.2

53.4

1,916.4

69.9

2,163.6

79.0

2,740.2

125

1,079.4

48.2

1,462.9

65.3

1,681.0

75.0

2,241.7

137

1,298.4

48.5

1,757.2

65.7

2,015.3

75.3

2,675.0

146

2,260.4

51.6

3,036.1

69.3

3,434.8

78.5

4,378.1

156

2,780.2

51.9

3,739.9

69.9

4,252.0

79.4

5,352.3

154

2,795.7

52.8

3,728.4

70.4

4,141.0

78.2

5,292.5

159

2,987.1

54.7

3,935.7

72.1

4,332.8

79.4

5,458.3

164

2,498.1

55.4

3,364.3

74.6

3,719.3

82.5

4,507.4

137

3

4

5

6

& l\5

OO

CD

CD

CO

C O

-a

I

C D

C D

O

OO

@

CO

CO

<y>

CO sr

J 3

CD

CO

Table 9

1995 Worldwide Wafer Fab Equipment Market Summary Statistics (Millions of Dollars)

Lithography

Contacty Proximity

Profectian Aligners

Steppers

Direct-Write

Lithography

Maskmaking

Lithof^raphy

X-Ray Aligners

Worldwide Wafer Fab Market

1994

2,036.5

27.7

50.8

1,832.7

34.9

1995

3,531.5

33.9

35.8

3,331.6

38.0

Change

(%)

73.4

22.2

-29.5

81.8

8.8

Performance

Index

-643.2

-1.7

-9.3

855.0

-3.3

Americas

25.6

16.2

31.0

25.5

10.5

1995 Regional Distribution (%)

Japan

37.6

12.9

38.8

37.6

67.9

Europe,

Middle

East, and

Africa

10.7

40.8

26.3

10.1

19.7

Total

Asia/

Pacific

26.1

30.1

3.9

26.8

1.8

Korea

16.3

Other

Asia/

Pacific

9.8

6.5

0

16.9

0

23.6

3.9

9.9

1.8

1995

North

American

Companies

8.5

0

69.3

6.4

0

Jap

Comp

79.1

11.2

82.0

10.2

3.7

-9.3

-17.6

-0.4

28.9

100.0

37.6

0

12.9

0

20.6

0

12.2

0

8.4

0

71.2

27.6

Resist Processing (Track)

695.1 1,412.5 103.2 672.1 27.2 30-2 14.5 28.1 14.5 13.6

23.4

Etch and Clean

Auto WetStatior\3

Spray Processors

Vapor Phase Clean

Post-CMP Clean

Other Clean Process

Chemical Mechanical

Polishing

Dry Strip

Dry Etch

Low-Density Etch

1 High-Density Etch

2,548.5

467.6

60.6

8.8

21.7

121.4

64.2

4,723.8

927.6

122.8

12.6

47.4

206.1

196.7

212.5

1,591.6

1,139.6

452.0

368.8

2,841.8

2,032.7

809.1

85.4

98.4

102.8

42.2

118.5

69.8

206.1

2,826.6

245.7

5.0

-0.1

1.1

-4.8

35.5

73.5

78.6

78.4

79.0

-6.8

238.2

110.9

23.7

26.9

25.6

48.9

28.8

60.7

36.1

68.7

32.4

30.7

21.2

27.4

15.7

31.7

17.4

24.4

22.5

22.1

23.5

32.5

34.8

31.5

43.0

12.2

9.3

22.7

12.1

15.0

8.9

9.4

9.0

13.5

14.4

11.2

28.6

34.4

7.1

31.7

8.7

23.4

4.6

34.1

29.3

32.0

22.3

16.5

19.2

0.7

1.5

4.9

8.2

2.0

22.0

17.5

17.5

17.7

12.0

15.1

6.4

30.2

3.8

15.1

2.5

12.1

11.7

14.6

4.6

54.9

18.3

89.5

76.8

70.7

33.2

90.3

45.8

65.3

69.2

55.5

O o

CO

CO

CO

CJ>

I

CO

0 5

CD

CO

@

C D

CO

CT> o

Q } sr

J 3

Table 9 (Continued)

1995 Worldwide Wafer Fab Equipment Market Summary Statistics (Millions of Dollars)

Etepositiart

CVD

TubeCVD

Horizontal Tiibe LPCVD

Vertical Tube LPCVD

Horizontal Tlibe PECVD

NontubeCVD

APCVD/SACVD

High-Density Plasma

CVD

Dedicated LPCVD

Reactors

Dedicated PECVD

Reactors

Sputtering

Silicon Epitaxy

Otiier Deposition

Worldwide Wafer Fab Market

1994

2,553.2

1,327.0

441.5

23.3

392.3

26.0

885.4

191.4

8.7

1995

4,479.9

2,582.9

779.6

18.6

733.2

27.9

1,803.3

425.0

13.1

Change

(%)

75.5

94.6

76.6

-20.2

86.9

7.3

103.7

122.0

50.6

Performance

Index

-368.8

1,622.5

-0.5

-1.9

79.2

-1.8

1,110.1

91.3

-0.1

Americas

29.7

26.8

19.7

44.6

19.2

17.9

29.9

24.1

84.7

1995 Regional Distribution (%)

Japan

31.2

30.3

28.9

23.1

28.9

Europe,

Middle

East, and

Africa

13.1

13.1

13.4

28.0

11.7

Total

Asia/

Pacific

26.0

29.7

38.1

4.3

40.2

Korea

16.9

21.1

28.3

0

30.1

32.3

31.0

31.7

12.2

46.2

13.0

10.2

3.1

3.6

26.1

34.0

0.0

1.1

18.0

21.1

0.0

Other

Asia/

Pacific

9.1

8.6

9.7

4.3

10.1

2.5

8.1

12.9

0.0

North

American

Companies

64.0

60.5

13.4

41.4

13.3

0

80.8

62.1

93.9

1995

Jap

Comp

218.1

467.2

1,011.5

114.1

100.7

509.3

855.9

1,566.5

206.7

123.7

133.5

83.2

152.8

71.0

54.9

81.2

22.9

-1,016.2

3.0

-22.5

31.6

31.0

32.5

42.8

32.3

36.2

27.7

32.5

30.2

34.1

6.4

18.5

11.6

21.3

16.2

25,7

22.7

23.4

5.7

17.4

20.2

15.4

12.8

1.3

7.7

5.5

7.3

10.6

4.4

9.7

77.8

91,0

71.4

57.8

NM

Tlierma! NondeposiBon

Rapid Tliermal

Processing

Diffusion

Vertical Diffusion

Horizontal Diffusion

566.5

76.2

490.3

421.5

68.8

926.9

153.5

773.4

700.8

72.7

63.6

101.4

-195.9

7.5

57.7

66.2

5.6

-209.2

-87.2

-13.1

26.4

45.5

36.4

23.9

22.6

20.0

47.3

38.9

39.3

34.8

11.0

161

10.0

9.5

15.6

26.2

14.4

15.7

2.7

28.5

31.2

2.3

18.3

20.2

0.0

10.5

11.8

10.2

11.1

2.3

21.1

58.5

13.7

12.2

28.1

Ion Implantation

Medium-Current

Imp la titer

659.4

241.6

1,053.2

383.8

59.7

58.9

-340.3

-47.9

27.7

27.8

31.4

30.3

15.2

15.6

25.8

26.4

14.7

14.1

11.0

12,3

78.1

78.0

a

T)

I

t o

Oi

<z> oo

@

CD

0 3

CD

a cu

*-+

Oi

r i

o o f-fr o

C T

CD l>0

OO

CO

CO

a>

Table 9 (Continued)

1995 Worldwide Wafer Fab Equipment Market Summary Statistics (Millions of Dollars)

High-Current Implanter

High-Voitage Implanter

Worldwide Wafer Fab Market

1994

391.2

26.6

1995

550.3

119.2

Change

(%)

40.7

348.2

Performance

Index

-240.6

16.5

Americas

29.2

20.0

1995 Regional Distribution (%)

Japan

Europe,

Middle

East, and

Africa

Total

Asia/

Pacific Korea

32.1

32.1

15.5

12.2

23.2

35.7

11.3

32.5

Other

Asia/

Pacific

11.9

3.2

1995 M

North

American

Companies

79.7

71.2

Jap

Comp

Process Control

Optical Metrology

CDSEM

Thin-Film Measurement

Auto Patterned Detection

Auto Review antt

Classification

Manual Detection and

Review

AutoUnpatterned

Detection

Other I'rocess Control

1,103.3

66.8

153.9

99.7

281.0

119.6

54.8

56.4

271.1

1,853.5

97.6

312.9

196.1

535.1

194.2

68,9

110.6

338.3

68.0

46.2

103.3

96.6

90.4

62.3

-498.2

-6.0

33.1

10.2

55.0

-9.5

25.6

96.2

-6.4

3.2

24.8 -158.2

28.3

27.7

23.1

36.2

24.9

27.0

35.2

37.2

34.7

26.3

38.8

39.3

37.2

35.2

22.5

30.9

34.1

35.0

11.5

12.3

10.4

14.5

8.8

10.5

19.7

11.4

13.6

25.0

22.8

31.7

23.0

27.5

23.2

20.6

19.3

20.6

14.2

20.7

15.2

14.8

15.6

17.6

9.7

9.2

9.5

11.0

10.1

11.1

10.8

2.1

16.5

8.2

12.0

5.6

65.5

88.9

16.8

%.3

87.9

22.9

0

84.1

NM

Factory Automation 412.0

686.0 66.5 -81.2

OUier Wafer Fab

Equipment

212.9 386.1 81.3 10.5

Total Wafer Fab

Equipment

10,787.3

19,053.5 76.6

Note: Totaf Asia/PacKIc equals Korea plus other Asia/Pacific.

NM = Not meaningful

Source: Dataquest (October 1996)

NM

17.8

34.3

27.3 30.3

27.2

33.3

6.6 41.4 27.1 14.3

13.3 29.1 19.2

10.0

12.2 27.3 16.5 10.8

NM

NM

46.2

24 Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

#

For More information...

George Shiffler, Market Research Analyst (408) 468-8196

Internet address [email protected]

Nader Pakdaman, Principal Analyst (408) 468-8417

Internet address [email protected]

Via fax (408) 954-1780

^'' The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness.

^ ^ It does not contain material provided to us in confidence by oiu- clients. Reproduction or disclosure in whole or in

IJOTO/^I I^CT part to other parties shall be made upon the written and express consent of Dataquest.

* ^ * * * * * ^ " s S r ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and Materials

Woridwide

Market Analysis

CVD Market Analysis

Abstract: The worldwide chemical vapor deposition (CVD) equipment market shot up 95

percent in 1995, achieving U.S.$2.58 billion in sales, registering the highest growth recorded since Dataquest began tracking the equipment market in 1982. Japan remained the largest market, and Asia/Pacific rose to prominence. Americas region companies extended their dominance, maintaining more than 60 percent of the worldwide CVD market share.

By Calvin Chang

The CVD Market in 1995

For the chemical vapor deposition (CVD) equipment market, 1995 was a banner year, with numerous sales records shattered. Explosive sales led to revenue milestones that read like a long string of multiples: worldwide CVD equipment sales almost doubled in 1995, nearly tripled in two years, and close to quadrupled in three years. The CVD equipment segments also outperformed the overall wafer fab equipment market, which grew 75 percent in 1995. Driven by a continued ramp in 16Mb/64Mb DRAM capacity investment, atmospheric pressure CVD/subatmospheric CVD

(APCVD/SACVD) and low-pressure CVD (LPCVD) reactors turned in stellar sales growth performance that exceeded more than 120 percent. Also impressive, the plasma-enhanced CVD (PECVD) reactor and LPCVD tube markets turned in 83 percent and 81 percent growth, respectively.

DataQuest

Program: Semiconductor Equipment, Manufacturing, and iVIateriais Worldwide

Product Code: SEIVIM-WW-DP-9607

Publication Date: September 30,1996

Filing: Perspective

(For Cross-Teclinoiogy, file in tiie Semiconductor Regional i\/larl<ets and IVIanufacturing binder)

Semiconductor Equipment, IVJanufacturing, and l\/laterials Worldwide

Regional Markets: Japan Remains the Largest, while Asia/Pacific Rises to

Prominence

Figure 1 shows the change in the regional markets' percentage contribution to worldwide CVD equipment sales from 1993 to 1995. Asia/Pacific has grown progressively in share, while the other three regions have declined during the three-year period. Consuming nearly 30 percent of world's total

CVD shipments, Asia/Pacific is now the second-largest regional market.

Japan is still the largest market, with a shade more than 30 percent of the world's CVD total. This compares with the 50 percent share that Japan enjoyed only four years ago. Asia/Pacific's rise in CVD equipment sales resulted from increases in both tube CVD and nontube CVD sales in the region. As Figure 2 illustrates, Asia/Pacific has become the world's largest tube CVD market during 1993 through 1995, principally at the expense of the Japanese market. Large DRAM investments in Korea, Taiwan, and

Singapore in recent years have led to Asia/Pacific's garnering the largest share, 38.1 percent of 1995 worldwide tube CVD sales. At the same time,

Asia/Pacific sales of nontube CVD equipment, which includes APCVD,

PECVD, LPCVD, and high-density plasma (HDP) CVD reactors, also rose as a percentage of the worldwide total. As Figure 3 shows, Asia/Pacific's share of nontube CVD consumption increased to more than 26 percent of the world's total from less than 20 percent in 1993. Japan and Americas are, however, larger markets, each with about 30 percent of the world's total nontube CVD sales.

Figure 1

Regional Contribution to the Woridwide CVD Equipment Market

Percentage Share

35 i r

30

25

m

1993

H

1994

1995

20

15

10

Asia/Pacific Americas

Source: Dataquest (September 1996)

SEiviivi-ww-DP-geo?

Japai

©1996 Dataquest

Europe

September 30,1996

20

15

10

5

0

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 2

Regional Contribution to the Worldwide Tube CVD Equipment Market

Percentage Share

40-ff

35

30'

m

1993

H

1994

1995

25

Asia/Pacific

Americas Japan Europe

Source: Dataquest (September 1996)

Figure 3

Regional Contribution to the Worldwide Nontube CVD Equipment Market

Percentage Share

35 tl

30

25

1993

^

Bgfll

1994

H

1995

20

15

10

968439

0-1'

Americas Japan Europe Asia/Pacific

96M90

Source: Dataquest (September 1996)

SEMM-WW-DP-9607

There was little change in 1995 in the ownership of the CVD eqtiipment market by company base. As shown in Figure 4, Americas vendors remained

©1996 Dataquest September 30,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide the dominant suppliers of CVD tools, with 60.5 percent of the worldwide market, just a bit less (0.2 percent) than the previous year. Japanese companies' market share rose to 28.5 percent from 25 percent in 1994. This is largely the result of Tokyo Electron Ltd.'s absorbing the Varian/TEL joint venture, which caused the joint ventures' share to drop to 5.1 percent. Joint ventures, considered in the past a favorable business model for sharing costs and complementary resources, are becoming less popular in the equipment industry. Greater support of and access to the customers is requiring suppliers to maintain close control of their product development, marketing, and services. Figures 5 and 6 show the change in market share by company base for tube CVD and nontube CVD, respectively. In tube CVD, which was

30 percent of the total CVD market, Japanese suppliers TEL and Kokusai remained dominant, with a 72.7 percent share of the tube business, compared to 66.2 percent in 1994. The U.S. suppliers' share also rose, to 13.4 percent, an increase of 3 percent, courtesy of Silicon Valley Group Inc. (SVG) and Semitool Inc. In the U.S.$1.8 billion nontube market, U.S. vendors are still the preeminent suppliers but have seen their share slip from 85.8 percent to 80.8 percent. The lost share was won by TEL and Amaya, whose tungsten silicide CVD and APCVD, respectively, contributed to Japanese companies' attaining 9.3 percent of the nontube market. European suppliers are prindpally represented by ASM International, whose tube and nontube sales contributed to Europe's single-digit share of the market. Asia/Pacific equipment system suppliers, all of them in Korea, have a negligible share of the world market.

Figure 4

Worldwide CVD Market Share by Company Base

Worldwide Market Percentage Ownership

7011

1994

60

1995

50

40

30

20

10

0

Americas Companies Japanese Companies European Companies Joint Ventures

96W91

Source: Dataquest (September 1996)

SEMM-WW-DP-9607

©1996 Dataquest

September 30,1996

Semiconductor Equipment, l\^anufacturing, and IVIaterials Worldwide

Figure 5

Worldwide Tube CVD Market Share by Company Base

Worldwide Market Percentage Ownership

Americas Companies Japanese Companies European Companies Joint Ventures

96^492

Source: Dataquest (September 1996)

Figure 6

Worldwide Nontube CVD Market Share by Company Base

World Market Percentage Ownership

yu-

80-

70-

'm^

M 1994

H 1995 bU-

50-

40-

30-

20-

10-

0-

mm^^m m.jw.m^m^

^ —

Americas Companies Japanese Companies European Companies Joint Ventures

9664^3

Source: Dataquest (September 1996)

SEMIVI-WW-DP-9607

©1996 Dataquest September 30,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Company Rankings

Table 1 shows the individual CVD equipment company rankings based on the worldwide 1995 CVD revenue. The table indicates each company's performance in 1995 compared to 1994, as well as individual CVD market segment results.

Table 1

1995 Worldwide Chemical Vapor Deposition Equipment Company Ranking

(Millions of U.S. Dollars)

Applied Materials

Tokyo Electron Ltd.

Novellus

Kokusai Electric

Watkiris-Johnson

ASM International

Alcan/Canon/Quester

Silicon Valley Group

Genus

Amaya

Ulvac

Lam Research

Bruce Technologies

E.T. Electrotech

Materials Research Corp.

Others

14.9

56.2

Total

2,582.9

1994-1995 Segment

Growth (%)

Source; Dataquest (Septerr^r 1996)

CVD

Revenue

849.9

393.3

340.2

223.2

202.0

136.5

115.0

76.5

46.9

42.3

36.7

17.7

16.6

15.0

1995

Share (%)

Change(%)

1994-1995

Tube CVD

Horizontal

LPCVD

Vertical

LPCVD

32.9

15.2

13.2

8.6

1.7

0.6

-1.2

-0.9

0

0

0

0

7.8

5.3

4.5

3.0

1.8

1.6

1.4

0.7

0.6

0.6

0.6

2.2

100.0

-1.4

0.1

-0.4

-0.1

0.1

1.3

0.2

-0.1

-0.1

-0.3

0.2

-2.8

-

0

0

0

6.8

0

0

2.5

0

6.6

0

0

2.2

18.6

0

319.1

0

223.2

11.7

63.3

0

69.7

0

0

6.0

0

10.0

0

0

30.1

733.2

APCVD

Reactor

61.9

0

202.0

0

0

0

115.0

0

0

42.3

0

0

3.7

425.0

0

0

0

-20

87

122

Nontube CVD

LPCVD

Reactor

240.9

74.1

100.1

0

0

PECVD

Reactor

HDPCVD

Reactor

544.3

0

240.1

0

0

2.7

0

0

46.9

0

0

0

22.6

45.3

0

0

0

0

0

0

Q

Q

P

b

0

0

d

Q

8.1

0

9.6

9

0

14.9

1.7

509.3

15.0

0

0

0

11.2

855.9

0

0

0.8

13.1

134 83 51

Applied Materials Inc.

Applied Materials solidified its CVD leadership in 1995, increasing its CVD equipment market share by 1.7 points to 32.9 percent. Applied Materials continued to be the No. 1 supplier in PECVD and LPCVD reactors. PECVD is the largest segment in CVD, with U.S.$856 million sales in 1995, and

Applied Materials is the undisputed leader, with nearly two-thirds of the market. The rapid growth of the Asia/Pacific PECVD reactor market and

Applied Materials' nearly imchallenged position in the region contributed to the company's dominance in worldwide PECVD shipments. In the U.S.$0.5 billion LPCVD reactor market. Applied is the leader in CVD tungsten and shares the top position in tungsten silicide with TEL.

Another noteworthy Applied Materials achievement has been its quick penetration into new markets. In a few years since its SACVD introduction,

Applied Materials has garnered 15 percent of the APCVD/SACVD market.

A number of factors contributed to the success of the company's SACVD

SEMM-WW-DP-9607 ©1996 Dataquest

September 30,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide program. First, its SACVD was developed on the well-established Precision

5000 platform. Precision 5000's position as the most widely installed and proven single-wafer processing platform played well with APCVD's migration into single-wafer multichamber processing. Second, Applied

Materials' extensive sales and support infrastructure served as a powerful marketing machine in bringing the company's new SACVD into the global market. This is an important strength because APCVD/SACVD is a DRAMsensitive market, where a large part of worldwide demand is in the Japanese market and the burgeoning Asia/Pacific market. Applied Materials' global presence played an essential role in the rapid acceptance of its new products into the world market.

HDP CVD is an emerging market that is poised to grow dramatically as a gapfill application for the sub-0.35-micron generations. Applied Materials'

Centxira is going through a redesign and should remain one of the leading contenders in the HDP CVD market.

Tokyo Electron Ltd. and Kokusai Electric Co.

In the first year since dissolving the Varian/TEL joint venture, TEL expanded its lead over its tube CVD rival, Kokusai Electric, by achieving

U.S.$319 million in sales in 1995, or 44 percent of the vertical tube CVD business. It is from the two largest tube CVD markets, Asia/Pacific and

Japan, that TEL derived more than 88 percent of its tube revenue, and these two regions will continue to be the battleground where leadership in tube

CVD will be fought. TEL also has strong capabilities in nontube CVD with its offering of CVD tungsten silicide and tungsten tools. Ctuxently sharing the top position with Applied Materials in CVD tungsten silicide sales, TEL may have a slight advantage going forward as a result of its entrenched position in Japan and Asia/Pacific, the largest CVD tungsten silicide markets.

Kokusai remains the second-largest tube CVD supplier. It grew its tube revenue 75 percent to U.S.$239.8 million. The bulk of its sales growth came from Asia/Pacific, which now represents 62 percent of Kokusai's revenue base. The other three regions, including home base Japan, turned in lackluster performance. Kokusai may need to do something quickly to prevent losing more market share to TEL in Japan and Silicon Valley Group in North America and Europe. One solution could be to build a stronger global sales and support infrastructure. Kokusai recently became the sole owner of North America-based Bruce Technologies International Inc., which should help forge a global network of marketing and services for the strong suite of vertical and horizontal furnaces offered by the two companies.

Novellus Systems Inc.

After an explosive year in 1994, in which its revenue more than doubled and its nontube CVD market share rose more than six points, Novellus' performance in 1995, although respectable, was overshadowed by the megagrowth of the overall nontube CVD market. Novellus' nontube revenue grew 78 percent, which in the context of a market that doubled in sales translated into a loss of nearly 3 percentage points in nontube CVD market

SEMM-WW-DP-9607 ©1996 Dataquest September 30,1996

Semiconductor Equipment, iVIanufacturing, and Materials Worldwide share for Novellus. Despite that, Novellus' position in nontube CVD remains strong. In both PECVD and LPCVD reactors, Novellus' well-secured No. 2 supplier position makes it the most viable, low-cost-of-ownership alternative to Applied Materials.

Also, Novellus' HDP CVD product, SPEED, is making headway and has attained early leadership in the emerging HDP CVD market. HDP CVD could be the market in which Novellus will no longer have to play second fiddle.

Watkins-Johnson Company

Watkins-Johnson grew its CVD revenue by 65 percent in 1995 to U.S.$202 million and maintained its top position in the APCVD/SACVD market.

However ,with the APCVD/SACVD market growing much faster, WJ's market share slipped from a dominant 64 percent to 48 percent, a whopping

16 percent loss. Applied Materials' SACVD, Amaya's new ATO-8000, and

Alcan/Canon/Quester were all eating into WJ's long-held control of the

APCVD BPSG and PSG ILD markets. With DRAM investment waning in the near term and competition in APCVD/SACVD rising, WJ will need to tread carefully and use its resources adequately to prevent its market share from slipping further.

There is a new market opportunity for APCVD/SACVD in the adoption of shallow trench isolation (STI) in the next generation of semiconductor devices. APCVD, with its unmatched step coverage, is the ideal solution for

STI. WJ should leverage its leadership position in APCVD and aggressively pursue this new opportunity. Also, WJ has successfully developed its first single-wafer multiciiamber process platform, the WJ2000, targeting the HDP

CVD market. The WJ2000 is getting good reviews and, if marketed successfully, should provide fresh ammunition in WJ's defense of its CVD market position.

ASM International N.V.

ASM International raked in U.S.$136.5 million in sales in CVD in 1995, a 98 percent increase over the previous year. ASMI fotind dramatic success with its Eagle 10 PECVD system, which grew more than four times in sales to reach U.S.$45.3 million. Although this represents only 5 percent of the

U.S.$855 million PECVD market, it does serve notice to Applied Materials and Novellus that there may be cracks developing in their lock of the

PECVD market.

Sales of ASMI's tube CVD equipment reached U.S.$91.2 million, equaling growth of 49 percent, a subpar performance in light of the 77 percent jump in total tube CVD revenue. ASMI's tube CVD growth is masked by the company's horizontal PECVD tube shipments, a market with declining sales but which ASMI owns without having to pump in any significant resources.

Factoring out horizontal PECVD tube, ASMI's other furnace offerings, the vertical furnace cluster tools, reached U.S.$63.3 million in sales, or 80 percent growth. ASMI's vertical furnaces are gaining rapid acceptance in North

America and are the most serious challenge to SVG, the reigning Americas region furnace supplier.

SEMM-WW-DP-9607 ©1996 Dataquest September 30,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

. Alcan/Canon/Quester

Alcan/Canon/Quester captured 4.5 percent of the overall CVD market by increasing its sales to U.S.$115 million, a 78 percent rise over 1994. The jointventure APCVD supplier continued to find success where it has strong, entrenched marketing and support; the company holds the No. 1 and No. 2

APCVD positions in Japan and Asia/Pacific, respectively. But in the

Americas region and Europe, the company continues to lag behind Watkins-

Johnson and the formidable newcomer. Applied Materials.

Alcan's challenge in the APCVD market, which is becoming increasingly crowded by the recent entry of new players, is to build differentiation based on low cost of ownership and enhanced productivity in its product offerings.

Alcan will need to solidify its strong presence in Japan and Asia/Pacific against encroaching APCVD/SACVD upstarts Applied Materials and

Amaya. In North America, where Watkins-Johnson is fighting on its home turf against giant Applied Materials, severe competition could translate into a price war. Alcan may be better off focusing on the regional markets where it has an entrenched advantage.

Genus Incorporated

Having exited the CVD tungsten market a few years ago and having begvm focusing solely on tungsten silicide. Genus has broken its long stagnation and grown its CVD revenue significantly for the first time in four years.

Genus' market share grew for the first time as result of the doubling of its

CVD revenue in 1995. Genus is now a viable tungsten silidde equipment vendor and has carved out a 25 percent share of the worldwide CVD tungsten silicide market. Much of Genus' recent success is attributable to its dichlorosilane (DCS) LPCVD tungsten silicide process. The preferred choice for bitline fabrication in 64Mb/256Mb DRAM, DCS tungsten silicide could be the ticket with which Genus regains its past prominence in metal CVD.

Silicon Valley Group Inc.

Silicon Valley Group grew its tube CVD sales by 90 percent in 1995, reaching

U.S.$76.5 million in sales. SVG remained the market leader in Americas and

Europe for the second straight year. However, SVG is facing a serious challenge in Europe from tube giant TEL, which staged a dramatic entry last year and quickly gained 30 percent of the Exiropean market and the secondplace spot. Another and bigger challenge for SVG in the furnace market is to forge a strategy for penetrating the lucrative Asia/Pacific and Japan markets, which hitherto have been dominated by TEL and Kokusai.

CVD Market Forecast

Table 2 presents historical data and Dataquest's projection of the worldwide

CVD equipment market segmented by film applications. After three years of hypergrowth, the CVD market is expected to take a bit of a breather in

1996—market growth moderates in the face of growing hesitation in chip capadty investment that will likely lead to a significant market contraction in

1997. Bucking the trend, however, could be technologies that are viewed as critical technology enablers. One such technology is HDP CVD oxide, which

SEMM-WW-DP-9607 ©1996 Dataquest September 30,1996

10

Semiconductor Equipment, Manufacturing, and Materials Worldwide is expected to provide the most exciting growth in CVD during the next five years. Dubbed the gapfill solution of choice for advanced logic devices at the

0.35- and 0.25-micron generations, HDP oxide is projected to achieve nearly

U.S.$0.5 billion in sales by the year 2001. The adoption of growing layers of interconnects in IC devices and the need to fill increasingly small gaps between metal lines with superior films is driving the growth of HDP CVD.

PECVD oxides, both silane and tetraethylorthosUicate (TEOS), will also benefit from the continued rise in metal interconnect layers. PECVD silane oxide, in particular, is expected to experience a resurgence because of its superior characteristics as a capping layer for chemical mechanical planarization polishing. APCVD/SACVD will likely be pulled out of the running as a gapfill or IMD solution as the HDP CVD/PECVD oxides combination gains acceptance. The silver lining in APCVD/SACVD, however, w^ill be found in the adoption of shallow french isolation, for which

TEOS/ozone is the prime candidate as the oxide material.

Little growth is projected for LPCVD polysilicon, as DRAM capacity investment is expected to stay in a hiatus for a good part of 1996 through

1998. However, growth is expected to resume beginning in 1999, when investment for the next generations of memory fabs kicks in. In CVD polysilicon, a technology shift will take place from the current batch furnaces to minibatch or single-wafer systems. Consequently, single-wafer CVD polysilicon (and polycide) tool offerings are likely to expand. LPCVD metal comprises both the logic-sensitive (CVD tungsten and CVD titanivun/titanium sUicide) and DRAM-sensitive (CVD tungsten silicide) segments. This helps mitigate LPCVD metal's exposure to the market cyclicity generally caused by the dramatic swings in DRAM capacity investment. Consequently, LPCVD metal is projected to experience growth that is both healthy and stable.

Table 2

1995 Worldwide CVD Equipment Market History and Forecast by Film Application

(Millions of U.S. Dollars)

APCVD + SACVD Oxides

LPCVD Oxides

PECVD Oxides

HDP Oxide

LPCVD Nitride

PECVD Nitride

LPCVD Polysilicon

Single-Wafer CVD

Poly/Polycide

1992

81.9

81.2

176.8

4.8

1993

103.4

106.1

240.0

12.7

1994

191.4

181.9

326.1

7.9

1995

425.0

320.4

615.5

13.1

1996

491.7

367.9

808.4

78.5

1997

397.0

320.6

730.3

123.7

1998

388.3

357.9

756.3

180.8

1999 2000 2001

469.4

442.5

666.5

612.8

878.1

973.1 1,513.7 2,133.9

225.9

348.4

766.0

467.6

CAGR(%)

1995-2001

12.9

15.6

23.0

81.5

29.5

98.4

36.7

123.1

47.2

167.1

83.2

268.3

87.8

308.5

75.9

266.3

84.5

279.8

99.1

332.7

138.3

389.4

178.5

483.8

13.6

10.3

87.9 104.6

1.1

188.5

18.8

356.3

41.9

389.1

47.8

327.9

40.5

322.6

78.9

407.2

126.8

577.3

151.6

726.9

183.3

12.6

27.9

LPCVD Metal

103.0

142.2 207.1 479.2

498.4 438.9 446.0 516.2 727.0 1,002.4

Total 663.5

Source: Dataquest (September 1996)

869.9 1,326.2 2,583.0

3,054.1 2,700.9 2,855.6 3,529.5 5,049.0 6,728.8

13.1

17.3

SEMM-WW-DP-9607

©1996 Dataquest

September 30,1996

12 Semiconductor Equipment, Manufacturing, and Materials Worldwide

For More Information...

Calvin Change, Senior Industry Analyst (408) 468-8605

Internet address [email protected]

Via fax (408)954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness. It does not contain material provided to us in confidence by our clients. Reproduction or

I J ^ T ^ f ) ! l ^ ^ S t disclosure in whole or in part to other parties shall be made upon the written ai\d express consent of Dataquest.

A r. o ^ ^ * S ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and Materials

Worldwide

Event Summary

How Long and Deep Will the Slowdown In Wafer Fab Equipment

Be? What Will Be the Shape of the industry Recovery?

Abstract: Overcapacity in the DRAM market, created by the massive spending from 1994

through the first part of this year and the normal DRAM product migration, is finally taking its toll, resulting in a faster than expected contraction in the wafer fab equipment market. This document is the complete transcript of the telebriefing held by Dataquest on

July U, 1996, with the release of the forecast update on capital spending and wafer fab equipment.

By Clark J. Fuhs, Nader Pakdaman, and Calvin Chang

Opening Statement by Dataquest

MR. FUHS: Welcome to the Dataquest telebriefing this morning. Our topic is from the Semiconductor Equipment, Manufacturing, and Materials group

(SEMM). Our group tracks most aspects of the actual manufacturing of semiconductors worldwide. Today we will be discussing the outlook and forecast for wafer fab equipment and capital spending. We will also present our forecast on silicon wafers supported by a recent supply-and-demand analysis in the silicon market.

My name is Clark Fuhs, Director and Principal Analyst in the group, and your host for the next hour. With me today are Nader Pakdaman and Calvin

Chang, who are also senior analysts from the group. After a brief 15- to 20minute review of our forecast, we will open up the session to your questions.

Dataquest

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-DP-9606

Publication Date: September 16,1996

Filing: Perspective

(For Cross-Technology, file in the Semiconductor Regional Markets and Manufacturing binder)

Semiconductor Equipment, Manufacturing, and Materials Worldwide

You should have received a set of two tables and five figures that visually describe some specific topics we will be reviewing.

Our group has just released its midyear semiconductor capital spending and wafer fab equipment forecast. These are summarized in Tables 1 and 2 of your handout. Our forecast process has several cornerstones, including semiconductor production by region, a worldwide database of existing and planned fabs, and independent comprehensive surveys of the equipment and semiconductor companies. We have just completed an internal audit of the fab database in which we have scrutinized the planned fab activity worldwide.

One comment we should make here regarding our capital spending survey.

Our survey was initially done in early May. By mid-June things had changed dramatically, particularly in Taiwan and Korea. We have made an attempt to update these figures before the forecast release, not only for announced cutbacks, but for anticipated cutbacks later in 1996 and in the following year.

In many cases, we have reverified the survey responses with companies.

However, since we are in a dynamic market where companies are still evaluating their investment levels, some of our judgment is in the numbers.

Forecast Overview

The survey results are just one input into several forecasting models we have, which include analysis of trends and semiconductor production, raw silicon consumption, spending ratios, investment cycle over and under investment, stepper vmit growth to DRAM price-per-bit analysis and semiconductor revenue-per-square-inch growth patterns. Our forecast shows the following highlights.

• The year 1996 is a year that turns from growth to decline, but will net out with about a 17 percent growth in wafer fab equipment year over year.

• The only reasons 1996 remains a double-digit-growth year are that there were such strong backlogs coming out of 1995 that the first half remains strong, and a few companies such as IBM, Texas Instruments, TSMC,

Siemens, and Chartered Semiconductor have continued to grow investments in the first half of the year.

• Push outs, cancellations, and lack of new orders in wafer fab equipment are commonplace today, virtually guaranteeing 1997 to be a down year.

• In order to quantify and characterize some of the dynamics of the market, we have established for the first time a top line quarterly shipment forecast for wafer fab equipment. This is shown in your handouts and Figures 1 and 2. We expect a decline to accelerate into 1997 and should perhaps bottom in the first half of the year. We believe the first sequential growth quarter after the decline will be in the fourth quarter of next year, but the growth will be constrained tmtil late in 1998.

• This quarterly outiook has produced the annual forecast shown in Table

2 where we show 1997 decliiung by nearly 16 percent and 1998 as a single-digit-growth year.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and l\/iaterials Worldwide

Table 1

Capital Spending Forecast, 1995-2001 (Millions of U.S. Dollars)

Total Capital Spending

Percentage Growth

Americas

Percentage Growth

Japan

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Source: Dataquest (July 1996)

1995

38,308

73.5

12,147

68.8

9,910

48.6

4,137

65.2

12,115

111.8

1996

45,344

18.4

14,828

22.1

9,692

-2.2

5,058

22.3

15,766

30.1

1997

40,942

-9.7

13,461

-9.2

9,566

-1.3

4,444

-12.1

13,470

-14.6

1998

42,915

4.8

14,599

8.5

10,331

8.0

4,317

-2.9

13,668

1.5

1999

53,954

25.7

18,453

26.4

12,730

23.2

5,637

30.6

17,134

25.4

2000

75,777

40.4

24,571

33.2

16,829

32.2

7,725

37.0

26,653

55.6

2001

96,715

27.6

29,721

21.0

20,495

21.8

9,718

25.8

36,781

38.0

CAGR (%)

1995-2001

16.7

-

16.1

-

12.9

-

15.3

-

23.6

-

Table 2

Wafer Fab Equipment Forecast, 1995-2001 (Millions of U.S. Dollars)

Total Wafer Fab Equipment

Percentage Growth

Americas

Percentage Growth

Japan

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Source: Dataquest (July 1996)

1995

19,054

76.6

5,179

62.3

6,352

73.5

2,316

69.1

5,208

102.9

1996

22,309

17.1

6,217

20.1

6,423

1.1

2,642

14.1

7,026

34.9

1997

18,760

-15.9

5,453

-12.3

5,272

-17.9

2,428

-8.1

5,607

-20.2

1998

19,900

6.1

6,163

13.0

5,635

6.9

2,486

2.4

5,616

0.2

1999

25,380

27.5

7,797

26.5

7,011

24.4

3,124

25.7

7,448

32.6

2000

36,219

42.7

10,530

35.1

9,679

38.1

4,253

36.1

11,757

57.8

2001

46,889

29.5

13,105

24.5

11,694

20.8

5,520

29.8

16,570

29.5

CAGR (%)

1995-2001

16.2

-

16.7

10.7

-

15.6

-

21.3

-

As the absorption of capacity happens during this two-year pause, we would expect a robust resumption of growth in 1999 with the wafer fab equipment market growing to about the $36 billion level by the year

2000.

What Happened to Trigger the Slowdown?

One year ago at this time, we were asking questions like, "Are we ready to call the silicon cycle dead? Are we in a different world today than we were in

1984 and 1989?" At that time, we pointed out that the cycle today is similar to the historical cycles from the perspective of the DRAM business and how

DRAM product cycles and silicon area are related. An equipment "pause" was predicted, as we classified it then, and it was expected to be triggered by price softness in the DRAM market, which would depend on when the 16Mb

SEMI\/l-WW-DP-9606

©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 1

Wafer Fab Equipment Quarterly Revenue Forecast

Billions of U.S. Dollars

Q-S

5 -

4 -

3 -

2 -

1 i

I u

I I i

I i

• l l

" • I

I

1

I

I

I

^ ^

-i 1 I ' I I "-^r'''^ 1 " " ' I 1 1 r

I

I

Q1/95 Q2/95 Q3/95 Q4/95 Q1/96 02/96 Q3/96 04/96 01/97 02/97 03/97 04/97 01/98 02/98 03/98 04/98

Source: Dataquest (July 1996)

966233

Figure 2

Wafer Fab Equipment Quarter-to-Quarter Sequential Growth Rates

Percent

6 0 -

5 0 -

4 0 -

3 0 -

I

2 0 -

1 0 -

i

0 r ^ i

^ ^ .

1 r

-mr

| i ^ l ^ i " i ' ^ i i * ^ ! ^ ! 1

-10-

01/95 02/95 03/95 04/95 01/96 02/96 03/96 04/96 01/97 02/97 03/97 04/97 01/98 02/98 03/98 04/98

966234

Source: Dataquest (July 1996)

SEMM-WW-DP-9606

DRAM was being produced at economic jdelds. We thought that ramping to a yield of 65 percent would be a key level, on the road to historical full yields of above 80 percent. This was extremely important for us to tmderstand because the memory spending accounted for nearly 50 percent of capital spending in 1995, up from 40 percent in 1994, and 33 percent in 1993.

In the second half of last year, a complicating factor entered the picture. The overanticipated demand for memory created by the introduction of

©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and !\yiaterials Worldwide

Windows 95, which, taken with the short supply of memory last year, caused inventory to inflate to enormous levels. This situation started correcting in the final quarter of last year, slightly in advance, but essentially on top of the yield trigger for the 16Mb DRAM conversion. Today we have a double-barrel driver, both demand and supply, to push the DRAM price slide and capital spending downturn. These events together have increased the anticipated severity of the downturn.

The Shape and Length of the Downturn

We expect this contraction to be sharp, relatively deep, but slightly shorter than historical norms—in the 18- to 24-month time frame. The key reason for the belief of a shorter than normal slowdown is the continued robust forecast for PC shipment unit growth worldwide through this decade, driving the need for silicon.

We also do not see a stoppage in advanced technology investments indicating a belief in the customer base of a strong end-user market for semiconductors. Let's review some specifics.

• It is very normal in this type of downturn to get a pocket of companies who will stay and continue to invest in the infrastructure. We mentioned a few names earlier, and Korean companies should continue to grow spending in moderation this year.

• Japanese companies who held spending up in the last cycle for a year or more have already shut off spending. These companies as a group will spend slightly less in dollar terms in 1996 than they did in 1995.

However, they are still concentrating what they are spending to build the shells for the next upturn, along with pilot investments for advanced products.

• Outside of TSMC and Chartered, most all other Asian companies are cutting back dramatically, starting in the last couple of months.

• Our most severe reasons for the downturn for 1997, therefore, are Japan and Asia, as these are the DRAM production centers. As a group, the

DRAM manufacturers are responding to the situation by cutting spending faster this time than in previous cycles.

An important facet of the semiconductor revenue downturn is that it is driven more by lower margins than the end-user market and lower unit demand. The equipment industry now must rely on a fundamental demand in PCs and telecommunications to catch up to supply. Any hiccups in the end-user markets for semiconductors and we must batten down the hatches for a much longer downturn.

The Shape and Timing of the Recovery in Spending

How will the equipment industry recover? What are the anticipated dynamics? The industry today is primarily experiencing a DRAM oversupply coupled with a product transition. In order to determine how capital spending may recover, it is important to understand how this excess capacity may migrate or trickle into other areas in semiconductor capacity.

Please refer to Figure 3 in your handout.

SEI\/ll\/i-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Figure 3

Semiconductor Market in "Capacity Trickle" Mode

Old 4MD

Fabs: 0.6-0.8

Micron

DRAM Overcapacity

1

J

Recovery Last

laie l o M D

Fabs: 0.4-0.5

Micron

Commodity SRAM, 1

Flash Memory 1

MCU, Analog,

Mixed Signal

Recovery Second

1

X ^

1

Recovery First

Power, Discretes 1

- X

Fast SRAMs. 1

Advanced Logic 1

96623S

Source: Dataquest (July, 1996)

There are two general blocks of capacity now available from today's conditions. The first comprises old 4Mb DRAM fabs that carmot run 16Mb chips. These are limited to two-level metal processing and are generally in the 0.6- to 0.8-micron linewidth technology. Microcontrollers, telecommimications chips, mixed signal, and analog ICs are quite happy being processed in this type of fab. It is likely that most of these fabs in Japan and some in Korea will and have already started to migrate into this area.

The power and discrete chips, which we might think also might have some capacity come their way from this first block, have very specialized processes not found in the older DRAM fabs. So these segments are relatively isolated from extraneous supply impact, since significant time and money is required to convert, and specialized technology and knowledge within the company is required.

The second block of capacity comprises idle or underutilized advanced 16Mb capacity, which is again limited today to two-level metal, rurming about the

0.4- to 0.5-micron linewidth technology. Because these fabs generally lack the process sequences needed to support self-aligned silicide and three-or-morelevel metal, they cannot effectively be redirected to advanced logic or fast

SRAM. Therefore, they are limited to commodity SRAM, flash, other nonvolatile memory, or a limited span of logic products that do not require those processing techniques. We expect most of these plants, quite frankly, to remain somewhat idle.

Therefore, we believe the first areas of spending recovery will be in the advanced logic area. An equipment company positioned in these markets

SEMM-WW-DP-9606

©1996 Dataquest

September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide should have a more moderate slowdown and will start to recover perhaps as early as the mid-1997. The microcontroller, analog, mixed signal, and telecom chip capacity will be next to recover, but probably not until the end of 1997 or early 1998. Motorola will be a key company to watch here. The

DRAM segment, the root cause of the problem, is not expected to resume robust spending until the middle or late 1998.

Outlook on Silicon Consumption

Now I would like to shift gears and talk about our silicon outlook, since it offers a somewhat different view of the market. Figure 4 shows our silicon forecast in terms of square inches. A detailed alert will be sent to our clients shortly. The silicon market demand picture is only slightly changed from our outlook six months ago.

Dataquest has predicted a slower growth picture for square inch consumption over the next few years as the industry migrates to the 16Mb

DRAM, which uses silicon area more efficiently per bit than the 4Mb DRAM.

The dynamics of our forecast have been modified slightly to account for the quicker than expected price decline in migration timing, which does lag capital spending growth by about 6 to 12 months. The forecast growth for

1996 in silicon has actually been raised slightly, since lower DRAM prices are resulting in more memory installed per PC. Since the industry has only started to migrate to the 16Mb DRAM, the short-term dynamic will be an increase in silicon area consumption.

The accelerated schedule of the product migration has caused our growth outlook to be slightly more cautious for 1997 and 1998, but stronger into

1999. Still, in the slowest-growth year of 1998 it is expected to be about 9 percent growth, as shown in Figure 5, primarily because PC unit demand worldwide is expected to remain robust through the decade. Generally, silicon area is much more stable than capital spending.

How can the silicon industry grow in the face of a slowing semiconductor and equipment market? The answer lies in the cyclical dynamic of the semiconductor industry revenue per square inch. Revenue per square inch is cyclical, and fluctuates with DRAM prices and semiconductor margins.

Because of cyclical pricing behavior, silicon consumption tends to meet more closely following unit demand rather than the revenue of the semiconductor industry.

The capital spending and equipment markets are more directly affected by factory utilization than demand. For example, we believe the semiconductor industry runs at maximum at about 92 percent factory utilization. As utilization in an upturn coming from low levels goes above the 85 percent level, spending accelerates and does not typically slow dov^ni until it falls back below this 85 percent level. There are constant overshoots, meaning capacity in the industry is added in spurts, whereas silicon is consumed in a more steady fashion. It is the tie between the two that we believe is important to watch during turning points in the industry.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Figure 4

Silicon Area Consumption Forecast

Millions of Square Inches of Silicon

Semiconductor Equipment, Manufacturing, and Materials Worldwide

1993 1994 1995 1996 1997 1998 1999 2000 2001

966236

Source: Oataquest (July 1996)

Figure 5

Silicon Consumption Growth Is Less Volatile and Lags Capital Spending Slightly

Percent

80-ff

-20-"

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

966237

Source: Dataquest (July 1996)

That concludes our prepared comments. We would now like to open this briefing to your questions.

SEMM-WW-DP-9606 ©1996 Dataquest

September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Question and Answer Period

OPERATOR: Thank you. If you would like to ask a question, please press one on your touch-tone phone. I'll take your questions in the order that they are received. #30, you may ask your question.

CALLER #30: Good morning. I would like to go back to your Figure 2, and if you can explain. I have a little problem understanding what this says, and also if you can walk through again Figure 3, which is very important to me.

Thanks.

MR. FUHS: To explain clearly, please refer to Table 2 which is our aimual forecast for wafer fab equipment, presenting our revenue by region in an annual fashion. Figure 1 shows the sales represented in quarterly numbers, so the quarterly numbers for the four quarters of 1996 would add up to the aimual number in Table 2. What we have done in Figure 2 is prepared what the growth rate is from quarter to quarter. In other words, if I go and look at the third quarter of 1996, it is the first negative quarter. Basically, what we are sajdng is that on a seasonally adjusted basis, the third quarter of 1996 will be below the second quarter of 1996.

CALLER #30:1 understand. Thanks.

OPERATOR: Thank you. #18, you may ask your question.

CALLER #18: Yes, I was just curious. You were saying earlier that obviously, for Dataquest, continued strong growth in the PCs is an important factor in how long the downturn is going to be, and I'm just wondering, if you are starting to rethink that in terms of the news starting to come out of people like Hewlett-Packard about a slowdown.

MR. FUHS: That is really a call for our PC group to make. They're standing by their growth rate of about 18 percent or 19 percent this year in unit growth, and I believe their compound aimual growth rate through the year

2000 is roughly 16 percent or so. In past semiconductor slowdowns, the PC unit growth has gone down into the low single digit-growth, and that is basically what happened in the extended downturn in the early 1990s. We are not viewing that to repeat right now.

CALLER #18: Thank you.

OPERATOR: Thank you. #25, you may ask your question.

• CALLER #25: Hello, Clark. I noticed that you stated that you believed there's still going to be investment in advanced technology, and I also noted that you didn't really talk about the details on equipment, but do you see any particular areas where you believe there will be strong investment, for example, batch tools to single wafer tools or cluster type tools? Any comments?

MR. PAKDAMAN: This is Nader Pakdaman. Perhaps maybe we can't take it at this time to the level of granularity in terms of any particular equipment.

We are seeing in our siirvey of fabs and information on capital spending that

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

_10 Semiconductor Equipment, Manufacturing, and Materials Worldwide people are not just canceling, but are delapng capacity investments into the latter part of 1997 and 1998 across the board, even for DRAM manufacturers.

But they are continuing capital investments for equipment with capability of

0.3 nucron and below, obviously not in the same volumes. These include higher ASP equipment and more advanced equipment to set up for later production, and as Clark mentioned in the initial discussion the industry is building a lot of shells. So a lot of the fab announcements, yes, actually fabs are likely to be started in the year of production currently planned for the most part, but capacity will probably be pushed out and the ramp rate is going to be slower. The technology that will be introduced will be more advanced than what was initially anticipated.

MR. FUHS: One additional comment. The 16Mb DRAM is likely in the later stages of the cycle to be processed at 0.3 or 0.25 micron, and what the DRAM companies are doing, particularly the Japanese, is to take this time to investigate advanced technologies. 1 just saw a release a couple of days ago on Gasonics announcing an order for advanced remote plasma strip systems to go into Japanese companies. We believe that the Japanese companies in particular are going to take this opportunity to look at advanced systems to determine their capability in order to design into their later stages of their

16Mb DRAM process flows.

OPERATOR: Thank you, and #13A, you may ask your question.

CALLER #13A: Two questions for you. I noticed that in Table 1, it looks like in 1997 the wafer fab equipment negative growth was larger than the capital spending. Can you go through some of the assumptions behind those two different growth rates? Are you suggesting the semiconductor manufacturers are building more shells?

MR. FUHS: The wafer fab equipment is a subset of the capital spending.

That's correct.

CALLER #13A: That is the main assumption behind that. And my second question. You mentioned that DRAM and oversupply were probably recovering in mid to late 1998. Now, can you make some comments on the supply and demand situations for 1997 and 1998, you know, maybe first half and second half, kind of time frame?

MR. FUHS: The comments that we made are based on the supply demand analysis in DRAM. The current outlook is that DRAM will remain in oversupply at least into the first half of 1998.

CALLER #13A: Thanks.

OPERATOR: Thank you. #5, you may ask your question.

CALLER #5: I've got a two part question. If you look at Figvire 2, you show going from a negative to a positive sequential growth rate. Would you review again the reasons you see it going positive starting the end of 1997, that is, what causes the inflection point between the negative to positive cycle and if you could, what portion of that change in growth rate is

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide 11 attributed to technology kinds of buys versus the portion that's attributable to capacity buys?

MR. FUHS: Well, we generally see that the cuts will be fully reflected, with shipments at their lowest levels in the middle of 1997 driven primarily perhaps by the DRAM segment. What we're looking at is a recovery in the advanced logic area in the last half of 1997. The business of wafer fab equipment is driven primarily by capacity needs. The R&D budgets are much smaller than the capital spending budgets for many of these companies when it comes to equipment issues and, in general, the average is roughly probably three or four times as much equipment bought for capacity than it is for R&D. During slow capacity times, that average will probably come down in the closer to two to one to three to one ratio. The driving force behind this stability in R&D spending is growth in the end-use markets for semiconductors. R&D spending will probably continue to have stable growth through this period.

So what we expect at the end of 1997 for capacity buys is a stabilization in the DRAM area, a return to growth in the advanced logic area, with a continued increased investment in R&D systems for 0.3-micron technology.

Through the first half of 1998, you start to see the capacity increases at the microcontroller, analog, mixed signal areas and then the end of 1998 into

1999, we expect we will start to see a sequential acceleration in the overall market again as DRAM expansion returns.

CALLER #5: Thank you.

OPERATOR: #13, you may ask your question.

CALLER #13: Yes, Clark, can you comment on the equipment lead time, how lead times are changing? For example, six months ago or so they were saying that lead times were like 12 to 18 months. How is it now?

MR. PAKDAMAN: I will discuss the stepper situation. I think it is a prevalent indicator for the rest of the equipment market. On the advanced systems, we do not see the backlog or the lead times really shrinking perceptively. For advanced deep-UV systems, the lead time is still rather long. However for mainstream products, primarily above 0.3 micron and iline, we are seeing definitely seeing the pressure alleviated and the lead time coming down to historical levels. We are thinking that by next year around this time, it will come down to about rune months. Right now, it is definitely above a year.

MR. FUHS: One other comment. Implanters and steppers tend to have the longer lead times. Processing equipment, the CVD and etching equipment and so on generally have much shorter lead times. I would classify that an average lead time in this industry is between four and five months. I think at the end of 1995, we were looking at an average six month lead time and I think we'll go out of 1996 with a four to three-and-a-half-month lead time.

OPERATOR: #26, you may ask your question.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

_12 Semiconductor Equipment, Manufacturing, and Materials Worldwide

CALLER #26:1 have three questions and I'll make them brief. First, is your original growth by headquarters location or the actual region of spending?

Second question is, I'm wondering if, and I want to get your feel for the risk of the DRAM downturn being longer. What I see going on is that the DRAM guys are converting to 64Mb rather quickly, and they could create a major price war in 64Mb. If they're profitability totally goes away, I would think that they may want to hold back longer before they start spending again, which could make this downturn in DRAM a long one rather than a short one. And the third question is, what impact on 300mm development do you think this downturn will have?

MR. FUHS: First question. The capital spending is by region of spending, so the Americas region would include anything that Samsung spends here in the United States, and thus would be considered American spending.

The second question had to do with DRAM and the impact of the industry trying to convert to 64Mb. Every time that there is a downturn like this, the immediate cry is, "Let's leapfrog and go to the next generation." I am very suspicious of that argument, primarily because there is not a 64Mb design that a PC manufacturer will buy and put in their systems. A 64Mb DRAM is fully 8MB, and the PC industry wants to ship PCs in with megabyte increments which are marketable in the business and particularly the consumer markets. I believe we are projecting that the average PC will have

48-56MB per box, and the shipment mix is expected to be 32,48, and 64MB systems. In order to get this 16MB "nugget" between systems, and use 64Mb

DRAMs, manufacturers will need to make a x32 configuration for the 64Mb in order to penetrate the market. This granularity issue is one of the key reasons behind the long length of the 4Mb DRAM cycle. And just to put this in perspective, nobody makes a x32 today! So to presume that you can drive the end-use industry by the supply side is unrealistic and I'm suspicious of that outcome. I think that by the time we get to the 1999 or 2000 time frame, the 16Mb DRAM will still be the dominant chip being shipped.

MR. PAKDAMAN: I want to also add a comment to that. The fact is that perhaps maybe we should not concentrate on device density of 16 or 64Mb, but look the issue per square inch of silicon. There will be continued investments to put more bytes per square inch, whether it's in a 16Mb or a

64Mb configuration. We are definitely seeing this in the patterns of spending and the price of equipment. But there is one way or another today a byte glut, first because of increased jrields, second because of the conversion that has happened from 4Mb to 16Mb. So for that glut to be absorbed we expect to have another at least three or four quarters if not more ahead. In the interim, companies will be making advanced shrink designs to put more bjrtes per square inch. We will see the "64Mb technology" being applied to

16Mb DRAM as 16Mb technology was very much utilized for the 4Mb product. The fact that the margin has really decreased and the prices have fallen so far recently for DRAM obviously brings the prospect of going to the next device generation closer into focus, but we expect with the robust enduser market that the prices will stabilize eventually, and 16 Mbit will have a similar life cycle as previous generations.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide 13_

MR. FUHS: Just one more comment and then we'll answer youx third one really quickly here. There is a risk that this DRAM slowdown could last longer, and that depends on how many of the 6-inch fabs out there convert to 8-inch wafers, delaying equipment purchases. We have actually increased our product demand for 200mm wafers in the 1997-1998 time frame because we were starting to see a lot more 6-inch to 8-inch conversions happening in that time frame. What this does is add square inch capacity to the industry without the benefit of additional equipment, so that may push the equipment purchase cycle out by six months or so into 1999.

Regarding 300mm wafers, there has been some acceleration in the wafer industry to put in 300m capacity. Four or five companies have announced pilot lines to produce SOOrnm wafers, and from the semiconductor side NEC is becoming a little bit more aggressive. 1 do not believe that the industry's slowdown will change the pace appreciably. I think we will continue to move toward the first 300mm R&D pilot line for semiconductor production online by 1999 or 2000, placing the major production ramp of 300mm in the

2003-2004 time frame. This picture is consistent with what we have been saying for the last 18 months and we still see that path continuing.

OPERATOR: #30, you may ask your question.

CALLER #30: Figure 3, could you walk me through the timing of these recoveries?

MR. FUHS: At this point, we expect the advanced logic and fast SRAM part of the market to begin recovery by mid-1997. The arialog, mixed signal, and microcontroller area by the end of 1997 and early 1998, and then the DRAM recovering last in the latter half of 1998, perhaps with the risk to go into 1999.

CALLER #30: Okay.

OPERATOR: #15, you may ask your question.

CALLER #15: Looking again at Figure 3 with the early recovery of fast

SRAMs and advanced logic, I'm assuming that's why you're showing in

Table 2 the Americas kind of leading the recovery in 1998. Can you go into a little more detail on the companies/products you see leading this recovery?

MR. FUHS: Generally speaking, logic chips connected with the multimedia

PC come to mind—the graphics controllers, anything having to do with microperipherals, logic controllers, and so on. The advanced logic foundries like TSMC and Chartered will continue to do pretty well. They will probably have some pricing pressures in the next one to two years because of the general overcapacity. The fabless companies are going to have a field day, and as far as their cost structure is concerned, and they are the ones that are actually inaking out like a bandit with this downturn. We would expect some of the foundries to do relatively well if they are focused in advanced logic, but as I said there is going to be pricing pressure for wafers out there.

The other area that will probably recover from the capacity perspective would be the advanced ASIC areas, especially the PLD and ASSP areas.

These devices tend to use the self-aligned silicide process, and it is that kind

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

_14 Semiconductor Equipment, Manufacturing, and Materials Worldwide of capacity, along with the multilevel metallization capacity, that we expect to see coming back first.

OPERATOR: #7, you may ask your question.

CALLER #7: Continuing on about the rate, or how quickly we will change and enter a growth period, I'd like to understand your viewpoint on what from the end-user standpoint, either events or issues, will trigger either an acceleration or a deceleration of the conversion to growth in 1997 and 1998.

MR. FUHS: Whatever makes a PC box ship. It comes down to that.

Unfortunately, that is the simplistic and most significant answer. There are other areas to look at however, such as telecommunications and networking.

The implementation of MPEG-2, for example.

MR. PAKDAMAN: I think one significant difference between this downturn and previous cycles is that we are riding on a much larger semiconductor and end-user market, and our forecast for the year 2000 is still above a $300 billion semiconductor market, even if prices do not tend to stay as robust as they did in 1994 and 1995, particular for the DRAM market. We really do believe that unit demand and end-user markets are very, very strong. All our forecasts show some sort of flattening or softness in the market for the next year or two, but every indication in the market is that we will see a pretty strong growth toward the latter part of the decade. So the truth of the matter is that the semiconductor revolution that we have experienced in the past two to three years is real, and today we are dealing with fluctuations at the top of a much larger pool.

MR. FUHS: Some of the products that may contribute to the advanced logic recovery might be some of the digital consumer products, DVD systems, digital cameras. These will help fuel the move to digital and are part of the underlying strength of the market. How that translates to equipment spending, those tend to be more secondary but related to the dynamic of why we believe advanced logic will probably be the first one to recover.

OPERATOR: #21, you may ask your question.

CALLER #21: We talk a lot about the downturn, but through 1997, there's still going to be about $5 billion of quarterly spending. I just wanted to get your perspective on who will be the remaining winners in that period: frontend versus back-end or higher technology. Who's going to actually be selling this $5 billion worth of eqmpment?

MR. FUHS: The industry continues to invest in infrastructure and capacity because the industry is growing, albeit at a slower growth rate. The industry just does not shut down to zero, but companies will be primarily investing in advanced technology systems, such as intermetal dielectric 0.35-micron "gap fill" such as Novellus' SPEED product, as an example of a classification of a product that would do well. Deep-UV steppers are actually going to grow significantly through this time period. We are showing the number of deep-

UV steppers growing from about 105 in 1996 to about the 270 level in 1998, and this is an example of a segment of the market that is showing strength.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide 15

What we are expecting is a shift to invest focused on trying to understand how to manufacture productively at 0.25-micron, and companies are going to invest in that technology and set up pilot production capacity for that.

MR. PAKDAMAN: In the area of process control, going beyond the 0.3micron realm will entail a leap into a lot of unknowns. This pause period should allow the semiconductor manufacturers to make investments to understand what will be required. What we are seeing in terms of metrology and all areas of process control, not just for R&D but for pilot production, is really getting ready for an environment of volume production at these advanced linewidths. Those are the key drivers for the investments that we were seeing in 1997 and even into the first part of 1998.

CALLER #21: Would you anticipate branching out or some experimentation with new processes to cut costs in the future like someone that might have modular portable clean areas to cut down overall cleaning room costs, things like that, that might change the process, but in the long run, be cheaper?

Would you anticipate much experimentation in this time period?

MR. FUHS: Yes, I think there will be some experimentation done. The production pressures are kind of off for the next couple of years. That is going to be an opening to perform that kind of an experimentation. As far as new processes and restructuring the fab, so to speak, there might be some of that. But I think that is going to be primarily in the R&D center first, early on and perhaps some of the fabs that are announced later in the cycle will implement some of those changes. Because the shells are being built and the fabs are being designed today, changing the structure of the fab is a little bit risky, a little bit slower to be accepted.

MR. PAKDAMAN: What we are starting to see and would expect in terms of investments have to do with statistical process control, and investments in information systems. We are seeing many of the common maladies of increasing capacity in a tight environment, where equipment is purchased to add capability rather than focusing on efficiency. Fabs are actually experiencing smaller throughputs because companies are not able to utilize the fab in the most efficient manner. With all of the past investment, there needs to be time now to go in and make sure that it is running efficiently. I think a lot of the investments, as far as changing how fabs operate, are going to be done as information systems and focus on trying to increase efficiency.

I do not think it will go as deep as trying to change processes. We're still dealing with a rather conservative manufacturing environment and things do not change as fast as they do in R&D in what is actually seen on the production floor.

OPERATOR: #5, you may ask your question.

CALLER #5:1 have a question about the Windows NT for impending release shortly. What effect does that have on demand and timing?

MR. FUHS: I'm not going to answer that question specifically with respect to how it will affect the capital spending recovery because I do not think it will have much of an impact. The issue with anything in the operating system

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

_16 Semiconductor Equipment, Manufacturing, and Materials Worldwide area is, will the consumer buy it? Will the business people buy it and the answer is, yes, in time. If it drives PC units, then it will make the capital spending recovery happen a little faster, perhaps. But then look at all the anticipation everybody had on demand for Windows 95.1 got a question last year at our Semicon/West seminar, "What do I think about the memory requirements for Windows 95 just blowing out the top of the memory demand?" My comment was, "I'm memory-constrained today because of the

PowerPoint presentation I've got to do rather than going to a new operating system." The applications are what is going to drive the memory requirements necessarily and the PC demand. A year after NT comes out, applications will be developed for the NT to require more memory. That is what is going to drive demand directly related to NT. That is the normal course of events in the PC industry driving how many bytes go into a PC.

OPERATOR: #28, you may ask your question.

CALLER #28: You mentioned that you expect R&D expenditure of semiconductor makers to be stable. Do you expect the same for R&D expenditure of semiconductor equipment manufacturers?

MR. FUHS: We would expect that to remain fairly stable. Some companies that are in fairly good shape would probably increase their R&D relative to sales in this time period.

CALLER #28:1 see. Thanks.

OPERATOR: #26, you may ask your question.

CALLER #26: I'm wondering, in Tables 1 and 2 you show from 1995 to 2001 that capital spending CAGR is greater than wafer fab equipment by about half a percent and in a couple of regions, Japan and Asia/Pacific, it's considerably higher, about 2-1/2 points higher. This is counter to my understanding of what's been going on in the last 10 years or so where an increasing portion of the capital spending is going toward wafer fab equipment, but you're showing that trend over a pretty long period of time here, six years, reversing. I wanted for you to comment on that. Is that something you think is going to continue?

MR. FUHS: Actually, it's remained fairly stable in history—I actually have that information here. By the way, that's been a consistent part of our forecast in every forecast we've published in the last two years. In the late

1980s, the average was just over 50 percent of the capital spending went into front-end, or wafer fab, equipment. In 1995, which was the peak year in the current cycle, it was slightly under 50 percent, so the trend in history is gradually declining, even if you go back further in history. The reasons for it are a couple. One has to do with the fact that the test and assembly part of the market, which is not in the wafer fab equipment number, but is in the capital spending number, includes testing equipment that has become more sophisticated. If I were to look at the equipment spending as a whole, I would probably say that it was stable relative to capital spending. Some of the assembly pieces of equipment and packaging issues associated with that are getting more complicated and complex, and that's growing a littie bit

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide 17_ faster. The second has to do with computer systems. Ten years ago, computer systems that were capitalized were a very small part of the total.

Today, they account for between 7 and 9 percent of the total capital spending dollar. That's the answer in a nutshell.

MR. FUHS: We're right at one hour now, and we have limited this to one hour. I apologize to anyone who did not get a chance to ask their question.

We are available to take questions after this call, but I want to thank you all for participating. I'd like to invite you to attend our Semicon/West seminar next week. It's on Thursday morning, July 18. For more information, please contact our conference desk at 800-457-8233 in the United States, or outside the U.S., the number is 805-298-3262. It will be held in the Sheraton Palace in

San Francisco. Thanks for participating. Good-bye now.

END OF CONFERENCE CALL.

SEMM-WW-DP-9606 ©1996 Dataquest September 16,1996

18 Semiconductor Equipment, Manufacturing, and Materials Worldwide

For More Information...

Clark Fuhs, Director/Principal Analyst (408) 468-8375

Internet address [email protected]

Via fax (408)954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness. It does not contain material provided to us in coi\fidence by our clients. Reproduction or

1^ y 3 T ? t C J l . I ^ S t *^<^l°s"''^ ^ whole or in part to other parties shall be made upon the written and express consent of Dataquest.

r> ^ ^ * S ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and Materials

Woridwide

Market Analysis

The Dynamic Etch Equipment Market: 1995 Market Review and

Projection

Abstract: After two years of spectacular growth, the dry etch market in 1995 continued its

skyward ascent with an eye-popping 79 percent growth, reaching U.S.$2.84 billion in sales and completing a three-year market expansion that resulted in the quadrupling of worldwide etch revenue. The market remains on track to achieve a solid growth in 1996. But

1997 will likely be a year of significant decline in the dry etch market as the chip industry continues to absorb the large capacity put on line during the fab-building frenzy of the prior years.

By Calvin Chang

Dry Etch Market Summary

The worldwide dry etch equipment market grew 78.5 percent in 1995, the strongest growth in more than 10 years. Booming demand for dry etchers in

1995 was driven by a combination of memory and microcomponent capacity expansion, as well as increased average selling prices (ASPs) in all segments of the dry etch product mix.

LU

N

Z

LU

>

Q .

^

8<

LLI CC

d <

Dielectric etch, including oxide and nitride applications, became a billiondollar market in its own right and contributed the most to the overall etch growth. A rise in polysilicon and silicon trench etcher sales was in large part driven by the seemingly incessant addition of new memory capacity construction. Metal etchers also turned in a respectable showing, although with a comparatively modest 63.8 percent growth. Table 1 presents the makeup of the 1995 dry etch market, segmented by film applications, and their respective growth rates.

DataQuest

Program: Semiconductor Equipment, Manufacturing, and iV^aterials Worldwide

Product Code: SEMM-WW-DP-9605

Publication Date: Septembers, 1996

Filing: Perspective

(For Cross-Technology, file in the Semiconductor Regional Markets and Manufacturing binder)

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 1

1995 Dry Etch Segment Sales and Growth (Millions of U.S. Dollars)

Dry Etch Segment

Estimated 1995 Sales*

Dielectric Etch

1,264.3

Polysilicon/Silicon Trench

913.5

Metal Etch

663.9

Total Dry Etch

2,841.8

*Sales are for calendar year 1995 and exclude service and spares.

Source: Dataquest (August 1996)

Percentage of Overall

Dry Etch Market

44.5

32.1

23.4

100.0

1994-1995

Growth (%)

87.3

78.7

63.8

78.6

Polysilicon Etch

The polysilicon etch market surged in 1995 as capacity was added to strengthen DRAM production. Nearly all new DRAM capacity added during the year was designated for 16Mb/64Mb production, usually with a minimum of four levels of polysilicon to etch. Riding on the success of its

TCP 9400 polyetcher. Lam Research has gained the leadership spot with a rising market share in high-density polyetch, which now represents 32 percent of the overall polysilicon etch market (from 21 percent just two years ago). In low-density polysilicon etch. Applied Materials' workhorse P-5000 became the undisputed market leader, having taking the advantage of Lam's shifting focus toward the higher-ASP high-density etch market. In total polysilicon etch sales, including both high-density and low-density polyetch,

Lam and Applied Materials were neck and neck, each with about a 36 percent share of the market. Hitachi was a distant third with a 13.3 percent share. With single-digit market shares were Tokyo Electron Ltd. (TEL), Tegal

Corporation, and Sumitomo Metals, although all have increased sales in polysilicon etch during the year.

Long absent from the high-density polysilicon etch market. Applied

Materials has completed its lineup of high-density etchers with its recent unveiling of the Silicon Etch DPS (Decoupled Plasma Source) Centura. A lot is riding on the DPS; Applied hopes that the new high-density source will allow it to extend its low-density dominance into the faster-grovdng highdensity market.

Metal Etch

Applied Materials became the market leader in overall metal etch sales in

1995, displacing Lam Research, the metal etch leader of the previous year.

Applied accomplished this by increasing its dominance in the low-density metal etch segment, where it attained nearly 74 percent in market share. The second-place low-density metal supplier. Lam Research, in contrast, had only a 12 percent share. Here again. Lam's conscious shift toward the highdensity etch market is evident. Lam's shortfall in low-density metal was more than made up in high-density metal, where its share reached more than 50 percent, landing it the No. 1 spot from the previous high-density metal leader, Hitachi.

SEMM-WW-DP-9605 ©1996 Dataquest September 9,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Metal etch is where high-density has attained the highest penetration, contributing more than 52 percent of the metal etch market, and the battle for future metal etch market share will be fought in the high-density arena.

Although Lam has sprung ahead of Hitachi in high-density metal, Hitachi still has a formidable presence with its ownership of the largest high-density metal etcher installed base. Early this year. Applied Materials formally entered the high-density metal market with its introduction of the Metal Etch

DPS Centura system. The battle for metal etch market share for the next few years will be fought among the three etch titans. Tegal, Sumitomo Metals, and the few other metal etch suppliers, however, will likely be significant but will remain niche players.

Dielectric Etcli

Tokyo Electron, with its successful concentration of sales efforts into the

Asia/Pacific region, retains its top spot in dielectric etch, with 34 percent of the market. Lam remains in the runner-up position, holding onto a 27.7 percent share. Breathing down Lam's neck, and getting closer, is Applied

Materials, with 27.1 percent. These three vendors will continue to be fierce competitors in the largest etch segment. Dielectric etch accounts for 44 percent of the total etch market and is poised to take on a much greater share when the chip industry adopts damascene processes, starting in several years. High-density oxide etch will be a key technology enabler for the damascene process. Lam's recently introduced TCP 9100, targeting critical oxide etch applications below 0.35 micron, is the second serious contender for the high-density oxide market, after Applied's HDP Dielectric Etch

Centura, which was introduced some three years ago. Compared with polysilicon and metal etch, high-density dielectric etch has been a laggard, with relatively meager 1995 sales of U.S.$119 million, representing only 15 percent of the high-density etch market. A principal reason is the limited choice of truly production-class tools. Lam's TCP 9100 and the new oxide entrants expected from other vendors within the next year could be what it takes to jump-start the high-density oxide market.

Table 2 presents 1995 dry etch revenue for the major suppliers. Low-density, high-density, and total dry etch sales results are shown with the percentage change in 1995 sales compared to 1994. The year 1995 was successful for the entire dry etch industry, with every vendor, big or small, reporting doubledigit to triple-digit growth in sales.

Regional Segmentation

Figure 1 presents the change in the regional segmentation of the total dry etch equipment market from 1994 to 1995. Japan continues to be the largest etch market at $988 million, or 34.8 percent of the world market, down from

36 percent in 1994. The Americas region's share also declined, to less than 23 percent, from 26 percent in the previous year. The decline of the Japan and

Americas' share of the etch consumption market is in contrast to the continuing rise of Asia/Pacific, which has now become the second-largest

SEMM-WW-DP-9605 ©1996 Dataquest September 9,1996

Semiconductor Equipment, Manufacturing, andMateriais Worldwide etch market. Sales of dry etchers in Asia/Pacific reached U.S.$831.3 million, or 29.3 percent of the total market, a far cry from the 12.7 percent the region represented in the world market only three years ago. Large capacity expansions, including DRAM and foundry in Korea and Taiwan, drove the bulk of the etch sales increase in Asia/Pacific. Europe's contribution to the etch consumption market remains steady at levels just a shade above the

1994 levels.

Table 2

Estimated 1995 Sales and Rankings for Dry Etch Equipment Manufacturers

(Millions of U.S. Dollars)

1995

Rank

1994

Rank

Company

Low-Density Dry Etch Eqmpment

3

4

1

2

1

2

3

4

Applied Materials

Lam Research

Tokyo Electron Ltd.

Tegal

Others

Total Low-Density Dry Etch

Estimated

1994 Sales

389.9

346.9

279.8

28.9

94.1

1,139.6

Estimated

1995 Sales

832.1

532.5

481.2

34.4

152.5

2,032.7

Change (%)

113

54

72

19

62

78

High-Density Dry Etch

Equipment

1 2 Lam Research

2

3

4

5

1

3

4

5

Hitachi

Applied Materials

Sumitomo Metals

Tegal

Others

Total High-Density Dry Etch

Total Dry Etch

Eq

uipment

1

2

3

4

5

2

1

3

4

Applied Materials

Lam Research

Tokyo Electron Ltd.

Hitachi

6

6

5

Sumitomo Metals

Tegal

Others

Total Dry Etch

*Note: Safes are for calendar year 1995, and exclude service and spares

Source: Dataquest (August 1996)

415.0

510.7

279.8

208.4

22.2

36.2

119.3

1,591.6

163.8

207.9

25.1

22.2

7.3

25.7

452.0

904.5

881.5

481.2

285.5

51.2

46.8

191.1

2,841.8

349.0

285.5

72.4

51.2

12.4

38.6

809.1

118

73

72

37

131

29

60

79

113

37

188

131

70

50

79

SEI\/iM-WW-DP-9605

©1996 Dataquest

September 9,1996

Semiconductor Equipment, IVIanufacturing, and Materials Worldwide

Figure 1

Change in Regional Contributions to the Worldwide Total Dry Etch Market, 1994-1995

15

10

5

0

Percentage of Worldwide Market

40

35

30

25

20

Asia/Pacific

Total Diy Etch,

1994

Total Dry Etch,

1995

Europe Americas

Japan

965886

Source: Dataquest (August 1996)

Low-density etch in 1995 constituted 71.5 percent of the total dry etch market. Asia/Pacific jumped ahead of Japan to become the largest lowdensity etch market, with 32 percent of the worldwide market. Americas lost a significant share to end up with a little over 22 percent of the total lowdensity revenue. Europe grew modestly to finish 1995 with 14.4 percent of the market. Figure 2 shows the change from 1994 to 1995 in regional contributions to the worldwide low-density etch market. Figure 3 shows the corresponding distribution for the high-density etch market. All four regions in 1995 maintained their respective share of the worldwide high-density etch market. In terms of their percentage contribution to the high-density etch market, Americas and Asia/Pacific grew just a shade, while Japan and

Europe gave up a modest amount.

Dry Etch Equipment Forecast

Dry etch equipment market growth is moderating after three years of dizzying megagrowth Dataquest estimates that the worldwide dry etch equipment market will grow by 17 percent on a revenue basis in 1996 compared to 1995, versus a 17.1 percent rise in the overall wafer fab equipment market. Although Dataquest expects the dry etch market, as well as the overall equipment market, to show a significant contraction in 1997, the long-term outlook is very positive for the dry etch market.

SEMM-WW-DP-9605

©1996 Dataquest

September 9,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 2

Change in Regional Contributions to the Worldwide Low-Density Dry Etch Market,

1994-1995

Percentage of Worldwide Market

35

Low Density,

1994

Low Density,

1995

Asia/Pacific

Source: Dataquest (August 1996)

Europe Americas Japan

965SB7

Figure 3

Change in Regional Contributions to the Worldwide High-Density Dry Etch Market,

1994-1995

15

10^

5

0

Percentage of Worldwide Market

45-

40

35-j

30-1

25

20

High Density, 1994

High Density, 1995

Asia/Pacific

Europe

Americas Japan

965888

Source: Dataquest (August 1996)

As device geometries shrink and the manufacturing process flow becomes more complex, the number of dry etch steps required in the process grows.

More dry etch steps means greater unit demand of dry etch systems. Also, in the sub-0.5-micron linewidth regime, high-density plasma etchers will be

SEMM-WW-DP-9605

©1996 Dataquest

September 9,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide imperative for critical etch applications. The adoption of high-density etch systems will displace the traditional low-density reactive ion etch (RIE) etch sources and provide the impetus for continuing growth in etch system shipments. The requirement of high-density etch in all film etch applications will also resxilt in increases in average selling prices. Thus, growth in unit system demand and the higher ASPs that accompany the higher mix of advanced etch systems are the principal driving forces for the long-term growth in dry etch sales.

Table 3 presents Dataquest's forecast for dry etch equipment sales, segmented by film application. The year 1996 is shaping up to be solid, with growth expected in every dry etch segment except silicon trench etch. The current slowdown in DRAM investment is putting a damper on silicon trench etch, almost exclusively used for fabrication of DRAM trench capacitors. Looking beyond the slump of 1996 through 1997, however, silicon trench etch growth is projected to pick up steam quickly from the adoption of shallow trench isolation as well as the continuing use of trench capacitors in next-generation DRAM. Dry etch for polysilicon and polysilicon/silidde stack etch applications is also a DRAM-sensitive market.

Dataquest projects a significant slowdown during the next couple of years while the chip manufacturers are busy absorbing the large memory IC capacity put on line during the fab-building frenzy of the past few years.

Investment should begin to kick in for the 64Mb/256Mb generations of

DRAM in 1999. Projected growth for polysilicon/polycide dry etch from

1995 to 2001 is a healthy 15.2 percent compound annual growth rate

(CAGR).

Table 3

Worldwide Dry Etch Equipment Market History and Forecast, by Film Tj^e

(Millions of U.S. Dollars)

Oxides

Plasma Nitride

LPCVD Nitride

Others

1993

376.8

15.9

23.9

12.2

1994

600.8

24.6

34.3

15.4

1995

46.7

64.4

18.9

1996

51.9

73.1

19.9

1997

1,134.4 1,370.8 1,191.5

46.4

64.1

14.4

1998

1,217.0

46.4

62.3

14.8

1999

1,485.4 2,132.5

58.9

74.6

14.6

2000

77.9

100.5

16.9

2001

2,822.0

99.7

128.6

19.7

CAGR (%)

1995-2001

16.4

13.5

12.2

0.7

Metal

339.4 511.1 913.5

1,033.8 884.3

910.4 1,106.0 1517.7 1,928.6

Poly/Polycide 302.4

Silicon/Trench

12.3

Total Dry Etch Market

Source: Dataquest (August 1996)

1,083.0

386.3

19.1

625.9

38.0

744.6

29.9

646.0

24.4

1,591.6 2,841.8 3,324.0 2,871.0

662.3

40.8

2,954.0

846.6

52.0

1,210.1

71.8

1,462.2

99.1

3,638.0 5,127.3 6,559.8

1 3 3

15.2

17.3

15.0

As shown in Table 3, oxides etch is expected to remain the largest etch application segment and is projected to grow at a healthy clip, with a 16.4 percent CAGR from 1995 through 2001.

SEMM-WW-DP-9605 ©1996 Dataquest September 9,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

The trend in the semiconductor industry toward increasing use of multilevel intercoimection is driving the demand for oxide. The number of interconnect layers in logic/ASIC devices could rise from an average of three layers in

1996 to five layers by the year 2000. In DRAM manufacturing, three interconnect layers, as opposed to today's two layers, will begin to be introduced at tiie shrink 64Mb generation a few years from now. Moreover,

ASPs for oxide etch systems will continue to rise, with new advanced oxide etchers delivering superior performance that permits etch of ever-smaller contact and via holes with good profile control and selectivity. Increases in

ASP and unit shipments of oxide etch are expected to help dielectric etch

(including oxide, LPCVD nitride, PECVD nitride, and others) to raise its contribution to the overall dry etch market revenue, achieving more than

U.S.$3 billion in sales, or 47 percent of the total etch market, by the year 2001.

A major concern surrounding the metal etch technology in the sub-0.5micron regime is the reliability of metal interconnect. Multilayer stacked structures consisting of different metal films are now commonplace. Many of these films require different etch chemistries integrated in a flexible, multichamber platform. Also reqtiired is an in-situ clean module for the prevention of corrosion from etch residues. The increasing technology demand will drive the ASPs of future metal etch offerings. This, in conjunction with the continued rise in interconnect levels that drives the unit demand of metal etcher, bodes well for the overall health of the metal etch market for the remainder of the 1990s.

The long-term threat to metal etch is the possibility of an eventual adoption of (dual) damascene processes in semiconductor manufacturing. The rapid success in recent years of chemical mechanical polishing as an accepted planarization technology is fueling the belief that, as device geometries continue to shrink, fabrication of metal interconnect will be easier to achieve by deposition rather than by dry etch. It is almost a foregone conclusion that damascene processes will eventually be adopted in IC manufacturing, initially in advanced logic/ASIC devices and later in other devices. The question is when damascene will enter into production, and, hence, when metal etch sales will begin to be impacted. Dataquest believes that implementation of damascene processes will most likely begin at the 0.18micron generation, which is projected to enter the scene after the year 2000.

This means that there should be at least five more years of stable market opportunities for metal etch systems.

Mix and Match: High-Density versus Low-Density Etch

The dry etch market is now a two-tier market, with growth split between the conventional low-density plasma etch systems and advanced, high-density plasma etchers. The segmentation between low-density and high-density etch applications falls at about the 0.5-micron linewidth level.

The 4Mb/16Mb DRAM capacity expansions of recent years have boosted the demand for low-density etchers capable of fabricating features down to

SEI\/IM-WW-DP-9605 ©1996 Dataquest September 9,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide about the 0.5-micron level. For 0.35-micron etch and critical etch applications that require stringent profile control, selectivity, microloading, and pattern sensitivity, high-density etchers are imperative. High-density etchers, sporting a variety of sources (for example, Inductive-Coupled Plasma,

Trarisformer-Coupled Plasma, High-Density Reflected Electron, and

Decoupled Plasma Source), must be capable of operating in low pressure

(from a few to 10 mTorr) without compromising wafer throughput.

Dataquest believes that high-density etch will be the principal beneficiary of the growth in dry etch as device geometries continue to scale down. Figure 4 projects the worldwide dry etch equipment market, segmented by lowdensity and high-density etch sales.

Figure 4

Worldwide Dry Etch Equipment Market Segmented by Technology

Millions of U.S. Dollars

/

3,000-

High-Density Dry Etchers

2,500 -

2,000-

V ^ /

/ ~ /

1,500-

1,000-

^ ^ ^ ^ ^ ^

5 0 0 -

0 -

1991

1 1 1 1 1 1 1 1 1

1992

1993

1994 1995

1996 1997 1998 1999 2000 2001

965889

Source: Dataquest (August 1996)

During the first half of the 1990s, high-density etch emerged from a modest beginning to become a vital part of the etch market, garnering more than 28 percent of total dry etch revenue in 1995. With the adoption of the 0.35micron/0.25-micron processes and the spread of critical etch applications driving high-density etch demand, high density sales are expected to grow rapidly and capture nearly half of all dry etch sales by the year 2000.

Although the wind is filling the sails of high-density etch, low-density etch will continue to be a mainstay of seiniconductor manufacturing. Future semiconductor processes will consist of a combination of critical and noncritical steps. While the number of noncritical etch steps will decline as a percentage of overall process sequences, the decline will moderate and

SEMM-WW-DP-9605

©1996 Dataquest September 9,1996

10 Semiconductor Equipment, Manufacturing, and Materials Worldwide eventually taper to a stable level. The mix and match of high-density and low-density etchers to achieve the optimal overall productivity and cost of ownership will be the overriding theme that shapes the dry etch market.

For More Information...

Calvin Chang, Senior Industry Analyst (408) 468-8605

Internet address [email protected]

Via fax (408)954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness. It does not contain material provided to us in confidence by our clients. Reproduction or

f\rM^fi^g^m mg^C't' '^^^°^'^^^ ^ whole or in part to other parties shall be made upon the written and express consent of Dataquest.

mJkX l < M k . ^ U W 3 L ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and Materials

Worldwide

Market Analysis

Polysilicon Supply/Demand Update: The Industry Responds and

Relief Is in Sight, but only after 1997

Abstract: Polysilicon, the razv material for all silicon wafers supplying the semiconductor

industry, has been under threat of shortage. In a Focus Report published in November 1995,

Dataquest outlined the details of an impending shortage starting late in 1996. With the

recent downgrading of semiconductor forecasts, the ifuestion iswill this shortage happen?

With a continued strong wafer forecast driven by unit demand primarily, the shortage in

polysilicon is inevitable. Since the report in November, there have been several announce-

ments for increased polysilicon capacity. In this article, Dataquest provides an update to the supply and demand outlook for polysilicon.

By Clark ]. Fuhs

Why is Polysilicon Production Critical?

All single crystal silicon wafers manufactured for the semiconductor industry come from the same raw material—pure silicon. Industrial metal grade is not pure enough for semiconductor applications, so the silicon must be purified through chemical processing. The method initially converts the industrial-grade metal silicon to a gaseous or liquid silicon compound, such as silane or trichlorosilane, and then processes the chemical through distillation or another purification method to produce a very high-grade chemical. Once purified, the silicon-containing material is then broken down to its components through a form of thermal decomposition to produce semiconductor-grade silicon in the shape of rods or granules.

Because this crystallization process produces a polycrystalline material, the industry refers to this material as polysilicon. This material, after being

DataQuest

Program: Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Product Code: SEMM-WW-DP-9604

Publication Date: July 22, 1996

Filing: Perspective

(For Cross-Technology, file in the Semiconductor Regional IVIarkets and Manufacturing binder.)

Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide broken into usable-size pieces, is purchased by wafer manufacturers to melt before the single crystal growing process. The purification process is a semiconductor industry requirement, and therefore it becomes a potential bottleneck in the production of silicon wafers.

Supply-Side Forecast Through the Year 2001

Based on known and estimated expansion plans, Dataquest has developed a supply forecast for polysilicon production through the year 2001. This forecast is shown in Table 1. It should be noted that these figures refer to actual annual production estimates and not facility capacity at the end of the year. In many cases, we have factored ramp-up schedules to arrive at annual production estimates. We should note that these figures do not include production out of Ukraine, which has recently started again.

Dataquest estimates current production at 100 tons per year, but we are unclear on ramp plans.

Demand Forecast for Polysilicon

Before presenting our actual demand forecast, a discussion about the approach, methodology, and assumptions used is in order. All our forecast analyses have as a starting basis the wafer slice forecast by wafer size, presented in Table 2 and in Dataquest's midyear 1996 forecast Market Trends

Report (SEMM-WW-MT-9602, expected in early August 1996).

Using this data and forecast as a basis, the next step was to apply an estimate for the average number of grams of polysilicon used to manufacture one good polished wafer. In terms of grams per square inch, this figure varies among the different wafer sizes. For example, just over 3.1 grams per square inch was assumed for 100mm wafers, increasing to about 4.4 grams per square inch for 200mm wafers. The average per year varies based on the wafer size mix in the forecast. When these assumed values are multiplied and added, the result is gross polysilicon demand in kilotons per year.

Factors to Consider: Test Wafer Use and Wire Saws

In Dataquest's November report, we suggested two factors that could reduce polysilicon demand from the above methodology. The first was demand-related—the reduction in consumption of virgin test wafers, either by reduced test wafer use or by the alternate use of reclaimed wafers. For this update, we have factored in this reduced use in the overall silicon and wafer size forecast and therefore do not need to consider it in a special way here.

The second factor was the implementation of wire saw technology in the production process, which has the capability of increasing ingot yield. This was technology in its infancy a year ago, but technology leaders such as

Shin-Etsu Handotai and Wacker Siltronic are understood to be applying the technology fairly aggressively today. In the process of making wafers, the grown ingot needs to be sliced into individual wafers. The most common approach is to use a diamond edge rotary saw with the cutting edge in the interior diameter for vibration stability, commonly referred to as an ID saw.

In any sawing operation, sawdust is created, referred to as "kerf" loss, representing wasted polysilicon material.

SEMI\^-WW-DP-9604 ©1996 Dataquest July 22,1996

Semiconductor Equipment, IVlanufacturing, and IVIateriais Worldwide

Table 1

Worldwide Actual and Projected Polysilicon Annual Production by Supplier, 1993-2001

(Thousands of Metric Tons per Year)

1992 1993

U.S. Production

Advanced Silicon Materials

Hemlock Semiconductor

MEMC Granular Polysilicon

Mitsubishi Materials

Japan Production

Kojundo Silicon (Hi-Silicon)

Komatsu Electronic Metals

Sumitomo Sitix

Tokuyama Corporation

European Production

MEMC (Huls)

Wacker Siltronic

Worldwide Total

Source: Dataquest (July 1996)

1.10

1.40

0.60

0

0.80

0.03

0.50

1.10

0.40

2.30

8.23

1.10

1.40

0.65

0

0.90

0

0.45

1.15

0.70

2.30

8.65

1994

1.10

2.30

0.85

0

1.10

0

0.45

1.65

0.75

2.40

10.60

1995

1.20

2.80

1.10

0

1.20

0

0.45

1.65

1996

1.60

3.40

1.30

0

1.40

0

0.45

2.20

1997

2.10

4.20

1.90

0

1.40

0

0.70

3.00

1998

3.20

5.00

2.50

0.80

1.40

0

0.70

3.00

1999

4.00

5.40

2.70

1.50

1.40

0

0.70

3.00

2000

5.40

6.00

2.70

1.50

1.40

0

0.70

3.50

2001

5.90

6.60

2.70

1.50

1.40

0

0.70

4.00

0.75

2.40

11.55

0.75

2.50

13.60

0.90

2.75

16.95

0.90

3.35

20.85

0.90

3.60

23.20

0.90

5.00

27.10

0.90

5.00

28.70

Table 2

Worldwide Wafer Forecast by Diameter, 1993-2001 (Millions of Wafer Slices per Year)

1996 1997 1998 1999 2000 2001 1992

Unit Distribution by Wafer

Starts (Millions of Wafers)

2 inches

3 inches

100mm

125mm

150mm

200mm

300mm

Total Wafers (M)

Average Wafer Diameter

(Inches)

Source: Dataquest (July 1996)

110.3

4.92

1.0

6.3

32.3

36.3

33.2

1.2

1993

1.0

5.6

32.2

37.0

42.0

3.2

0

121.0

5.07

1994

5.21

1995

0.9

5.6

35.3

37.7

49.9

7.3

0

136.6

0.9

5.5

37.9

39.3

58.1

14.0

0

155.7

5.37

0.4

5.2

34.4

37.8

64.0

24.1

0

165.9

5.61

0.3

4.7

32.2

35.5

66.3

33.4

0

172.4

0.2

4.5

30.7

34.3

68.1

41.5

0

179.4

0

4.3

29.9

33.1

71.7

50.2

0.2

189.4

0

3.7

30.5

33.8

83.0

58.8

0.4

210.2

0

3.6

29.9

33.3

96.3

65.5

0.7

229.3

5.79

5.92

6.05

6.14

6.20

Thin metal wire rotating at high speeds, in conjunction with an abrasive slurry/coolant, is the core concept of the wire saw. The cut consumes less distance between wafer slices, with less wasted material, and therefore polysilicon use becomes more efficient in terms of grams per wafer. We have assiHned that kerf losses can be reduced by about 30 percent (we have heard of reductions from 20 to 50 percent).

SEI\/IM-WW-DP-9604

©1996 Dataquest July 22,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Probable Demand Scenario Compared with Supply

Wire saw technology is factored into our polysilicon demand scenario; however, the technology will not be fully implemented across the entire installed production base for 200mm wafers and will be only moderately implemented into 150mm and smaller wafers (see Table 3).

By comparing the supply from Table 1 with the demand methodology and assumptions, we can create a supply and demand analysis that calculates either a shortage or the industry inventory levels from 1995 to 2001. This is shown in Table 4.

Dataquest estimates that the industry's stock inventory of polysilicon was

2,730 metric tons at the end of 1994 (1,730 in Japan), down to about 900 metric tons at the end of 1995 (725 in Japan). By the middle of 1996, we would expect this inventory to have been depleted, and indeed we are starting to hear of instances in which some wafer manufacturers will not be getting all the polysilicon they desire in the coming year We believe these instances may be limited to smaller companies and companies that do not have an internal source of supply.

Table 3

Wire Saw Technology Market Penetration Assumptions

(Percentage of Capacity at a Particular Diameter)

1992 1993 1994 1995 1996

Percentage of Market

by

Wafer Size

2 inches

3 inches

100mm

125mm

150mm

200mm

300mm

Source: Dataquest (July 1996)

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

5

0

0

0

0

0 0

2

20

0

0

0

0

1997

2

4

45

0

0

0

2

1998

5

8

60

100

0

4

5

1999

5

10

75

100

0

4

5

2000

5

10

85

100

0

4

5

2001

5

10

90

100

0

4

5

Table 4

Polysilicon Supply/Demand Scenario—Probable Case (Wire Saw Technology Impact

Included; Metric Kilotons per Year)

Polysilicon Demand

Polysilicon Production

Net Annual Surplus

(or Deficit)

Industry Stock

(or Shortage)

Industry Stock Inventory

(Months)

Source: Dataquest (July 1996)

1992

7.60

8.23

1993

9.01

8.65

0.63 -0.36

1994

10.88

10.60

1995

13.38

11.55

-0.28

2.73

1.7

-1.83

0.90

0.7

1996

15.72

13.60

-2.12

-1.22

0

1997

17.24

16.95

1998

18.52

20.85

-0.29

-0.29

0

2.33

2.33

1.4

1999

20.16

23.20

2000

22.84

27.10

2001

25.35

28.70

3.04

5.37

3.0

4.26

9.63

4.8

3.35

12.98

6.1

SEMM-WW-DP-9604 ©1996 Dataquest July 22,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

By 1998, however, the production of polysilicon will again be above projected consumption because the polysilicon suppliers have stepped up investment in capacity. Announcements in the last nine months include

Advanced Silicon Materials, MEMC Granular Polysilicon, and Mitsubishi

Materials in the United States, and Wacker Siltronic in Europe. These capacity additions will put polysilicon production into the safe zone near the end of 1997, and we would expect polysilicon prices to ease somewhat at that time. However, we would expect buyers of polysilicon to maintain the position of purchasing all that is produced until 1999 in order to build the industry's capacity to a more comfortable two-month to three-month level, meaning that more significant price reductions will wait vintil then.

One interesting thing to note is that the industry has the capability to swing back into glut by the year 2000 should all plants announced be fully ramped. The price pressures from such a condition would cause investments in greenfield polysilicon plants to have vmfavorable returns.

Dataquest believes that polysilicon producers are likely to cut back the expansion plans noted in Table 1 after 1999 for these reasons.

A Look at the Highest Demand Scenario

Just for curiosity, what would the supply and demand picture look like if wire saws were not implemented beyond the 1995 level? This is a highly unlikely scenario, but it is one that would put an upper limit on polysilicon demand, in our view. This supply and maximum demand picture is presented in Table 5.

Clearly, this maximum demand picture represents a case for tight polysilicon supply until the year 2000, but it also shows that the plans for capacity additions already announced will adequately cover this demand. It is now highly probable that the next announcement in the polysilicon capacity arena will be about reduction rather than expansion.

Table 5

Polysilicon Supply/Demand Scenario—Maximum Demand Case

(Minimum Wire Saw Technology Impact; Metric Kilotons per Year)

Polysilicon D e m a n d

Polysilicon P r o d u c t i o n

N e t A n n u a l S u r p l u s

(or Deficit)

I n d u s t r y Stock

(or S h o r t a g e )

I n d u s t r y Stock I n v e n t o r y

( M o n t h s )

Source: Dataquest (July 1996)

1992

7.60

8.23

1993

9.01

8.65

1994

10.88

10.60

1995

13.38

11.55

1996

15.95

13.60

0.63

-0.36

-0.28

2.73

2.7

-1.83

0.90

0.7

-2.35

-1.45

0

1997

17.94

16.95

1998

19.72

20.85

1999

21.93

23.20

-0.99

-0.99

0

1.13

1.13

0.7

1.27

2.40

1.2

2000

25.16

27.10

2001

28.10

28.70

1.94

4.33

2.0

0.60

4.94

2.1

SEMM-WW-DP-9604 ©1996 Dataquest July 22.1996

Semiconductor Equipment, IVIanufacturing, and IViaterials Worldwide

Dataquest Perspective

With wafer unit demand forecasts intact, the inevitable shortage in polysilicon will happen later this year. The fact remains that industry inventories of polysilicon raw material have decreased in 1994 and 1995 because consumption exceeded production. Industry inventories in polysilicon will be exhausted during 1996. PC unit growth would have to decline to about

5 percent in 1996 to avoid this situation (which, by the way, would tend to drive the semiconductor market to an actual decline of more than 10 percent in 1996). Dataquest just does not see this happening.

Since the report on the supply/demand situation in November, there have been several announcements about increased polysilicon capacity. This will drive production to exceed consumption starting in late 1997 and continuing through the decade. However, the industry operates efficiently with some inventory in the channel, and it will take until 1999 for the industry to get back to the two-month inventory level considered to be the minimum for comfortable operation. So we would expect tight (although not short) conditions to exist in the market through 1999.

It appears that our forecast last November of a polysilicon shortage starting the second half of 1996 but lasting only a relatively short time (one year) is coming to pass. The industry effectively responded without the introduction of new suppliers into the market, thereby confirming a strong industry infrastructure. It is also interesting that most of the expansions and new plants are in the United States, which we believe is a response driven by two factors: the lower cost of capital in the United States than in Japan and the migration of production capacity recently of both fabs and wafer manufacturing facilities by foreign companies to the United States.

Will the impending polysilicon shortage hamper semiconductor industry growth? The danger is there, but we stop short of saying that semiconductor market revenue will suffer purely as a result of this particular shortage.

The semiconductor companies will do everything in their power to prevent that. What can they do? Consumption of test wafers will be reduced relative to prime wafers, forced by the lack of availability. We suspect that specifications for prime wafers will have to be loosened temporarily in order to accept more of the production mix into prime product wafer starts.

As virgin test wafer consumption is forced, the reclaimed wafer market should continue booming, although this should subside by the end of 1997.

The initial shift may cause yields to suffer temporarily, but semiconductor companies will choose this route over wasting the wafer for no revenue.

A Review of Poiysiiicon Producers

There are only a handful of polysilicon producers in the world, and the top two producers supplied just about 45 percent of the world's requirements in 1995. The suppliers fit into three basic categories: captive, hybrid-captive, and merchant.

Captive Producers

Captive producers are silicon wafer manufacturers that have polysilicon production within their companies and consume virtually all the material internally. Sumitomo Sitix and MEMC Electronic Materials (Hiils) are

SEIWM-WW-DP-9604 ©1996 Dataquest July 22,1996

Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide considered captive producers by Dataquest. Also, Mitsubishi Materials has just announced a new joint-venture plant in the United States starting production in 1998. For now, Dataquest classifies this as a captive producer, although in the future a portion of this production may be made available to companies outside Mitsubishi Materials. All these producers use the trichlorosilane process in the production of polysilicon.

Hybrid-Captive Producers

The category of hybrid-captives designates silicon companies that produce polysilicon and that consume the majority of their production. A minority portion, therefore, is sold to other silicon wafer producers. However, for this class of producer, distribution outside the company is limited to a small number—either the amount of polysilicon or the number of companies to which polysilicon is sold. Two companies fall into this category.

MEMC Granular Polysilicon (formerly Albemarle Corporation)

The Granular Polysilicon division of MEMC Electronic Materials is the only producer of polysilicon in granular, or pellet, form. The nature of the proprietary fluidized bed reaction process creates polysilicon in this form rather than conventional rod or block form. Albemarle first commercialized this process in 1987 when it was a division of Ethyl Corporation. The plant in Pasadena, Texas, and the polysilicon operations were spun off to

Albemarle Corporation in 1993. In 1995, MEMC acquired 85 percent of the operation and has restructured it as the Granular Polysilicon division of

MEMC. Because this form of polysilicon requires special feeding equipment to be attached to the crystal growers, the material has gained only limited acceptance, and during 1995 it is estimated that MEMC has consumed 90 percent of the production from this plant. Because the amount of material now being shipped outside of MEMC represents less than 10 percent,

Dataquest considers this operation to be a hybrid-captive. We understand that the granular polysilicon produced will be available for sale outside of

MEMC, although it is expected that the vast majority of production will continue to be locked up by MEMC.

MEMC has just announced that it will invest in more internal production from this facility for production starting in 1997. The key advantages of the granular form are believed to be clearly evident for 300mm wafers, but we do not expect demand for polysilicon driven by 300mm wafers to be significant before 2002. After then, we would expect MEMC to aggressively expand the capacity of this operation.

Kojundo Silicon (Hi-Silicon)

Hi-Silicon Co. Ltd. (also known as Kojundo Silicon) is located in Yokkaichi

City, Mie Prefecture, Japan. Hi-Silicon was formed in 1967 as a 50-50 joint venture between Mitsubishi Metal (now Mitsubishi Materials) and Osaka

Titanium Company (now Sumitomo Sitix). In October 1987, Osaka Titanium announced that it had sold its 50 percent position in Hi-Silicon to Mitsubishi Metal for an undisclosed sum. Mitsubishi Materials Silicon currently consumes more than half the production from this operation, with the remainder distributed to a limited number of companies. For this reason, Dataquest considers Hi-Silicon a hybrid-captive producer.

Dataquest estimates that Hi-Silicon's production in 1995 expanded slightly to 1,200 tons, which will then grow and be capped at 1,400 tons in 1996 through 1998. Hi-Silicon uses the trichlorosilane process in producing

SEMM-WW-DP-9604 ©1996 Dataquest July 22,1996

Semiconductor Equipment, IWanufacturing, and iVIaterials Worldwide polysiHcon. A major expansion is forecast, with production starting to ramp during 1999.

Merchant Producers

Merchant producers are those companies that have a wide distribution of siHcon wafer companies as customers. As is customary in this industry, consumption of polysilicon can be concentrated and can account for a large block of production capacity for even the largest producer. There are four merchant producers, and these four also happen to rank as the four largest producers.

Advanced Silicon IMaterials, inc.

Even though Advanced Silicon Materials (ASiMI) is owned by Komatsu, a silicon wafer manufacturer, and Komatsu consumes a significant portion of capacity (but estimated at less than 50 percent), ASiMI's wide distribution qualifies it as a merchant supplier with over a dozen outside customers.

ASiMI's plant is located in the Pacific Northwest region of the United

States, in Moses Lake, Washington.

ASiMI manufactures polysilicon by a proprietary method that converts trichlorosilane first to monosilane gas. This method, known as the Komatsu method, refines the silane gas to high purity, then decomposes it into very high-purity polycrystalline silicon material. This high-grade material has advantages in the float-zone crystal growing process. An example of devices that use float-zone crystal is high-voltage power devices.

Dataquest estimates that ASiMI is running its plant at essentially full capacity, producing 1,200 metric tons during 1995. ASiMI is expanding its facilities to a 2,100-metric-ton capacity, to be completed later this year, and has announced a major new facility in Montana starting production in 1998.

When both phases of the new plant are complete, it is expected that ASiMI's capacity will be 5,900 metric tons for all plants. It will be after 2000 before this capacity is realized, however.

In addition to its internal use of silane in polysilicon production, ASiMI is also a major supplier of high-purity silane to the semiconductor industry for processing applications such as epitaxy and chemical vapor deposition

(CVD).

Hemiocic Semiconductor Corporation

Initially formed as a wholly owned subsidiary of Dow-Coming, Hemlock built its first polysilicon plant starting production in 1963. Hemlock gets its name from the city where the plant was built—Hemlock, Michigan, a bit northwest of Detroit. In 1984, Hemlock became a joint-venture company, with two major silicon wafer manufacturers purchasing interests in the company. Shin-Etsu Handotai and Mitsubishi Materials Silicon own

24.5 percent and 12.25 percent, respectively, while Dow-Coming retained

63.25 percent ownership.

Hemlock takes the honor of being the largest producer of polysilicon, with an estimated production of 2,800 tons in 1995 (about 24 percent of the world's production) and capacity of 3,400 tons by the end of the year.

Hemlock is expanding aggressively, bringing Phase I of an expansion fully on line in 1995, with a Phase II recently accelerated to be ready to start production at the beginning of 1997. No formal commitment has been made

SEMM-WW-DP-9604 ©1996 Dataquest July 22,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide to Phase III, but Dataquest estimates that Hemlock will implement Phase III to start production in late 1999 or early 2000. This expansion will propel

Hemlock clearly ahead of Wacker Siltronic as the largest producer of polysilicon.

Although a good portion of its capacity is allocated to the joint-venture companies. Hemlock enjoys a broad list of customers as a merchant supplier. Hemlock uses the trichlorosilane process to produce polysilicon rods and chunks. Its manufacturing techniques are linked to a chemical manufacturer, Dow-Coming, for a supply of the liquid trichlorosilane. Dow-

Coming's plant in Midland, Michigan, is responsible for the production of this liquid chemical from metal silicon.

Hemlock is also a supplier of high-purity silicon tetrachloride, serving a niche market in fiber optics and some medical applications. This is a natural byproduct of the manufacturing process.

Tokuyama Corporation

Tokuyama is unique in the polysilicon market in one respect—it is the only company that is not at least partly owned by a silicon wafer manufacturer, and it is ranked as the third-largest polysilicon producer, with an estimated

1,650 tons of production in 1995. An improvement and expansion of the company's plant will take this total to 2,000 tons in 1996, and Tokuyama recently announced a new plant that will expand its capacity to 3,000 tons starting at the beginning of 1997.

Tokuyama Soda entered the polysilicon market in 1984, and by 1985, its

1,200-ton plant was the largest of its kind in Japan. Tokuyama also uses the trichlorosilane process in the production of polysilicon. Silicon tetrachloride, a high-grade byproduct of the process, is used by Tokuyama Soda as the raw material for a fumed silica used in several applications in electronics, including encapsulation materials for semiconductor packaging.

Wacker Siltronic AG

Wacker Siltronic AG has the distinction of being the most experienced company in the polysilicon manufacturing business. Initially built in the late 1950s, Wacker Siltronic's Burghausen facility continues to use the trichlorosilane process to produce polysilicon. Wacker has a completely integrated production system, similar to Tokuyama's, starting with raw silicon metal.

Wacker Siltronic has narrowly lost the lead in the production of polysilicon to Hemlock, producing an estimated 2,400 metric tons in 1995 (versus

Hemlock's 2,800). Wacker Siltronic has continually expanded to meet its own and its customer's demands and is expected to double production by the year 2000.

Even though Wacker Siltronic consumes more than half its own production of polysilicon, it has a broad customer base in the silicon wafer market. This broad base of customers qualifies Wacker Siltronic as a merchant supplier.

Wacker is also a major manufacturer of high-grade chlorosilanes and ultrapure hydrogen chloride, natural by-products of the manufacturing process, and through the Wacker Chemie group also produces fumed silica, used in electronics packaging and other consumer goods.

SEMM-WW-DP-9604 ©1996 Dataquest July 22,1996

10 Semiconductor Equipment, Manufacturing, and Materials Worldwide

For More Information...

Clark Fuhs, Director/Principal Analyst (408) 468-8375

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness.

^ ^ It does not contain material provided to us in confidence by our dients. Reproduction or disdosure in whole or in

IjOTO^^l l^*Cf part to other parties shall be made upon the written and express consent of Dataquest.

^ ' ^ * ^ " ' ' ^ ^ % S r ^ ' ^ * ' ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and Materials

Woridwide

Market Analysis

A Fresh Look at 200mm Wafer Supply and Demand: Still Tight

Longer Term

Abstract: The lOOmm wafer market has been tight for many months. How are suppliers

ramping to meet the need? In a Focus Report published by Dataquest in November 1995, we described an undersupply shaping up for 1996. Since that time, suppliers have significantly increased and accelerated their 200mm plant construction plans. Also, semiconductor manufacturers have placed emphasis on the reduction of test wafer consumption. But the lower and still-decreasing factory utilization rates in the semiconductor industry (caused by too much fab equipment) have given fobs the opportunity to convert from 150mm wafers to process the more cost-effective 200mm wafers. So where does that leave the balance for wafer supply? In this article, Dataquest provides an update to the supply and demand situation for 200mm wafers.

By Clark J. Fuhs

Methodology for 200mm Wafer Supply and Demand

The analysis for this part of the market is relatively straightforward in concept. Demand is driven by tabs, their ramp-up schedules, whether the products produced require epitaxial wafers, and what the test wafer consumption ratio is during the ramp and life of the fab. Dataquest's analysis takes advantage of our comprehensive fab database to drive a bottom-up derivation of ttie 200mm wafer demand, using a fab-by-fab process. This demand is segmented by polished wafers, epitaxial wafers, and test/ monitor wafers. Test and monitor use was based on each semiconductor company's usage styles and products produced (a higher percentage is used for higher revenue per square inch product segments, such as microprocessors), as well as the age of the fab. (Near the start-up, a higher percentage of test wafers is used.) Availability of supply was not considered.

I>ataQuest

Program: Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Product Code: SEIVIIVI-WW-DP-g603

Publication Date: July 29,1996

Filing: Perspective

(For Cross-Teciinology, file in the Semiconductor Regional Markets and Manufacturing binder.)

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Supply is driven by silicon wafer manufacturing plant construction by the silicon wafer industry. New silicon wafer plants are required because the existing 150mm wafer plants carmot be upgraded cost-effectively—an entirely new toolset is required for manufacturing 200mm wafers.

Dataquest recently completed a survey of the planned capacity expansions for the industry, and we have assumed some ramp schedules associated with these new plants.

Comparing these two parts of the market can describe the situation cleanly when considered within the concept of silicon plant utilization.

The Demand for 200mm Wafers Remains High

Demand for 200mm wafers is expected to increase to over 4.6 times the 1995 consumption by the year 2001. At the end of 1995, there were 49 fabs in operation processing 200mm wafers. Including several conversioris from

150mm wafer fabs expected to occur in the next several years, such as all three Micron Technology fabs in Idaho, Dataquest expects 234 fabs to be in operation processing 200mm wafers by the year 2001. Not all these fabs have been announced to date, and we have taken a long-term approach when estimating this number. Figure 1 shows the capacity split by region in terms of 200mm wafer starts expected by the year 2001.

Figure 1

200mm Fab Capacity by Region in 2001

SEMM-WW-DP-9603

964I15S

Source: Dataquest (July 1996)

What about all the fab delays the industry is experiencing today? In our fab-by-fab analysis, we were extremely conservative when taking into consideration fab aimouncements in 1997 and 1998, particularly if the primary product was memory. Although many fab delays have been announced and been realized, we elected in this analysis to anticipate further announcements of delays. We analyzed this factor on a company-bycompany basis, giving the nod to industry technology and product leaders such as Toshiba, Samsung, IBM, Texas Instruments affiliates, and NEC while bringing under scrutiny many second-tier companies. Winbond, for example, with its connection to Toshiba, was viewed as a slow ramp-up fab during 1997 and 1998, while Hualon Microelectronics Corporation's Fab 2

©1996 Dataquest

July 29,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide project, originally planned for 1996, was assumed to be pushed out significantly into late 1998, although no formal commitment has been made. In this way, we have factored into our analysis the recent negative shorterterm forecast for capital spending just released by Dataquest.

The demand based on this analysis is summarized in Table 1 on a regional basis. We note that test wafer usage on a percentage basis is significantly lower than was the case in our November 1995 report. Although the absolute demand for test wafers is lower in our long-term outlook, the consumption of 200mm wafers for semiconductor products has been raised (by about 5 percent in the year 2000, as an example). This increased demand for product wafers is the direct result of the semiconductor industry's having the opportunity to convert fabs running 150mm wafers to the more costeffective 200mm wafers.

Supply of 200mm Wafers: Tlie Silicon Industry Has Responded

One year ago, silicon wafer suppliers were still getting used to being a profitable industry again after a very long dry spell. With much better free cash flow today, silicon wafer manufacturers have much longer visibility in their commitment to capacity expansion. A summary of the 200mm supplier survey is provided in Tables 2 and 3.

The committed supply of 200mm wafers is significantly higher than that outlined in Dataquest's November 1995 report (by about 30 percent for the year 2000, for example). Several of the major suppliers have recently announced major plants and expansion, including two by MEMC Electronic Materials. The newest joint venture in Malaysia, MEMC Kulim, was announced earlier this year, and the proposal of a second Taisil plant in

Taiwan has also been factored into this analysis. Also, both Sumitomo Sitix and Wacker Siltronic have raised their production commitment for later in this decade. The major difference has been the introduction of aggressive ramp-up schedules from the smaller but fast-growing companies, namely

UniSil, LG-Siltron, Showa Denko, and Toshiba Ceramics. The latter two are plarming to put in a significant epitaxial wafer capacity, as well, in anticipation of the requirements for epitaxial wafers by some Japanese companies for the 64Mb DRAM.

Supply and Demand Compared: Industry Plant Utilization Is Key

In Dataquest's November 1995 report there was such a mismatch between supply and demand that presenting raw numbers alone conveyed a potential shortage. Now that suppliers have responded with committed capital investments, we should actually take a look at factory utilization trends.

Silicon plant utilization is a metric that captures the supply/demand dynamics and the effect on prices and capital spending in the sector.

Table 4 shows the supply and demand comparison for both the total

200mm wafer market and the subset of epitaxial wafers. Also calculated is the implied factory utilization for the silicon industry. In 1996, Dataquest estimates that plant utilization is about 90.4 percent.

In today's environment, we are hearing comments like, "We are shipping everything we make." This indicates that 90 percent to 91 percent represents full capacity, and a calculated number significantly above 90 percent

SEMI\/l-WW-DP-9603 ©1996 Dataquest July 29,1996

Semiconductor Equipment, IVIanutacturing, and Materials Worldwide c?s

S o

s*3

0 *

U*^ r - i

s

• *

^

0 \ r H

CM i n

t

r - i

^

0 0

c^

Qi

r H r H

CO r H

O

o

t M vn

S^ so

CO rsi

0 0

CO

CO v O

Ttt

CM

Tji

0 0

"^v r H

CM i n r H

o

r H

I S . r H

5!

o

C N r H

CM

^

<

2

0 0

^ so

CO i n

^

<

2

CO

^ r H

^

NO

^ i n

^ t>

CO

CM

I N r H

NO ON

CO CM

O

( N

o

( N

o

t N i n

o

CO

• < * <

0 0

o

0 0 r - l r H

o d

t N t N

ON r H t N

CO t N

NO r n "

0 0

CO

CO

o

CO

1 * so

CO

0 0

SO cv!

0 0

m

r H

0 0 - ^

I N CNJ i n

{3 o

52

01

M H

(t

M ^

o

X

•o c rt w o

O

o o

Os

0 ^ o\

F H

OS

9 ^ ff^ r H t N

0 ^ a\ i H

\D i n

CO r j

CO

CO

CO r H v O

• < *

CM

CO f 7 r H

T — t od

9

0 0

CM

0 0

CM iS t N

•<j5 i n r H r H r H

0^

T - l

CO

0 0

0 0 r H

CM

0 0 v O

CM t>

SO

OS v£> r H

T - l cs

r H v O

O v

CO

CO

iri

r4 t N

v-^

d o

r H

o

O S so

CO r H i

CNl

SO r H

0 0 f:i ON

CO

• < t r H vd i n r H r H r<

CO

o

i n r H t N

C N r H

o

O N r-J r H

SO

CM

• *

OS

CO t s

1

O

C O

O N

CM

q

r H

CM

< N

r-t

CM r H

CO

NO

I N

o

t

CM

^^

0 0 r H

CO

CO

CM i n

0 0

o

0 0 r H

O

iri

CM r H

I N r H r H

CM

I N r H

CM i n

O

I N r H

q

r H

0 0

OS rsi

0 0 r H

ON

d

r H

o d

ON

ON

o d

NO

0 0

o

r p

0 0 r H

o d o

Nd

CM

CO

0 0

CO

CM

OS r H

CM r H

o

CO r H

SO

o

CM

CO

r H

CO i n

CO r H

0 0

ON

ON

O

d o

0 0 r H t N r H t N

ON

0 0

I N r H

0 0 t N

ON i n

CNl

CO

d

CM

NO r H

CO sd

O N

o

CM

• *

CO r H

I N

Os

CM t N in

CM

• *

O N

O i n r H

OS

CM

CM

•<1<

CM

CO r H

v-t

CM r H

CM

0 0 i n

r H

rtH r H i n Nd

O N

0 0

CO

NO c o s o ON

0 0 d

I N •>* r H

'dH'

(D

T 3

• «

C o u o c c

3

<o o

3

«*< t v

Nd iri tQ ^

CO

i

I

NO O i r i CTs

0 0 •<* t ^

CNT

(S

E

a>

"D

C>l

I

a o

bO

O)

>^ pa

ve

9 s as r - l i f i

9 s

OS

T H i n

m

"i

CM

r—*

CTs vd

o

i n i n

<3J vO

C S r H

0 0

CO

s

CM

o

CO

CO

T f

CM so

0 0 r H

OS

d

CO

O N

0 0 r H

NO

ON

2J

CO r H

SO sd r H

0 \ si

0 0 r H

'!»< s5

C O

C O

C 3 r H r H r H r H

^ t v

ON

T-*

0 0

ON

O

CM

m o

CO

CM

o

s

o

i n i n r H

o

CNl

q

r H

^-1

ON t N

o

CM

OS

o

i n

q

r H

I N r H

o

'*

NO r H so

CO

ON

d

0 0

CNl

Tji

O

CM

SO

O s tv ^

d cvi

r H i n

q

CM

•<}*

d

CO

0 0

CM r H

ON

CM r H

SO r H t N d 00 tx i n r H r H

a

Q .

0}

• o

1

c

CO c o c o

O d

a

D

<u I

•s <=

H n

*

9 s

9 s

<

i n f ^

( N

o

^

CM

• ^

r H t N

q

r H

o

t v

I N

r-l

r H r H

SO r H r H i n

CM r H

m

t <

o

CM

o

"li

CM

o

NO

I N

r^

i n t s r H

0 0

ON

s

ON

r-*

ON

0 0 r H

SO

^2

NO

I N

Os i n

CO

CM

CO

r-^

NO

a.

E

2 o

42

a.

1

•a

1

d.

EU

o

S

^^

4 F J

en

E2 o

E o x: .^ o <o

O a>

&-

U-i

O

•3

1

c '5

S

CH

( 0

'a,

la

J

-B

o

^ ,

E2

L O

0 )

g..l

a

"to

1

1

3

o o

N ^

*^

'G

^ 1

"N "B

"ffl

1

'a, w

o

"^

CA

01

E-H

05 cn

E-H

3

O

H

V

• d

3

•^ s

S

T3 -a

• £ ,

a

•E

o

en

^OJ

u

O

VH

^ "o a

3 g

1 o

"^

j - t

(A a « a o 3 vl

. . u

11 HI §

< "B o

H Z CO

SEMM-WW-DP-9603 ©1996Dataquest

July 29,1996

l\3

CO

CO

CO

CJ)

CO

• D

I

t o

OJ w

@

CO

CO

a> o

K.

(U

£2

c=

CD cn

Table 2

Estimated Total 200mm Wafer Production Capacity, 1994-2001

(Thousands of Wafers per Month Average, Including Those Used for Epitaxial Deposition)

1994

1995

1996 1997 1998

1999 2000

MEMC Electronic Materials

(and AH Joint Ventures)

United States

Japan

Europe

Posco-Huls, Korea

Taisil, Taiwan

MEMC Kulim, Malaysia

Total

Komatsu (and All Joint

Ventures)

United States

Japan

Taiwan

Total

LG Siltron

Korea

Total

Mitsubishi Materials

(Includes Siltec)

United States

Japan

Total

NSC Electron

Japan

Total

50

10

20

70

0

0

150

0

40

0

40

20

20

0

40

40

10

10

75

40

20

160

0

0

295

0

110

0

110

70

70

0

80

80

10

10

75

60

40

240

50

0

465

0

160

0

160

160

160

50

225

275

20

20

150

75

60

300

125

25

735

0

160

40

200

220

220

80

300

380

35

35

230

80

80

320

210

95

1,015

80

160

80

320

310

310

80

300

380

40

40

280

80

80

320

235

145

1,140

110

160

100

370

380

380

120

300

420

40

40

320

80

80

320

280

150

1,230

130

160

100

390

420

420

170

300

470

40

40

CO

CD

CD

tj>

@

CO

CD

a>

o

K.

cu

CD

Crt

-a

I

(O

a>

o

00

Table 2 (Continued)

EsBmated Total 200mm Wafer Production Capacity, 1994-2001

(Thousands of Wafers per Month Average, Including Those Used for Epitaxial Deposition)

1994 1995

1996

1997

1998 1999

2000

Shin-Etsu Handotai

United States

Japan

Malaysia/Taiwan

Total

80

120

100

300

120

180

200

500

120

180

220

520

160

220

250

630

220

290

280

790

240

320

290

850

240

320

290

850

Shawa Denko

Japan

Total

Sumitonio Sitix

0

0

0

0

0

0

25

25

45

45

50

50

80

80

United States

Japan

0

100

100

0

170

170

0

270

270

50

320

. 370

100

320

420

150

320

470

200

320

520

Total

Toshiba Ceramics

Japan

Total

30

30

60

60

125

125

240

240

350

350

450

450

550

550

UniSiJ

United States

Total

Wacker Siltronic

United States

Europe

Total

All Companies

Total

0

0

20

75

95

785

20

20

20

100

120

1,435

55

55

60

115

175

2,225

85

85

130

120

250

3,170

115

115

200

145

345

4,130

125

125

200

165

365

4,660

125

125

200

200

400

5,075

NA = Not applicable

Note: These annual figures represent average production estimates for each year. Because the industry is ramping up, the capacity of each company is end of each year.

Source: Dataquest (July 1996)

@

CO

to

o> o

Oi

CD c*!

Crt

a

-a

(b en o

CO

Table 3

Estimated 200mm Epitaxial Wafer Production Capacity, 1994-2001

(Thousands of Wafers per Month Average)

1994

1995

1996

1997

MEMC Electronic Materials

(and All Joint Ventures)

United States

Japan

Europe

Posco-Huls, Korea

Taisil, Taiwan

MEMC Kulim, Malays]^

Total

Komatsu (and All Joint

Ventures)

United States

Japan

Taiwan

Total

LG Siltron

Korea

Total

Mitsubishi Materials

(Includes Siltec)

United States

Japan

Total

Shin-Etsu Flandojfei

United States

Japan

Malaysia

Total

25

0

5

0

0

0

30

0

10

0

10

0

0

15

17

0

32

0

5

5

30

0

5

0

0

0

35

0

10

0

10

0

0

0

15

15

30

20

0

50

40

0

5

0

0

0

45

0

0

5

15

20

30

40

0

70

0

10

0

10

0

10

0

10

0

0

15

30

45

30

40

0

70

65

0

5

0

10

0

80

to

CO

CO

CO

O)

1998

85

0

5

0

20

0

110

40

30

70

35

60

0

95

6

10

0

16

10

10

2000

6

10

0

16

20

20

125

0

5

0

40

0

170

60

30

90

45

70

0

115

1999

6

10

0

16

10

10

100

0

5

0

30

0

135

60

30

90

45

60

0

105

ro

CO

CO

CO

oi

O

13

I

to at

o

CO

CO

m

@

tu

S3

c

CD t/l

<o

(O

Ol

D

Si

Table 3 (Continued)

Estimated 200mm Epitaxial Wafer Production Capacity, 1994-2001

(Thousands of Wafers per Month Average)

1994

1995 1996 1997 1998

1999 2000

Showa Denko

Japan

Total

0

0

0

0

0

0

15

15

20

20

30

30

45

45

Sumitomo Sitix

United States

Japan

4

19

23

23

20

43

62

22

84

74

28

102

74

28

102

80

22

102

80

22

102

Total

Toshiba Ceramicat

Japan

Total

0

0

0

0

10

10

35

35

70

70

90

90

130

130

WackerSiltronic

United States atiid Europe

Combined

30 30 50 60 75 90

100

Total

All Companies

30

30 50 60 75

90 100

Total 130

183 289

417

568 668 788

NA = Not applicable

Note: These annual figures represent average production estimates for each year. Because the industry is ramping up, the capacity of each company is end of each year.

Source: Dataquest (July 1996)

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Table 4

Worldwide 200mm Wafer Supply and Demand Stmimary, 1994-2001

(Thousands of Wafers per Month)

Total 200inin Wafer Demand

Total 200mm Wafer Supply

Implied Plant Utilization (%)

1994

613

785

78.1

1995

1,170

1,435

81.5

1996

2,011

2,225

90.4

1997

2,786

3,170

87.9

1998

3,456

4,130

83.7

1999

4,179

4,660

89.7

Total Epitaxial Demand

Total Epitaxial Supply

Implied Plant Utilization (%)

Source: Dataquest (July 1996)

64

130

48.9

125

183

68.2

204

289

70.7

304

417

72.9

423

568

74.5

573

668

85.8

2000

4,895

5,075

96.5

713

788

90.5

would suggest a shortage. In the second half of 1995, the industry probably crossed over the 85 percent utilization level, and about this time we heard of spot situations involving allocations. We would therefore regard the industry as having tight supply above 85 percent utilization. Prices began increasing from late 1994 to 1995 as the 80 percent utilization level was crossed. We would therefore regard industry pricing as firm or stable above

80 percent utilization.

How do these levels compare with the industry in the late 1980s, when prices were in decline? Given the fact that the key wafer size in demand was 150mm, that capacity was added in excess in 1984 and 1985, and that plant utilization for the industry was not likely much above 70 percent, it is evident that we are in a different environment today.

The analysis in Table 4 suggests that the DRAM slowdown (which has always been included in Dataquest silicon and capital spending forecasts in the last year for 1997 and 1998) will ease the supply/demand situation starting as early as later in 1996 before a renewed tight market in late 1998.

This is the case even though the supply capacity of 200mm wafers will have nearly tripled in three years from 1995! Equipment manufacturers supplying the wafer manufacturers will be very busy for the next couple of years.

Epitaxial plant utilization is likely to remain between 70 to 80 percent, even though our analysis indicates higher figures later in the decade. Epitaxial capacity is fairly easy to add through incremental equipment purchases. A significant portion of the demand starting in 1999 is anticipated to come from Japanese companies that may use epitaxial silicon layers for the 64Mb

DRAM. Because the epitaxial wafer is highly profitable, we would expect the industry to respond with capacity accordingly and quickly.

2001

5,458

5,215

104.7

836

848

98.6

Factoring in tlie "Soft Commitmenf'

In Dataquest's May 1996 survey of silicon companies, there were several companies that would not necessarily commit to increased capacity until these companies saw demand remain firm. We would classify this potential supply as "soft commitment." We have not included this capacity in the previous tables, but we have totaled the capacity impact of all those making soft commitments as a group. Table 5 shows the level and impact on factory utilization when this new capacity is added. The scenario is basically the same as the outlook without this capacity, except that actual shortages could perhaps be pushed out a year. Again, we expect the industry to

SEMIVI-WW-DP-9603 ©1996 Dataquest July 29,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

10

Table 5

Worldwide 200mm Wafer Supply and Demand Summary Including "Soft Commitment"

Supplies, 1994-2001 (Thousands of Wafers per Month)

Total Confirmed 200mm Supply

Total "Soft" 200mm Supply

Total Supply Base

Implied Plant Utilization (%)

Source: Dataquest (July 1996)

1994

785

-

785

78.1

1995

1,435

-

1,435

81.5

1996

2,225

-

2,225

90.4

1997

3,170

-

3,170

87.9

1998

4,130

45

4,175

82.8

1999

4,660

170

4,830

86.5

2000

5,075

445

5,520

88.7

2001

5,215

595

5,810

93.9 respond to the increased demand but would not expect to see factory utilization levels below 80 percent for a long time.

Dataquest Perspective

The supply of 200mm wafers is expected to be relatively tight for the remainder of this decade and into the next. The industry may receive a brief respite later in 1997 and into 1998, but we expect it to be short-lived, because the next capital spending cycle should be in place by late 1998.

Table 6 represents a concise summary of the annual characteristics of the market in terms of supply, pricing, and utilization.

Table 6

Summary of 200mm Market Supply/Demand Characteristics By

Year, 1996-2001

Year

1996

1997

1998

1999

2000

2001

Industry Plant

Utilization (%)

90+

88

83-84

87-90

89-90+

90+

Source: Dataquest (July 1996)

200mm Supply

Very tight

Tight but easing

Good but tightening

Tight

Very tight

Very tight

Price Trends

Increasing

Firm to slight easing

Stable to firm

Firm to increasing

Increasing

Increasing

Will semiconductor revenue be affected by the tightness in the 200mm wafer market? The answer is no. There is enough flexibility in this market in the area of test wafers that product runs will not be touched because of the lack of wafers. We feel coiifident that the silicon supplier will respond with increased capital spending, but we do expect plant utilization to be watched very closely.

Dataquest

A Gartner Group Company

For More Information...

Clark Fuhs, Director/Principal Analyst (408) 468-8375

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analjrsis of information generally available to the public or released by responsible individuals in the subject companies, but is i«3t guaranteed as to accuracy or completeness.

It does not contain material provided to us in cor\fidence by our clients. Reproduction or disclosure in wfhole or in part to other parties shall be made upon the written and express consent of Dataquest.

©1996 Dataquest—Reproduction Prohibited

Dataquest is a registered trademark of A.C. Nielsen Company

%

Dataquest

A Gartner Group Company 251 River Oaks Parkway, San Jose, CA 95134-1913

(408) 468-8000 Fax (408) 954-1780

August 5,1996

To our clients,

This report replaces Dataquest's "1995 North American Semiconductor Gas

Market" (SEMM-WW-DP-9602, May 6,1996), shipped previously. This version corrects errors in the market share information presented in Table 5.

Please place this report in your binder a n d discard the previous version.

Dataquest regrets the error and apologizes for any inconvenience. For further iiiformation, contact Senior Analyst Calvin Chang at (408) 468-8605 or [email protected] t com.

#

Perspective

#

S e m i c o n d u c t o r Equipment, Manufacturing, a n d Materials W o r l d w i d e

M a r k e t A n a l y s i s

The 1995 North American Semiconductor Gas IVJarlcet

Abstract: The North American semiconductor gas market chalked up a stellar year in 1995

•with estimated total sales of U.S.$421 million, a gain of 18.5 percent over 1994. Bulk gases and specialty gases both recorded the highest growth rate since Dataquest began tracking the semiconductor gas market in 1985. A monstrous increase of nearly 70 percent in North

American semiconductor capital spending produced an unprecedented double-digit growth in nearly all categories of electronic gases.

By Calvin Chang

The Bulk Gas Market

The North American bulk gas market is estimated to have reached

U.S.$252.1 million in 1995, a growth of 11.3 percent from 1994. This is the first double-digit growth rate and the fastest growth rate ever recorded in bulk gas sales in 10 years. Table 1 shows the semiconductor bulk gas sales in

North America from 1991 to 1995.

5

LU

N

Z

LU

Nitrogen: Liquid Loses Sliare to On-Site

For the semiconductor bulk gas market, 1995 was indeed a banner year in which all segments grew during the year. The last time this occurred w a s in

1989. Although liquid nitrogen sales grew nicely in 1995, liquid nitrogen's share of overall nitrogen sales continued to decline. From 1991 to 1995, the liquid nitrogen share has shrunk from 59 percent to 51 percent of total nitrogen revenue. On-site nitrogen continues its march, gaining big sales and market share. In the same period, on-site plants' contribution to total nitrogen sales increased from 27 percent to 36 percent. Dataquest believes that

LU CC

-J <

E S

OataQuest

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-DP-9602

Publication Date: August 5, 1996 (Replaces May 6, 1996)

Filing: Perspective

(For Cross-Technology, file in the Semiconductor Regional Markets and Manufacturing binder.)

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 1

North American Semiconductor B u l k G a s Sales (Millions of U.S. Dollars)

Nitrogen Total

On-Site

Pipeline

Liquid

Hydrogen

Oxygen

Argon

Other Bulk Gases

(Helium, Carbon

Dioxide)

Total

Year-to-Year Change (%)

'Preliminary

NA = Not available

NM = Not meaningful

Source: Dataquest (March 1996)

1991

162.3

44.2

22.8

95.3

29.0

7.4

15.1

NA

213.7

-

1992

164.0

51.5

21.4

91.1

26.9

8.7

16.1

NA

215.7

0.9

1993

160.9

54.9

19.3

86.7

30.3

8.2

19.8

NA

219.2

1.6

1994

163.7

55.8

29.6

84.4

31.4

8.5

22.1

0.9

226.6

3.4

1995*

181.7

65.4

29.9

92.1

35.5

11.0

22.9

1.0

CAGR (%)

1991-1995

2.9

10.3

7.1

-0.8

5.3

10.4

11.0

NM

252.1

11.3

4.2

-

this development/ which began several years ago, will continue. The continual migration tow^ard building larger fabs, the lowest cost in nitrogen generation, and a business environment of relatively low-cost capital have all contributed to the continued increase in the adoption of on-site nitrogen plants in the semiconductor industry.

Other bulk gases, including argon, oxygen, and hydrogen all enjoyed healthy growth in 1995. Robust demand plus modest increases in average selling prices (price increases began in 1994) have contributed to the best bulk gas market since 1985.

Semiconductor Capital Spending: Tlie Engine Propeiiing tlie Growtii

The 1995 bulk gas market data reaffirms the close correlation noted between the grov^th rate of semiconductor capital spending and that of the bulk gas market. The basis of this correlation has to do with new semiconductor fab construction and expansion, which will be populated with production equipment that uses bulk gases for purge, dilution, and processing.

It has been shown that the peak growth year for bulk gases lags the peak growth year in semiconductor capital spending by about one year. The oneyear lag is consistent with the general time elapsing between the construction of a new fab, during which capital is spent, and the onset of ramp-up production, which produces a surge in bulk gas consumption.

Figure 1 shows the North American semiconductor capital spendihg growth rates and the bulk gas market growth shifted back one year to adjust for the effect of the time lag. As the figure shows, the bulk g a s m ^ r k e t growth trends coincidentally (scale-adjusted) with the level of capifal (^ spending. When capital spending increases, bulk gas consumptioriY't ,

SEMM-WW-DP-9602

©1996 Dataquest

Augusts, 1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Figure 1

Semiconductor Capital S p e n d i n g as a Leading Indicator for the Bulk Gas Market

Bulk Gas Market

Annual Growth Rate (%)

Semiconductor Capital Spending

Annual Growth Rate (%)

1 1 1 1 1 1 1 1 1 1 1 —

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

1998

• — North America Bulk Gas Sales, Shifted Back 1 Year

North America Semiconductor Capital Spending sszeeo

Source: Dataquest (May 1996) increases—in the following year. When capital spending declines, so does the bulk gas market. This linic between semiconductor capital spending and the bulk gas market provides a convenient leading indicator for the health of the bulk gas market.

The figure shows that the semiconductor capital spending projection, as a leading indicator, suggests that the peak in capital spending in 1995 means that 1996 will be the last of a string of growth years in the bulk gas market.

The projected slowdown in North American semiconductor capital spending in 1996 through 1998 could lead to a corresponding slowdown in the bulk gas market in 1997 through 1999.

The Specialty Gas Market

The semiconductor specialty gas market in North America struck gold in

1995. The market grew at a n unprecedented 30 percent rate in 1995, with all major categories (silane precursors, dopants, plasma etchants, reactants, and atmosphere/purge) participating in double-digit growth rates. Table 2 shows the North American semiconductor specialty gas market in 1991 through 1995. Etchant gases continue to be the fastest-growing segment in specialty gases. Volume consumption of plasma etchant gases has been in large part driven by the increasing proportion of etch steps in IC fabrication processes. Silicon precursors saw an unusually high growth of 27 percent in

1995, as silane and dichlorosilane both registered sizable growth. Particularly notable is silane, which may be enjoying a resurgence from the

SEMM-WW-DP-9602

©1996 Dataquest Augusts, 1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Table 2

North America Semiconductor Specialty Gas Sales ( M i l l i o n s of U.S. Dollars)

Silicon Precursors

Silane

Dichlorosilane

Trichlorosilane

Silicon Tetrafluoride

Disilane

Dopants

Plasma Etchants

Reactant Gases

Atmosphere / Purge

Total

Year-to-Year Change (%)

'Preliminary

NM = Not meaningful

Source: Dataquest (March 1996)

1991

22.0

15.6

3.3

2.0

1.1

0

7.4

20.2

16.3

11.9

HA

22

1992

22.2

15.6

73.5

2.2

0.9

0.01

7.5

25.3

18.2

13.6

86.8

12

1993

25.5

17.3

3.9

3.0

1.0

0.01

12.7

37.6

19.5

14.7

109.7

26

1994

26.2

19.2

2.7

3.0

1.2

0.04.

14.9

50.8

22.6

15.3

129.8

18

1995*

33.3

23.5

4.8

3.5

1.4

0.1

20.2

70.1

27.5

17.8

168.9

30

CAGR(%)

1991-1995

10.9

10.8

9.7

14.8

5.9

NM

28.8

36.5

14.0

10.7

21.4

increasing use of plasma-enhanced chemical vapor deposition (PECVD) of silane oxide in the fabrication of intermetal dielectric. Silane oxide has shown favorable polishing characteristics in the chemical mechanical polishing of wafers. Dopants are another bright area of growth that produced a 36 percent increase in sales in 1995.

Etchant Gas Market Boom Continues

Etchant gases sales continued to charge ahead in 1995, producing a 38 percent growth for the year. This is the fourth consecutive year of hypergrowth in plasma etchant gases, and it took etchants from less than 30 percent of specialty gas sales in 1992 to more than 42 percent in 1995. At this rate, etchants are expected to contribute to more than half of all specialty gas sales in just a few more years. In so, leadership in the supply of etchants will be tiie king-maker for the market leadership in overall specialty gases.

Table 3 provides a breakout of etchant gases and shows the 1992 through

• 1994 compound armual growth rate (CAGR) in each etchant gas. Freon

14 (CF4), freon 116 {Q^d and freon 23 (CHF3) all grew at a CAGR of well over 35 percent during 1992 through 1994 and are expected to have shown even faster growth in 1995.

Dataquest believes that the perfluorocarbons (PFC) will continue to be a mainstay of semiconductor manufacturing processes in spite of the environmental concerns (for example, global warming) surrounding these plasma etchants. The industry has developed measures, in terms of abatement techniques such as PFC recovery, that will substantially reduce the emission of tiiese substances into the environment.

Consumption of freon 115, an ozone-depletion substance banned by the

Montreal Protocol, has essentially been eliminated. C^Fg is also experiencing drastically reduced demand.

SEMI\/l-WW-DP-9602 ©1996 Dataquest August 5,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 3

North America Semiconductor Plasma Etchant Gas Sales ( M i l l i o n s of U.S. Dollars)

CF4 (Halo 14)

CF4 (Halo I4/O2)

CHF3 (Freon 23)

C2F6 (Freon 116)

C2CIF5 (Freon 115)

Other Halocarbons (13,13bl, 21)

NF3

SFe

BCI3

C3F8

HF

CI2

CIF3

HBr

Others (SiF4 and Others)

North America Plasma Etchant

Market

Year-to-Year Change (%)

NM = Not meaningful

Source: Dataquest (March 1996)

1992

2.62

0.09

0.27

1.30

0.003

1.00

0.09

25.18

2.91

0.29

1.94

8.01

0.001

0.05

5.09

1.52

-

1993

4.11

0.41

3.12

11.20

0.004

0.18

7.38

1.99

4.53

0.14

0.33

1.95

0.004

2.02

0.25

37.61

49

1^*4

6.26

0.76

4.22

15.53

0

0.10

9.40

2.42

5.08

0.03

0.94

2.75

0.100

2.67

0.64

50.91

35

CAGR (%)

1992-1994

47

61

47

39

NM

44

36

26

39

-38

88

46

493

63

173

42

-

NF3, used in chamber clean, continues to be driven by increasing adoption of single-wafer chemical vapor deposition (CVD) and etch processes. Both

BCI3 and CI2 are used in metal etch, and their consumption will continue to be driven by the fabrication of ever-increasing levels of interconnect in ICs.

HBr (as well as C y is used in dry etch of polysilicon and silicon and will maintain steady growth. CIF3, a powerful etchant (and cleaning agent) relatively new to U.S. IC manufacturers, is expected to continue to rack up fast growth because of some of its unique properties and its currently modest sales.

SEMM-WW-DP-9602

Specialty Gases Growth Tied to Semiconductor Equipment Sales

We have also observed that there exists a connection between some of the specialty gases and the semiconductor production equipment that consumes them. This connection has been noted in the sales of etchant gases and the dry-etch equipment market. Figure 2 shows that the growth of the etchant gas market trends nicely with the dry-etch semiconductor equipment market. Unlike the case of capital spending and the bulk gas market, there is, however, no time lag in the relation between the etchant market and the etch equipment market. This suggests that a reliable dry-etch equipment market projection can be used as a basis for developing a useful market projection for etchant gas sales. The analysis can be extended to other specialty gas markets, relating those gases to the sales of the semiconductor equipment that consumes them. Figure 3 shows the relation in yearto-year sales growth between the dopant gases and the ion implant equipment markets.

©1996 Dataquest Augusts, 1996

Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide

Figure 2

Plasma Etchant Market Growth and the Dry-Etch Equipment Market

North America Etchant

Gas Sales Growth (%)

50

Dry-Etch Equipment

Sales (%)

80

- 7 0

- 6 0

^ 5 0

- 4 0

- 3 0

- 2 0

- 1 0

- 0

-10

1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

North America Etchant Gas Sales Dry-Etch Equipment Sales

962B61

Source: Dataquest (Mayl 996)

Figure 3

Dopant Gas Growth and the Ion-Implant Equipment Market

./-.

North America Dopant

Gas Sales Growth (%)

lon-lmplanter Equipment

Sales Growth (%)

Ann

10-

6 0 -

5 0 -

4 0 -

3 0 -

2 0 -

/ \

/ \

/ \

\

/N}

/V^"^

\

/ / \ •

/ / \ x

\ . ^ / / ^

\

\

\

\

* t ^ \ f

- 1 0 0

- 8 0

- 6 0

- 4 0

- 2 0

- 0

- - 2 0

0 f- 1 ,

1987 1988 1989

1 1 1 i 1 1

1990 1991 1992 1993 1994 1995

1 1

1996 1997 1998

^ ^ North America Dopant Gas Sales lon-lmplanter Equipment Sales

»szse2

Source: Dataquest (Mayl 996)

©1996 Dataquest

SEMM-WW-DP-9602 Augusts, 1996

Semiconductor Equipment, Manufacturing, and iVIaterials Worldwide

Reactant Gases Sales Are Led by WFg and N2O

The two pillars of the reactant gas market are WFg and N2O. WFg is used in the deposition of tungsten and tungsten-silicide films. Tungsten deposition is an integral part of the fabrication process for vertical interconnects (tungsten plugs). The drive toward adoption of an ever-increasing level of interconnects in ICs will continue to boost the consumption of WF^ in semiconductor manufacturing. Table 4 shows the sales of reactant gases for

1993 and 1994. As shown in the table, WFg has the largest niarket share and is also the fastest-growing reactant gas. N2O, a critical source gas for fabrication of advanced gate oxide, is expected to see even greater demand in the foreseeable future.

Table 4

North America Semiconductor Reactant Gas Market ( M i l l i o n s of U.S. Dollars)

NH3

HCl

N2O

WFg

CO2

Others

Total

Source: Dataquest (March 1996)

1993

1.6

5.5

4.8

6.3

0.5

0.2

18.9

1994,

1.6

6.1

5.8

7.9

0.4

0.2

22.0

Growth (%)

1993-1994

2

9

21

26

-14

18

16

North American Semiconductor Gas Market Share

Table 5 shows the North American semiconductor gas suppliers and their market share in the total, bulk, and specialty gas markets from 1992 to 1995.

The four major gas companies (Air Liquide, Air Products and Chemicals,

BOC Gases, and Praxair) together accounted for 88 percent of total electronic gas market sales in 1995. In bulk gases, the four majors collectively contributed 96 percent of the sales in North America.

In the fast-growing specialty gas market. Air Liquide and Matheson have gained market share in year from 1993 to 1995. Focusing on specialty gases,

Matheson has displaced Praxair in the No. 3 position in specialty gases. For everyone in the specialty gas market, 1995 was, however, a bonanza year, with double-digit growth the norm and five suppliers romping home with more than 25 percent growth.

Dataquest Classification of Semiconductor Bulk and Specialty Gases

The following are classified as semiconductor bulk gases:

• Nitrogen (N; on-site, multicustomer pipeline, bulk liquid delivery)

• Oxygen (O2)

• Hydrogen (H2)

• Argon (Ar)

SEMM-WW-DP-9602

©1996 Dataquest Augusts, 1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 5

North American Semiconductor Gas Suppliers Market Share Summary

(Millions of U.S. Dollars)

Air Liquide

Air Products and Chemicals

BOC Gases

Matheson

Praxair

Scott Specialty

Solkatroruc Chemicals

Tri-Gas

Others

Total North American Semiconductor

Gas Market

1992

27.1

121.2

57.6

10.9

72.0

3.9

3.1

NA

6.7

302.5

1993

32.1

131.7

61.8

16.3

68.1

5.7

4.1

2.5

6.9

329.2

1994

32.2

148.2

66.6

20.3

70.2

5.3

5.4

1.3

7.0

356.4

North American Bulk Gas Market

Air Liquide

Air Products and Chemicals

BOC Gases

Praxair

Tri-Gas

Others

Total North American Bulk Gas Market

24.8

101.3

26.8

57.2

NA

5.6

215.7

28.6

101.6

30.0

50.7

2.5

5.8

219.2

25.8

110.8

30.7

52.0

1.3

5.9

226.6

North American Specialty Gas Market

Air Liquide

Air Products and Chemicals

BOC Gases

Matheson

Praxair

Scott Specialty

Solkatronic Chemicals

Others

Total North American Specialty

Gas Market

"First six months

NA = Not available

Source: Dataquest (March 1996)

2.3

19.9

30.8

10.9

14.8

3.9

3.1

1.1

86.8

3.5

30.1

31.8

16.3

17.4

5.7

4.1

1.1

110.0

6.3

37.4

35.8

20.3

18.2

5.3

5.4

1.1

129.8

12.8

60.2

14.6

28.1

1.5

3.0

120.2

4.5

22.6

21.7

13.4

10.4

3.2

3.1

0.5

79.3

1995*

17.3

82.8

36.3

13.4

38.5

3.2

3.1

1.5

3.5

199.6

SEMM-WW-DP-9602

©1996 Dataquest Augusts, 1996

Semiconductor Equipment, iVianufacturing, and IVIaterials Worldwide

The following are classified as semiconductor specialty gases (delivered in cylinders):

• Silicon precursor gases a SiH4 a SiHjCl^ a SiHClj

• SiCU

• Dopant gases a AsHj

• PH3

Q BF3, BjjFg, PFg, and others

• Etchant gases a CF4

Q CF4/O2 (halo I4/O2)

Q CHF3 (freon 23)

Q CjFs (freon 116)

• C2CIF5 (freon 115)

Q Other halocarbons (13,13bl, 21)

• NF3

• SF,

3 BCI3

• CsFs a HF

• CI2

• CIF3

Q HBr a SiF4

• Reactant gases

Q N H 3

• HCl

Np

a WF,

co^

SEMM-WW-DP-9602 ©1996 Dataquest Augusts, 1996

10 Semiconductor Equipment, Manufacturing, and Materials Worldwide

For More Information...

Calvin Chang, Senior Industry Analyst (408) 468-8605

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness.

Dataquest

It does not cot^tain material provided to us in confidence by our clients. Reproduction or disclosure in whole or in part to other parties shall be made upon the written and express consent of Dataquest.

©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

\

Perspective

Semiconductor Equipment, Manufacturing, and

Materials Worldwide

Dataquest Predicts

What's Happening at Micron Teclinology?

Abstract: Micron rates as a relatively "pure DRAM play" among publicly traded semicon-

ductor companies. Starting with the recent abrupt resignation of Micron's top executive, this document assesses the 1996 to 1997 outlook for this supplier.

By Ron Bohn and Clark Fuhs

Appleton Returns!

Somewhat incredibly, Steve Appleton, the chief executive officer of Micron

Technology, stunned the industry twice over the course of about a week.

First, he "resigned" during the third week of January. Just over a week later, he returned as Micron's top executive. Likely he will be there for the foreseeable future, but a lot of "unverified" information circulated about Micron during late January.

Micron's experienced nianagement team has weathered prior dramatic challenges and shifts in the DRAM market. Dataquest expects the same for the long term. However, Micron clearly faces a stiff near-term challenge. For example, the spot market for 4Mb fast page mode (FPM) DRAM has abruptly shifted from the September 1995 sellers' market to a buyers' market today. Micron's stock fell during this period—from nearly $95 per share to the low $30s. In addition to the new DRAM pricing environment. Micron must manage a host of challenging 1996-to-1997 market transitions. These include transitions from 4Mb to 16Mb DRAM, from 6-inch to 8-inch wafers, and from FPM DRAM to extended data out (EDO) DRAM.

Dataquest

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-DP-9601

Publication Date: February 5,1996

Filing: Perspective

Semiconductor Equipment, Manufacturing, and Materials Worldwjde

"Lagging the Market" Strategy

Historically, Micron has successfully "lagged" the DRAM market in terms of technology. For example, when the market migrated from the 1Mb density to the 4Mb density during the early 1990s, Micron stayed with the older

1Mb density longer than many other suppliers and allowed other suppliers to lead the way to the then-emerging 4Mb product technology. At that time, some of Micron's competitors expected the 4Mb DRAM market to peak by

1993. By contrast, iVticron correctiiy and profitably anticipated that the 4Mb density would reach its peak stage during the post-1993 period.

Once again. Micron appears to be "lagging the market" as some competitors accelerate their migration to the 16Mfb density. Dataquest believes 1995 represented the high-water mark for unit shipments of 4Mb DRAM.

Worldwide unit shipments totaled 1.3 billion or more units last year, with shipments forecast to drop to about 1.1 billion units for 1996. Dataquest expects that more than 800 million imits of 16Mb DRAM will ship this year

(versus 325 million imits last year).

Micron also seems to be lagging the market shift from FPM DRAM to EDO

DRAM. For example, during 1995 Micron touted its burst EDO (BEDO)

DRAM. From Dataquest's perspective, BEDO to date has not won much market acceptance. Dataquest expects a 1996 market preference for 60ns

EDO DRAM that could stretch into 1997. Micron Ukely is focusing much effort on EDO DRAM, given the market's preference. Later this year, the market will initiate a long-term migration to synchronous DRAM

(SDRAM). Suppliers like Samsung, NEC, and Texas Instruments will lead the move to SDRAM.

Manufacturing Challenges

Micron faces critical manufacturing issues. Corporate attention focuses on converting Fab in (Boise, Idaho) from a 6-inch wafer facility to an 8-inch wafer facility.

8-Inch Wafer Conversion

A recent U.S. Securities and Exchange Commission (SEC) filing by Micron highlights this issue. Micron claims it had originally anticipated that Fab HI would be substantially converted to 8-inch wafer processing by the end of calendar 1995. While considerable progress has been made. Micron now does not expect substantial conversion of Fab HI before the middle of fiscal

1996. (The middle of Micron's fiscal 1996 is in late February and early

March.)

We should note Micron is not the only semiconductor supplier challenged by the 8-inch wafer conversion. Dataquest estimates that just eight to 10 leading-edge semiconductor suppliers worldwide have made the migration. Most suppliers making the 8-inch transition seem to have a "hope for the best, prepare for the worst" strategy.

The use of 8-inch wafers meshes well with a 16Mb ramp-up. The use of

8-inch wafers would mean higher operating DRAM margins than 6-inch wafers. The 8-inch wafers generate nearly twice the nvunber of DRAM die than 6-inch wafers. This applies to either 4Mb or 16Mb DRAMs. The use of

SEMM-WW-DP-9601 ©1996 Dataquest February 5,1996

}

{

Semicondiictor Equipment, Manufacturing, and l\/lateriais Worldwide

8-inch wafers would enable Micron to protect, to some degree, its 4Mb

DRAM margins despite pricing declines. Micron's expertise in shrinking die means the company has enjoyed extremely high DRAM margins.

By early March 1996, the market should learn whether Micron has achieved substantial conversion of Fab IE to 8-inch wafers. Should this conversion occur. Micron could rebound quickly and strongly. Should this not occvir,

Micron will remain under pressure until the 8-inch conversion does occur.

Capital Spending: $4 Billion by the End of Fiscal Year 1997?

The SEC filing also says that Micron's current expansion and capitalimprovement projects were estimated to total $4 billion in remaining cash outlays. These projects include the conversion of Micron's several Boise f abs to process 8-inch wafers and the construction of a new 8-inch wafer manufacturing facility in Utah (Dataquest believes that at least half of the $4 billion budget would be for the new Utah fab). Micron expected the majority of the $4 billion cost to be incurred prior to the end of fiscal 1997. Micron's fiscal 1997 ends in early September 1997.

The SEC filing says that substantially all Micron's near-term cash flow from operations was expected to be dedicated to capacity-improvement programs. Historically, Micron responds to DRAM market softness; however, it does so by moving to the lower end of the price curve. Micron is also evaluating debt financing as a source of additional funds for timely completion of these projects. Because of changed DRAM market conditions. Micron is likely evaluating whether it still wants to spend $4 billion—or a lesser amount.

We expect the projects to continue. The 8-inch Utah fab's construction schedule likely will be delayed, and instead of producing wafers later this year, production might not start till 1997. Instead of being a major revenue contributor by early 1998, the Utah fab might not generate big revenue streams until late 1998.

What's Ahead for Micron

By mid-1996, the market will gain signals on the ability of Micron's management team to handle Micron's 1996 market and manufacturing challenges.

A key second quarter 1996 milestone will be this: whether Micron makes a successful 8-inch conversion or encounters more delay. Another near-term milestone to watch for: Micron's ability to ramp up EDO DRAM, especially

60ns devices. Sooner or later, there should be some revision of the capital spending plans. Dataquest believes the Utah fab likely will move forward but under a slower schedule.

Again, Dataquest expects Mr. Appleton to remain as chief executive for quite a while. The recent episode, however, means Micron's public relations team will be busy—maybe for years—squashing rumors about "impending" management changes.

SEMI\^-WW-DP-9601 ©1996 Dataquest Februarys, 1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

#

^

SEMM-WW-DP-9601 o t v

Ms. Maria Valenzuela

Dataquest Incorporated

1-1100

• /

—INTERNAL D I S T . —

For More Information...

Ron Bohn, Principal Analyst (408) 468-8542

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness.

It does not contain material provided to us in cor»fidence by our clients. Reproduction or disclosure in whole or in

I j O T O / ^ I I ^ C T P ^ ' ' ° other parties shall be made upon the vmritten and express consent of Dataquest.

B ^ C 1 L U V { U V-bVL ©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

Perspective

Semiconductor Equipment, Manufacturing, and

Materials Worldwide

Dataquest Predicts

What's Happening at Micron Teclinoiogy?

AbStr3Ct: Micron rates as a relatively "pure DRAM play" among publicly traded semicon-

ductor companies. Starting with the recent abrupt resignation of Micron's top executive, this document assesses the 1996 to 1997 outlook for this supplier.

By Ron Bohn and Clark Fuhs

Appleton Returns!

Somewhat incredibly, Steve Appleton, the chief executive officer of Micron

Technology, stunned the industry twice over the course of about a week.

First, he "resigned" during the third week of January. Just over a week later, he returned as Micron's top executive. Likely he will be there for the foreseeable future, but a lot of "unverified" information circulated about Micron during late January.

Micron's experienced management team has weathered prior dramatic challenges and shifts in the DRAM market. Dataquest expects the same for the long term. However, Micron clearly faces a stiff near-term challenge. For example, the spot market for 4Mb fast page mode (FPM) DRAM has abruptly shifted from the September 1995 sellers' market to a buyers' market today. Micron's stock fell during this period—from nearly $95 per share to the low $30s. In addition to the new DRAM pricing environment. Micron must manage a host of challenging 1996-to-1997 market transitions. These include transitions from 4Mb to 16Mb DRAM, from 6-inch to 8-inch wafers, and from FPM DRAM to extended data out (EDO) DRAM.

Dataquest

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-DP-9601

Publication Date: February 5, 1996

Filing: Perspective

Semiconductor Equipment, Manufacturing, and Materials Worldwide

"Lagging the Market" Strategy

Historically, Micron has successfully "lagged" the DRAM market in terms of technology. For example, when the market migrated from the 1Mb density to the 4Mb density during the early 1990s, Micron stayed with the older

1Mb density longer than many other suppliers and allowed other suppliers to lead the way to the then-emerging 4Mb product technology. At that time, some of Micron's competitors expected the 4Mb DRAM market to peak by

1993. By contrast. Micron correctly and profitably anticipated that the 4Mb density would reach its peak stage during the post-1993 period.

Once again. Micron appears to be "lagging the market" as some competitors accelerate their migration to the 16Mb density. Dataquest believes 1995 represented the high-water mark for unit shipments of 4Mb DRAM.

Worldwide unit shipments totaled 1.3 billion or more units last year, with shipments forecast to drop to about 1.1 billion imits for 1996. Dataquest expects that more than 800 million vmits of 16Mb DRAM will ship this year

(versus 325 million units last year).

Micron also seems to be lagging the market shift from FPM DRAM to EDO

DRAM. For example, during 1995 Micron touted its burst EDO (BEDO)

DRAM. From Dataquest's perspective, BEDO to date has not won much market acceptance. Dataquest expects a 1996 market preference for 60ns

EDO DRAM that could stretch into 1997. Micron likely is focusing much effort on EDO DRAM, given the market's preference. Later this year, the market will initiate a long-term migration to synchronous DRAM

(SDRAM). Suppliers like Samsung, NEC, and Texas Instruments will lead the move to SDRAM.

Manufacturing Challenges

Micron faces critical manufacturing issues. Corporate attention focuses on converting Fab III (Boise, Idaho) from a 6-inch wafer facility to an 8-inch wafer facility.

8-Inch Wafer Conversion

A recent U.S. Securities and Exchange Commission (SEC) filing by Micron highlights this issue. Micron claims it had originally anticipated that Fab HI would be substantially converted to 8-inch wafer processing by the end of calendar 1995. While considerable progress has been made. Micron now does not expect substantial conversion of Fab III before the middle of fiscal

1996. (The middle of Micron's fiscal 1996 is in late February and early

March.)

We should note Micron is not the only semiconductor supplier challenged by the 8-inch wafer conversion. Dataquest estimates that just eight to 10 leading-edge semiconductor suppliers worldwide have made the migration. Most suppliers making the 8-inch transition seem to have a "hope for the best, prepare for the worst" strategy.

The use of 8-inch wafers meshes well with a 16Mb ramp-up. The use of

8-inch wafers would mean higher operating DRAM margins than 6-inch wafers. The 8-inch wafers generate nearly twice the number of DRAM die than 6-inch wafers. This applies to either 4Mb or 16Mb DRAMs. The use of

SEMM-WW-DP-9601 ©1996 Dataquest Februarys, 1996

^

Semiconductor Equipment, Manufacturing, and Materials Worldwide

8-rnch wafers would enable Micron to protect, to some degree, its 4Mb

DRAM margins despite pricing declines. Micron's expertise in shrinking die means the company has enjoyed extremely high DRAM margins.

By early March 1996, the market should learn whether Micron has achieved substantial conversion of Fab in to 8-inch wafers. Should this conversion occur. Micron could rebound quickly and strongly. Should this not occur,

Micron will remain under pressure until the 8-inch conversion does occur.

Capital Spending: $4 Billion by the End of Fiscal Year 1997?

The SEC filing also says that Micron's current expansion and capitalimprovement projects were estimated to total $4 billion in remaining cash outlays. These projects include the conversion of Micron's several Boise f abs to process 8-inch wafers and the construction of a new 8-inch wafer manufacturing facility in Utah (Dataquest believes that at least half of the $4 billion budget would be for the new Utah fab). Micron expected the majority of the $4 billion cost to be incurred prior to the end of fiscal 1997. Micron's fiscal 1997 ends in early September 1997.

The SEC filing says that substantially all Micron's near-term cash flow from operations was expected to be dedicated to capacity-improvement programs. Historically, Micron responds to DRAM market softness; however, it does so by moving to the lower end of the price curve. Micron is also evaluating debt financing as a source of additional funds for timely completion of these projects. Because of changed DRAM market conditions. Micron is likely evaluating whether it still wants to spend $4 biUion—or a lesser amount.

We expect the projects to continue. The 8-inch Utah fab's construction schedule likely will be delayed, and instead of producing wafers later this year, production might not start till 1997. Instead of being a major revenue contributor by early 1998, the Utah fab might not generate big revenue streams until late 1998.

What's Ahead for IVIicron

By mid-1996, the market will gain signals on the ability of Micron's management team to handle Micron's 1996 market and manufacturing challenges.

A key second quarter 1996 milestone will be this: whether Micron makes a successful 8-inch conversion or encounters more delay. Another near-term milestone to watch for: Micron's ability to ramp up EDO DRAM, especially

60ns devices. Sooner or later, there should be some revision of the capital spending plans. Dataquest believes the Utah fab likely will move forward but under a slower schedule.

Again, Dataquest expects Mr. Appleton to remain as chief executive for quite a while. The recent episode, however, means Micron's public relations team will be busy—maybe for years—squashing rumors about "impending" management changes.

SEMM-WW-DP-9601 ©1996 Dataquest Februarys, 1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

r

For More Information...

Ron Bohn, Principal Analyst (408) 468-8542

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, but is not guaranteed as to accuracy or completeness.

^ ^ It does not contain material provided to us in confidence by our clients. Reproduction or disclosure in whole or in

I J O T O / ^ I I ^ C T P ^ ' *° other parties shall be made upon the virritten and express consent of Dataquest. l ^ C I L C l V { U X > ' k 3 1 . ©1996 Dataquest—Reproduction Prohibited

A Gctrtner G roup Company Dataquest is a registered trademark of A.C. Nielsen Company

DataQuest

n f

M

I e.

>

FILE COPY:

MARIA VALENZUELA

DataQuest

IVIidyear 1996 Forecast: Capital

Spending, Wafer Fab Equipment, and Silicon

Market Trends

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-MT-9602

Publication Date: August 26,1996

Filing: Market Trends

I

I iVIidyear 1996 Forecast: Capital

Spending, Wafer Fab Equipment, and Silicon

Market Trends

>

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-MT-9602

Publication Date: August 26,1996

Filing: Market Trends

Midyear 1996 Forecast: Capital Spending, Wafer Fab Equipment, and Silicon

>

I

Table of Contents

Page

1. Executive Summary 1

Midyear 1996: Looking Past 1997, Setting U p for Growth in 1998.. 1

What Happened? 1

How Are Semiconductor Companies Responding? 1

When and H o w Will the Recovery Occur? 2

Highlights on the Wafer Fab Equipment Market 2

Highlights of the Silicon Market 3

Dataquest Perspective 3

2. Semiconductor Capital Spending Forecast 5

Highlights 5

Capital Spending Tables 6

And the Spending Binge Comes to a n End 6

How Long Will This Downturn Last? How Will Capacity Be

Absorbed? 10

Are Spending Levels Forecast Too High Relative to

Semiconductors? 12

Before 1985: A n Immature Manufacturing Industry 13

1985 through 1992: Becoming Manufacturing-Smart 13

1993 through 2001 (?): Growth 14

Beyond 2001: A Maturing Manufacturing Infrastructure 14

The Americas Market Will Exhibit Strategic Stiength Long

Term 14

Japan: DRAM Capacity Additions Stop, Investment in

Technology Under Way 16

Europe Sustains Presence as a Growth Market 17

Asia/Pacific Investing Focusing on Foundry as the DRAM Falls. 17

Who's Investing Where? 18

Dataquest Perspective 20

3. Wafer Fab Equipment Forecast 21

Highlights 21

Armual Investment Themes for 1996 through 2001 26

When Will D e m a n d Expand to Meet Capacity? An Update to the Overinvestment or Underinvestment Model 26

When Will D e m a n d Expand to Meet Capacity? A Fundamental

Set of Analyses 29

Highlights of Key Equipment Segment Markets and Forecasts .... 31

Steppers a n d Track 31

Etch and Clean: Dry Etch and Chemical Mechanical

Polishing 31

Deposition: CVD, PVD, and Silicon Epitaxy 32

Thermal Nondeposition Processes: Diffusion and RTP 33

Ion Implantation 34

The Segments That Will Fare Best in the Next Two Years 35

Dataquest Perspective 36

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table of Contents (Continued)

Page

4. Silicon Wafer Forecast 37

Silicon Forecast Tables 37

The 200mm Wafer Ramps Up—Suppliers Have Responded with Capacity 37

Polysilicon to Remain Short from Now through 1997 and

Tight in 1998 48

What about 300mm Wafers? 48

Epitaxial Wafer Trends: Are There DRAMs in the Future? 49

Highlights of the Americas Region Silicon Wafer Market and

Forecast 50

Highlights of the Japanese Silicon Wafer Market a n d Forecast 50

Highlights of the European Silicon Wafer Market a n d Forecast.... 51

Highlights of the Asia/Pacific Silicon Wafer Market and

Forecast 51

Silicon Wafer Revenue Forecast 52

Dataquest Perspective 53

5. Semiconductor Consumption Forecast 55

Semiconductor Consumption 55

6. Semiconductor Production Forecast 57

Historical Semiconductor Production 57

Captive Semiconductor Production 57

Restatement of Regional Semiconductor Production from 1992:

Foundry 58

The Move toward Asia Continues; European Growth Rests

Temporarily 59

Semiconductor Production Trends: The Accelerating Shift to

Asia/Pacific 60

Dataquest Perspective 61

Appendix A—Economic Assumptions, Second Quarter 1996 63

Appendix B—Exchange Rates 65

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Midyear 1996 Forecast: Capital Spending, Wafer Fab Equipment, and Silicon

»

List of Figures

Figure Page

2-1 Capacity Trickle in the Semiconductor Industry Today 11

3-1 Net Cumulative Overinvestment and Underinvestment of

Semiconductor Wafer Fab Equipment 28

3-2 Net Cumulative Overinvestment and Underinvestment of

Semiconductor Wafer Fab Equipment 28

>

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

iv Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

List of Tables

Table Page

2-1 Semiconductor Capital Spending—Top 20 Spenders, Comparison of 1994 and 1995 Worldwide Capital Spending 7

2-2 Semiconductor Capital Spending—Top 20 Spenders, Comparison of 1995 and Projected 1996 Worldwide Capital Spending 8

2-3 Worldwide Capital Spending History b y Region, 1988-1995 9

2-4 Worldwide Capital Spending Forecast b y Region, 1995-2001 9

2-5 Regional Investment Patterns of Semiconductor Manufacturers in 1995 19

2-6 Regional Investment Patterns of Semiconductor Manufacturers in 1996 19

3-1 Worldwide Wafer Fab Equipment Market History, by Region,

1989-1995 22

3-2 Worldwide Wafer Fab Equipment Market Forecast by Region,

1995-2001 23

3-3 Wafer Fab Equipment Revenue History b y Equipment Segment,

1989-1995 23

3-4 Wafer Fab Equipment Revenue Forecast by Equipment Segment,

1995-2001 25

3-5 Annual Driving Forces and Investment Themes for Wafer Fab

Equipment, 1996-2001 27

4-1 Forecast of Captive a n d Merchant Silicon* and Merchant

Epitaxial Wafers by Region 38

4-2 Forecast Growth Rates of Captive and Merchant Silicon* and

Merchant Epitaxial Wafers by Region 39

4-3 Forecast of Captive a n d Merchant Silicon* Wafers by Region 40

4-4 Forecast Growth Rates of Captive and Merchant Silicon* by Region 41

4-5 Forecast of Merchant Prime and Test/Monitor Wafers by

Region 42

4-6 Worldwide Wafer Size Distribution Forecast, 1993-2000 43

4-7 The Americas Wafer Size Distribution Forecast, 1993-2000 44

4-8 Japan Wafer Size Distribution Forecast, 1993-2000 45

4-9 European Wafer Size Distribution Forecast, 1993-2000 46

4-10 Asia/Pacific Wafer Size Distribution Forecast, 1993-2000 47

4-11 Worldwide Merchant Silicon Wafer Revenue Forecast,

1993-2001 52

5-1 Worldwide Semiconductor Consumption History by Region—

Merchant Semiconductor Companies Only 55

5-2 Worldwide Semiconductor Consumption Forecast by Region—

Merchant Semiconductor Companies Only 56

6-1 Worldwide Semiconductor Production History by Region—

Merchant and Captive Semiconductor Company Sales 58

6-2 Net Regional Semiconductor Production by Foundry, Dedicated and IDM Foundries Included 59

6-3 Worldwide Semiconductor Production Forecast by Region—

Merchant and Captive Semiconductor Company Sales 61

B-1 Exchange Rates per U.S. Dollar 65

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

I

Cliapter 1

Executive Summary

IVIidyear 1996: Loolcing Past 1997, Setting Up for Growtli in 1998

Overcapacity in the DRAM market, created by the massive spending from

1994 through the first part of 1996 and the normal DRAM product migration, is finally taking its toll, resulting in a faster-than-expected contraction in the wafer fab equipment market. We expect this contraction to be sharp and relatively deep but slightly shorter than historical norms, around 18 to

24 months. The key reason for the belief that the slowdown will be shorter than normal is the continued robust forecast for PC unit shipment growth through this decade.

What Happened?

Although Dataquest has been forecasting over the last year that a DRAM oversupply-driven capital spending slowdown would occur in 1997, the swift nature of its arrival still has shock value. We should point out why we were looking for a slowdown in capital spending.

The industry migration from 4Mb to 16Mb DRAMs would cause overcapacity even in the face a high bit demand growth. Why? Die size relationships mean that the average 16Mb DRAM has two to three times more bits per square inch than the 4Mb generation. Because capacity in the industry is built in wafers, this event causes a step-function increase in bit capacity of the industry. This usually happens in a hurry and is triggered by the new generation hitting a critical yield level of about 65 percent, estimated to have been hit by the end of 1995.

The complicating factor, and the reason for the severity of the downturn today, was the artificial demand created in 1995 by the anticipation of

Windows 95, which, taken with the tight supply of memory last year, caused inventories to inflate to enormous levels. This corrected in the final quarter of last year, slightly in advance, but essentially on top, of the conversion timing, and n o w we have a double-barreled driver (demand and supply) for the DRAM price slide and capital spending downturn. These events together have increased the anticipated severity of the downturn.

How Are Semiconductor Companies Responding?

It is very normal in this type of a downturn to get a pocket of companies that will stay and continue to invest in the infrastructure. These companies today are IBM, Texas Instruments, and, on a moderate level, the Korean companies. Japanese companies, which were the companies that held on in the last cycle for a year or more, have already shut off spending, and these companies as a group will spend slightly less in dollar terms in 1996 than in 1995. Except for Taiwan Semiconductor Manufacturing Co.,

Chartered Semiconductor, and Macronix, Asian companies are cutting back dramatically, starting in the last couple of months. As a group, the

DRAM manufacturers are responding more rapidly to the situation, cutting spending faster than in previous cycles.

SEMM-WW-MT-9602 ©1996 Dataquest

Semiconductor Equipment, Manufacturing, andMaterials Worldwide

The industry must now rely on the continued growth in personal computer unit sales, with added growth in telecommunications and networking products to create a unit d e m a n d picture that will keep this slowdown short-lived from a historical perspective. The wafer fab capacity bubble has b u r s t in all regions and for most semiconductor products, most notably DRAMs, mixed-signal devices, and analog devices. Whereas the 1995 spending growth was almost entirely driven by DRAM and microcomponent capacity purchases, 1996 is a year of transition, and 1997 will be a year of investment in technology.

When and How Will the Recovery Occur?

Based on how capacity is migrating among device types, we believe the first areas of spending recovery will be in the advanced logic area, as early as mid-1997. Equipment companies positioned for these markets will have a more moderate slowdown and perhaps can grow through this time if they have advanced technology. The MCU, analog, mixed-signal, and telecom chip capacity will be next to recover, but probably not until the end of

1997 or early 1998. The DRAM segment, the root cause of the problem and the very last to recover, is not expected to resume robust spending until mid-to-late 1998. The next major and broad investment cycle will have m o m e n t u m by 1999.

Data quest believes that capital spending may be influenced in late 1997 through 1999 positively with the facility construction and purchase of equipment toward the world's first 300mm wafer fab. We have built this inirastructure investment into our model. However, our outlook for a significant 300mm equipment market will wait until well after 2000.

Highlights on the Wafer Fab Equipment Market

Wafer fab equipment spending is expected to grow 17 percent worldwide in 1996. The only reasons that 1996 remains a double-digit growth year are that there were strong backlogs coming out of 1995 and increased spending by some companies early in the year.

After three and a half strong expansion years from 1993 through 1996, equipment purchases in 1997 should decline markedly, followed by an essentially flat 1998. Investment in DRAM capacity will be curtailed as producers elect to convert their 4Mb DRAM capacity to 16Mb, which a d d s bit capacity through the instant increase in bits per square inch. Also, m a n y Japanese DRAM facilities now running 150mm wafers will convert to 200mm wafers, further delaying the need for new equipment. DRAMsensitive equipment technologies or capital-intensive segments will be affected more than logic-sensitive technologies. The next expansion should kick in by 1999, driven by 0.3-micron to 0.35-micron capacity expansion.

During the coming slowdown, there will be two kinds of purchasing behavior that equipment companies can take advantage of to buffer sales declines. The first behavior is tied to which types of capacity will be required early in the recovery cycle. If there is a heavier dependence on advanced logic or supply to the materials industry, these segments will fare better than DRAM or semiconductor capacity dependent segments.

SE!\/l!\/l-WW-MT-9602 ©1996 Dataquest August 26,1996

Executive Summary

I

Included in this category are the segments of nontube CVD, sputtering, metal etch, silicon epitaxy, maskmaking lithography, process control systems in the materials business, and rapid thermal processing (RTP).

The second purchasing behavior that will be prevalent over the next 12 to

18 months will be new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. These segments will be those related primarily to deep-ultraviolet (deep-UV) lithography and inspection and 0.25-micron multilevel metallization schemes.

Highlights of the Silicon Market

The silicon market, driven by a strong long-term picture for semiconductor unit d e m a n d in general, will grow faster over the next six years than the last six years. As the industry transforms into a 200mm baseline, the outlook for silicon wafer manufacturers becomes brighter. Silicon manufacturers have answered the call for 200mm capacity with significantly increased capital outlays. We believe that silicon manufacturers' ramp plans in 200mm have been strategically and smartly measured, because the overcapacity of 1985 is being remembered, and w e are not expecting that scenario to repeat.

Activity in 300mm wafer development has accelerated, particularly in

Japan. Although no semiconductor company has yet to commit to purchasing equipment for a formally announced fab, w e would expect at least two pilot fabs to attempt to come o n line by 1999.

By the year 2000, Dataquest expects that fully 16 percent of merchant epitaxial silicon will be used for DRAMs, u p from about 2 percent in 1995. We believe that for late-generation shrink versions of 16Mb and for 64Mb

DRAMs, companies that employ the trench design will need to use epitaxial silicon. Also, other companies may decide to use epitaxial layers to solve specific yield issues at 0.25 micron; however, it may only be a singlegeneration solution.

The silicon has become recognized again as being strategic in the semiconductor manufacturing infrastructure. Will this continue? We believe it will, as long as silicon suppliers continue to concentrate on value-add processes a n d techniques as the equipment manufacturers have done, as well as adequately and smartly plan capacity additions.

>

Dataquest Perspective

Our forecast for capital spending and wafer fab equipment sales during the next six years assumes the excessive growth in 1995, which carried over into the first half 1996, will be sharply corrected in 1997. Our outiook for the future includes moderated growth in equipment spending in 1998, accelerating into the start of another boom in 1999, likely lasting into early

2001.

Where the PC goes, so go semiconductors. This is true from the perspective of the business forecast as well as the production line. Europe and

Asia/Pacific, with very large capital spending upticks over the last several years—and expected to continue that trend—will continue to gain share in world production over the next several years.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

The shifts and currents in semiconductor production trends m e a n that equipment and material suppliers will absolutely need a global presence in every sense of the word to remain competitive in the market. Product supply can no longer depend on local trends, as all major semiconductor companies have m a d e it clear they are investing on a worldwide basis.

However, local service a n d support are required to maintain customer satisfaction.

Taiwan is clearly the n e w major production growth area. We w o u l d expect Malaysia a n d Thailand to be the next major growth countries in three to five years. Evidence of this includes recent joint-venture fab announcements by Texas Instruments and others. Silicon plants are now being strategically placed, such as Shin-Etsu Handotai's Malaysian plant and recently announced joint venture in Taiwan, Komatsu's joint venture with Formosa Plastics in Taiwan, and MEMC Electronic Materials' joint ventures in both Korea (Posco-Hiils), Taiwan (Taisil), and Malaysia

(MEMC-Kulim).

Further, the concept of contract manufacturing in semiconductors is clearly here to stay. Equipment and material suppliers could find themselves selling their technical products to an international team from several companies, including the manufacturer and the designer. However the emergence of the dedicated foundry company taking ownership of the process and manufacturing flow will tend to centralize this activity.

Dataquest has started a research service—Semiconductor Contract

Manufacturing Worldwide—to continue exploring the key trends in contract manufacturing and foundries, including technology trends and supply/demand balance through the decade.

Project Manager: Clark Fuhs

Contributing Analysts: Calvin Chang, Joe D'Elia, Mike Glennon, C.S. Kim, Ben

Lee, Jason Lin, Takashi Ogawa, 'Nader Pakdaman, Beth Sargent, James Seay,

George Shiffler, Yoshie Shima, Yoshihiro Shimada, Yasumoto Shimizu, J.H. Son, and Jerry Yeh

SEMM-WW-l\/IT-9602 ©1996 Dataquest August 26,1996

^ Chapter 2

' Semiconductor Capital Spending Forecast

This chapter presents data on worldwide semiconductor capital spending by region. Capital spending in a region includes spending by all semiconductor producers with plants in that region. Components of capital spending are property, plant, and equipment expenditure for front-end and back-end semiconductor operations.

Chapter Highlights

This chapter will discuss the following highlights:

• On the heels of booming growth of 74 percent in 1995, global semiconductor capital growth will slow to 18 percent in 1996 to $45.3 billion.

With excess capacity clearly present in the industry for the next year or two, capital spending is expected to contract 10 percent in 1997 and modestiy recover in 1998 before the next capacity cycle starts in late

1998 into 1999.

• Capital spending in the Americas region grew at an accelerated 69 percent in 1995, with most of the investment growth in 1995 coming from

U.S. companies connected with ASIC a n d logic products. Capital spending is decelerating in 1996 into 1997, but we expect that investment in advanced technology, coupled with the earlier capacity upturn from the advanced logic segment, will stabilize the region's spending later in 1997 and will lead the market's recovery in 1998 and 1999 as it did in 1993. We expect the Americas region to be the second-fastestgrowing market as foreign multinationals and foundry company invest in capacity in the United States.

• Japan's 49 percent increase in capital spending in 1995 is only 37 percent on a yen basis, as Japanese companies look to invest outside Japan to optimize buying power. Japanese spending has essentially stopped growing and will actually decline about 2 percent in 1996. Because of the early cutback in spending, 1997 remains a flat capital spending year as well, with only modest growth in 1998. Lagging investment patterns in Japan are expected to continue throughout the decade.

• Japanese companies, however, grew spending during 1995 about

58 percent worldwide, spending a total of $12.0 billion, second only to

Americas company spending of $13.8 billion and well ahead of the

Asia/Pacific companies' $9.1 billion level. In 1996, however, with the quick brakes the Japanese have put on spending, Asia/Pacific companies will surpass Japanese company spending.

• Although spending on capacity has essentially stopped in Japan, two other types of investments are likely to be important in Japan now through 1998. First, Japanese companies will invest in any new technology and equipment targeted at the 0.25-micron production arena.

Second, the Japanese companies will build shells in 1996 and 1997, initially at very low run rates, as a preparation to ramp w h e n the market turns u p .

SEMM-WW-MT-9602 ©1996 Dataquest

Semiconductor Equipment, iVIanufacturing, and iVlateriais Worldwide

We are looking for continued growth for Europe of 22 percent in 1996 as production continues to ramp from Siemens, SGS-Thomson, and eight new fabs, most notably by GEC-Plessey a n d TEMIC. However, in part because we do not believe Siemens can sustain its current spending, we are calling for a 12 percent decline in spending in Europe for 1997 a n d a flat-to-down 1998 as w e expect multinationals will ramp domestic memory fabs before Europe. Longer term, w e still see Europe as a significant growth region for spending through the decade.

The often erratic but sustained semiconductor capital spending growth in the Asia/Pacific region continued at the explosive rate of 112 percent in 1995. We expect a "moderated" growth of about 30 percent in 1996 as several new fab projects are built and equipped. However, the tide has turned in the DRAM area, and we are forecasting the Asia/Pacific region to be hit the hardest in 1997, with a nearly 15 percent decline in capital spending. Longer term, we expect Asia/Pacific to exhibit one of the most aggressive growth rates in capital spending of any region.

However, several Asia/Pacific companies continue with their planned projects in expansion of foundry capacity, and new entries continue to be aruiounced, with more expected. The reason for the continued interest in spending capital in this area comes from the fact that the core business is dependent on logic and PC unit d e m a n d rather than DRAM.

The foundry industry is now a strategic industry rather than simply a tactical one.

Capital Spending Tables

A final list of the top 20 semiconductor capital spending companies in

1995 is presented in Table 2-1, and the comparable list for 1996 with projected spending is shown in Table 2-2. Capital spending details by region are provided in two tables in this chapter. Table 2-3 shows historical serrticonductor capital spending by region for the years 1988 through 1995.

Table 2-4 shows the capital spending forecast b y region for the years 1995 through 2001. Yearly exchange rate variations can have a significant effect on the interpretation of the data for 1988 through 1996. For more information about the exchange rates used and their effects, refer to Appendix B.

And the Spending Binge Comes to an End ...

After a 24 percent growth in semiconductor capital spending in 1993, accelerated growth of 54 percent followed in 1994, a n d growth has n o w peaked at 74 percent worldwide during 1995. Based o n our most recent capital spending survey, 1996 will have markedly slower growth of

18 percent. Nearly all of this growth has already occurred in the first half of this year, and we expect the spending contraction to begin in the second half of 1996, spilling into 1997.

The industry is now relying on the continued growth in PC unit sales, with a d d e d growth in telecommunications a n d networking products, to create a unit demand picture that will keep this slowdown short-lived from a historical perspective. The wafer fab capacity bubble has burst in all regions and for most semiconductor products, most notably DRAMs> mixed-sigrial devices, a n d analog devices. The 1995 spending growth was almost entirely driven b y DRAM and microcomponent capacity

SEMM-WW-MT-9602 • ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast

8

10

18

19

16

15

27

20

13

1994

Rank

1

2

9

3

6

11

14

12

5

7

4

Table 2-1

Semiconductor Capital Spending—^Top 20 Spenders, Comparison of 1994 and 1995

Worldwide Capital Spending (Millions of U.S. Dollars)

1995

Rank

1

2

3

4

8

9

10

11

12

5

6

7

13

14

15

16

17

18

19

20

Company

Intel

Motorola

LG Seinicon

NEC

Samsung

Toshiba

Fujitsu

Hitachi

H3mndai

IBM Microelectronics

Mitsubishi

Texas Instruments

SGS-Thomson

Micron Technology

Philips

Siemens AG

Matsushita

Chartered Semiconductor

Sanyo

Advanced Micro Devices

1995

3,550.0

2,530.0

2,258.1

2,010.1

1,946.6

1,624.1

1,592.1

1,497.6

1,492.4

1,150.0

1,118.2

1,079.3

1,001.0

960.0

959.0

850.0

846.6

786.7

647.0

634.0

969.9

700.0

525.0

675.3

825.0

780.0

387.0

385.0

1994

2,419.0

1,640.0

800.0

1,117.3

1,000.0

933.1

1,072.6

410.0

513.2

176.7

356.1

625.0

Change

(%)

46.8

54.3

182.3

79.9

94.7

74.1

48.4

54.4

113.2

119.0

65.6

30.8

28.3

148.1

149.1

107.3

65.0

345.2

81.7

1.4

Total Top 20 Companies

Total Worldwide Capital Spending

Top 20 Companies Percentage of Total

Source: Dataquest (July 1996)

28,532.7

38,308.1

74.5

16,310.2

22,085.2

73.9

74.9

73.5 purchases, 1996 is a year in transition, and 1997 will be a year of investment in technology.

The first comparues to cut back were the U.S. companies, as they tend to be more driven by short-term cost issues. The most severe slowdown in capital spending that is occurring today is in japan, as the overcapacity in

DRAM has caused Japanese companies to quickly draw the purse strings in hope of avoiding a more serious price erosion. The spending plans have already started to be cutback within DRAM companies in Taiwan, and we would expect this spending cutback to extend into Korea in late 1996 into

1997.

But meanwhile, the big three Korean companies are increasing spending a more modest 27 percent to a combined $7.2 billion in 1996, whidi is still above the industry in terms of growth. This above-industry growth has meant that all three large Korean companies are in the top five for capital

SEMM-WW-MT-9602

©1996 Dataquest

August 26,1996

Semiconductor Equipment, Manufacturing, and IVlaterials Worldwide

12

4

14

10

16

6

8

7

1995

Rank

1

3

5

9

2

22

17

13

11

18

15

20

Table 2-2

Semiconductor Capital Spending—Top 20 Spenders, Comparison of 1995 and Projected M

1996 Worldwide Capital Spending (MiUions of U.S. Dollars) ~

13

14

15

16

6

7

10

11

12

8

9

1996

Rank

1

2

3

4

5

17

18

19

20

Company

Intel

LG Semicon

Samsung

Hyxindai

Motorola

Texas Instruments

NEC

Micron Technology

IBM Microelectronics

Siemens AG

Toshiba

Hitachi

Fujitsu

TSMC

Matsushita

SGS-Thomson

Mitsubishi

Chartered Semiconductor

Philips

Advanced Micro Devices

1996

3,900.0

2,688.8

2,304.6

2,250.0

2,100.0

2,000.0

1,814.0

1,600.0

1,550.0

1,450.0

1,441.9

1,300.4

1,279.1

1,039.3

1,034.9

1,000.0

965.1

872.3

841.0

750.0

1995

3,550.0

2,258.1

1,946.6

1,492.4

2,530.0

1,079.3

2,010.1

960.0

1,150.0

850.0

1,624.1

1,497.6

1,592.1

583.9

846.6

1,001.0

1,118.2

786.7

959.0

634.0

34.8

70.6

-11.2

-13.2

-19.7

78.0

22.2

-0.1

-13.7

10.9

-12.3

18.3

Change

(%)

9.9

19.1

18.4

50.8

-17.0

85.3

-9.8

66.7

Total Top 20 Companies

Total Worldwide Capital Spending

Top 20 Companies Percentage of Total

Source: Dataquest (July 1996)

32,181.4

45,344.0

71.0

28,469.6

38,308.1

74.3

13.0

18.4 spending in 1996, as was shown in Table 2-2. As noted earlier, Japanese suppliers of memory cut back investment early in this cycle. Japanese companies as a group will actually spend 7 percent less in 1996 in dollar terms (6.5 percent growth in yen terms). As a result, only one Japanese company appears in the top 10 capital spenders in 1996—NEC. Most of the other Japanese companies do appear in the second 10. Intel still heads the list for 1996, as the microprocessor giant demand continues to be strong on a unit basis. Intel's capital spending growth has slowed primarily because yield ramps on its new fabs have been better than expected, so Intel therefore needs less equipment to produce the same unit volume. Motorola, the long-time No. 2 spender, has dropped to No. 5 as the demand for telecom-related chips softened in conjunction with the overcapacity in this area.

SEI\/ll\/l-WW-l\/IT-9602

©1996 Dataquest August 26,1996

Semicondiictor Capital Spending Forecast

Table 2-3

Worldwide Capital Spending History by Region, 1988-1995

(Includes Merchant and Captive Semiconductor Companies) (Millions of U.S. Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Japan (Billions of Yen)

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (July 1996)

1988

3,319

26.6

4,454

81.2

579

62.6

951

7.5

1,024

89.6

9,748

49.8

1989

3,833

15.5

5,415

21.6

748

29.2

1,198

26.0

1,884

84.1

12,331

26.5

10.4

1,598

33.4

1,580

-16.2

13,230

7.3

1990

4,320

12.7

5,732

5.9

826

-4.7

1,248

-21.9

2,300

45.6

13,145

-0.6

1991

3,895

-9.8

5,702

-0.5

787

1992

4,135

6.2

3,958

-30.6

500

-36.4

1,188

-4.8

2,318

0.8

11,599

-11.8

1993

4,943

19.5

4,413

11.5

491

-2.0

1,738

46.3

3,238

39.7

14,333

23.6

1994

7,194

45.5

6,667

51.1

679

38.3

2,504

44.0

5,720

76.6

22,085

54.1

1995

12,147

68.8

9,910

48.6

931

37.1

4,137

65.2

12,115

111.8

38,308

73.5

CAGR (%)

1989-1995

18.8

9.6

2.0

17.2

40.4

23.7

Table 2-4

Worldwide Capital Spending Forecast by Region, 1995-2001

(Includes Merchant and Captive Semiconductor Companies) (Millions of U.S. Dollars)

CAGR (%)

1995-2001

46.1

Americas

Percentage Growth

Japan

Percentage Growth

Japan (Billions of Yen)

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (July 1996)

1995

12,147

68.8

9,910

48.6

931

37.1

4,137

65.2

12,115

111.8

38,308

73.5

1996

14,828

22.1

9,692

-2.2

1,042

12.0

5,058

22.3

15,766

30.1

45,344

18.4

1997

13,461

-9.2

9,566

-1.3

1,028

-1.3

4,444

-12.1

13,470

-14.6

40,942

-9.7

1998

14,599

8.5

10,331

8.0

1,111

8.0

4,317

-2.9

13,668

1.5

42,915

4.8

1999

18,453

26.4

12,730

23.2

1,369

23.2

5,637

30.6

17,134

25.4

53,954

25.7

2000

24,571

33.2

16,829

32.2

1,809

32.2

7,725

37.0

26,653

55.6

75,777

40.4

2001

29,721

21.0

20,495

21.8

2,203

21.8

9,718

25.8

36,781

38.0

96,715

27.6

12.9

15.4

15.3

20.3

16.7

A mostly new crowd of Taiwanese companies that entered the DRAM manufacturing business, spending over $1 billion collectively in 1995, increased spending feverishly in tiie first half of 1996 and has likely spent more than in all of 1995 already. However, the spending planned for the second half of the year has been predominantly delayed, mostly into 1997, likely later, and, in some cases, indefinitely (read late 1998 or 1999).

SEMM-WW-MT-9602

©1996 Dataquest August 26,1996

10 Semiconductor Equipment, Manufacturing, and Materials Worldwide

TSMC debuts on the top 20 list for 1996 with a n estimated $1.04 billion spent on capacity, as foundry capacity expansion has now evolved into a major trend. TSMC and Macronix, a company that we understand does some second-source work for TSMC, are the two major Taiwanese companies still maintaining the spending plans set early in the year for 1996. This industry has transformed into a dedicated bona fide business and is no longer a specialized way to use excess capacity. There are several companies that we understand will try to enter the foundry business as a result of today's overcapacity. Unless these companies commit to the foundry business in the long term, their success at entering the market will be limited. Gone are the days when the "temporary" foundry can exist.

Customers of foundry are now requiring long-term relationships and contracts for winning their capacity business.

It is very normal in this type of a downturn to get a pocket of companies that will stay and continue to invest in the infrastructure or that have niches that maintain growth, thus supporting an increase in spending.

These companies in 1996 include IBM (advanced 16Mb DRAM to support systems), Texas Instruments (primarily digital signal processors, or DSPs, and logic), TSMC (foundry), Siemens (advanced DRAM), Lucent Technologies (modem chipsets and the Cirrus Logic foundry), and Chartered

Semiconductor (foundry). All of these companies have continued to increase investment during 1996. Micron Technology may be the surprise of 1996 to some, particularly because it has delayed the Lehi, Utah fab.

However, it has been spending aggressively in Boise, Idaho, upgrading the facilities for 200mm production and advanced technology for the

16Mb generation. We would expect Micron to return to 1995 levels at least in 1997.

With the cutback of the big Japanese players in the industry and with some smaller companies continuing to be aggressive in spending plans, the concentration of capital spending by the top 20 has decreased in 1996 b y a few percentage points to 71 percent.

How Long Will Tbis Downturn Last? How Will Capacity Be Absorbed?

Our longer-term forecast projects that this contraction will be sharp and relatively deep, b u t slightly shorter than historical norms, about 18 to

24 months. The key reason for the belief that the slowdown will be shorter than normal is the continued robust forecast for PC unit shipment growth worldwide through this decade, driving the need for silicon. We also do not see a stoppage in advanced technology investment, indicating a belief in the customer base of a strong end-user market for semiconductors.

Overall semiconductor product demand is expected to remain strong longer term, with sustained growth through 2001 with a compound annual growth rate (CAGR) of 15.5 percent (see Chapter 5).

H o w will the industry recover? What are the anticipated dynamics?

Today, the industry is primarily experiencing a DRAM oversupply coupled with a product transition. In order to determine how capital spending may recover, it is important to u n d e r s t a n d how this excess capacity may migrate or trickle to other areas in semiconductor capacity.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast

11

^

There are two general blocks of capacity now available from today's conditions (see Figure 2-1). These two blocks are being redirected into other semiconductor product areas today. The first comprises old 4Mb DRAM fabs that cannot run 16Mb chips. These are limited to two-level metal and are 0.6-micron to 0.8-micron technology. Microcontrollers, telecommunications chips, mixed-signal ICs, a n d analog ICs are quite happy being processed in these fabs. It is likely that most of these fabs in Japan and some in

Korea will migrate into this area. The power a n d discrete chips have specialized processes not found in old DRAM fabs, so these segments are relatively isolated from extraneous supply impacts because significant time a n d money is required to convert. We therefore expect capital spending patterns to be closely tied to demand in this specialty segment.

The second block of capacity comprises idle or underused advanced 16Mb capacity, which is limited today to two-level metal but at 0.4 micron to

0.5 micron. Because these fabs generally lack the process sequences of selfaligned silicide and three-level metal or more (w^hich requires chemical mechanical polishing at 0.5 micron), they cannot be effectively redirected to advanced logic or fast SRAM. Therefore, they are limited to commodity

SRAM, flash, other nonvolatile memory, or a limited span of logic products. W e expect most of these plants to remain somewhat idle through the overcapacity period.

I

Based o n how this capacity is migrating, we believe that the first areas of spending recovery will be in the advanced logic area, as early as mid-1997.

Equipment companies positioned for these markets will have a more moderate slowdown and perhaps can grow through this time if they have advanced technology. The MCU, analog, mixed-signal, and telecom chip capacity will be next to recover, b u t probably not until the end of 1997 or

Figure 2-1

Capacity Trickle in t h e Semiconductor Industry T o d a y

^

Old

4 M D

Fabs: 0.6-0.8

Micron

DRAM Overcapacity

1

J

Recovery Last

Idle ibiviD

Fabs: 0.4-0.5

Micron

1

r

Connmodity SRAM, 1

Flash Memory 1

MCU, Analog,

Mixed Signal

Recovery Second

1

X-^

Power, Disc

Recovery First

>

k

Fast SRAMs, 1

Advanced Logic 1

96571S

Source: Dataquest (August 1996)

August 26,1996

SEMM-WW-MT-9602 ©1996 Dataquest

12 Semiconductor Equipment, Manufacturing, and Materials Worldwide early 1998—Motorola will be a key company to watch. The DRAM segment, the root cause of the problem and the very last to recover, is not expected to resume robust spending until mid-to-late 1998. The next major and broad investment cycle will have momentum by 1999.

Our model does not include significantly more 16Mb DRAM capacity expansion until late 1998 into 1999. In the two "pause" years, w e believe

DRAM manufacturers will concentrate on converting capacity away from

4Mb toward 16Mb, which increases bits per square inch processed, then concentrate on shrinks to squeeze out value per square inch before a capital cycle starts again. Further, in Japan, w e expect that many 4Mb/16Mb fabs now running 150mm wafers will convert to 200mm wafers, further gaining efficiency and productivity from the capital investment. Although this will increase the demand for silicon and 200mm wafers, it will likely delay the capital spending cycle by six to nine months.

Through 2001, w e project a six-year worldwide capital spending CAGR of

16.7 percent, slightly ahead of the semiconductor consumption growth.

We believe that capital spending may be influenced in late 1997 through

1999 positively with the facility construction and equipment purchases for the world's first 300mm wafer fab. We have built this infrastructure investment in our model.

A couple of years ago, Dataquest introduced a model that quantifies the overinvestment and underinvestment picture for wafer fab equipment and semiconductor capacity. Although the early 1990s created and sustained a net underinvestment, this picture was corrected to create about a 37 percent overinvestment by the end of 1995 (see Chapter 3 and

Figures 3-1 and 3-2). Clearly this was in the danger zone, and w e are seeing the results of this overinvestment today. By the end of 1998, should our forecasts for investment and semiconductor demand be on target, we would expect the industry to return to a 16 percent net underinvestment, setting the stage for a robust recovery starting in 1999.

Are Spending Levels Forecast Too High Relative to Semiconductors?

As capital spending as a percentage of semiconductor revenue exceeds

•27 percent in 1996, clearly high by historical standards and creating overcapacity, a question is being asked often. What spending level is dangerous? W h a t spending level is normal?

The industry normally cycles through overcapacity and undercapacity, because there is an inherent lag time between capital investment decisions and productive capacity. This will never go away. But we are experiencing levels of spending today that we have not seen since 1984; should we return to the spending levels of the late 1980s, the current d o w n t u r n will clearly last longer than we are forecasting.

We think the industry is structured differently today, and there is a valid reason w h y higher levels of investment as a percentage of revenue than in the late 1980s are justified today.

We would split the market into four periods: before 1985,1985 to 1992,

1993 to 2001 (?), and 2002 and after. We will describe the conditions and

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast 13 trends in the production market and manufacturing infrastructure during these periods.

Before 1985: An Immature Manufacturing Industry

Characteristics of the period are:

• The semiconductor manufacturing infrastructure was fairly immature, characterized by large integrated systems companies, mostly in the

Uruted States.

• Manufacturing technology was favored over efficient use of capital, and device performance was favored over yield.

• Capital equipment manufacturers did not assume complete ownership of processes and system performance.

• These factors led to a capital spending ratio between 26 percent and

30 percent of revenue.

Two things happened to change the structure. First, the semiconductor downturn in 1985 ended up being extremely bloody, and, second, the emergence of the Japanese producer introduced a true manufacturing efficiency element into the irifrastructure.

1985 through 1992: Becoming Manufacturing-Smart

The Japanese manufacturing ethic and the losses incurred during the 1985 downturn introduced the need for the industry to become more efficient in its manufacturing infrastructure. Several things happened:

• A focus developed on manufacturing productivity and yield.

• SEMATECH was formed, in part as a result of the need to coordinate this effort in the Uruted States.

• Equipment companies were expected to take, and accepted, the ownership of the process and system performance parameters.

• Equipment performance and productivity increased substantially.

• Fab factory automation migrated from robotics to computer-controlled systems, and statistical process control became commonplace.

• The emergence of the Korean manufacturing power provided momentum for this transition.

By 1992, manufacturers had increased yields, and fab productivity was up dramatically. During this period, it was natural for the industry not to be required to spend as much on capital equipment, because the return per dollar spent was very high. Capital spending decreased to an artificially low 18 percent of revenue, on average.

Once yields achieved high levels and system productivity approached the point at which it was impractical to continue for most equipment types (to do otherwise would mean decreased equipment utilization), the return from these activities were diminished. The industry then entered the next period.

SEI\/iM-WW-MT-9602 ©1996 Dataquest August 26,1996

14 Semiconductor Equipment, Manufacturing, and Materials Worldwide

1993 through 2001 (?): Growth

As the industry built an efficient manufacturing infrastructure, it w a s now ready for the emergence of the semiconductor as the enabling technology in many electronic systems, and the industry entered an era of prosperity.

Semiconductors have become the productivity engine for the world's business, implementing communications systems and the power of the PC to improve worker efficiency. Characteristics we are now experiencing are:

• Unit growth in semiconductors has required manufacturers to invest in capacity for growth.

• Profitability has attracted new entrants with a concentration on manufacturing.

• Dedicated contract manufacturing has emerged as a new manufacturing model for the industry, enabled because of equipment efficiencies and the need to separate manufacturing from device innovation.

When the industry reached the point at which returns from yield and productivity from equipment had diminishing returns, the capital spending ratio could no longer be maintained at artificially low levels. We believe the current equilibrium level is about 22 percent of semiconductor revenue.

Further, because of the growth in the device market, this boom has been unprecedented in the industry in terms of length and levels of growth in capital spending. This has been the result of the industry adjusting to a new, higher spending level and of the increased unit demand.

Beyond 2001: A Maturing Manufacturing Infrastructure

With the emergence of dedicated contract manufacturing, Dataquest believes that the industry's manufacturing irifrastructure will evolve so that the foundry becomes an integral part of the manufacturing environment, as been the case with electronic equipment in general (Solectron is an example).

The foundry business model requires high equipment utilization, and we expect that sometime in the next five years this will influence the capital efficiency and decrease the capital spending ratio to perhaps 20 percent of revenue. This will not become evident until contract manufacturing increases in scope, from the estimated 9 percent of semiconductor production in 1995 to levels that could approach 35 percent to 45 percent by the year 2010.

The Americas Market Will Exhibit Strategic Strength Long Term

Capital spending in the Americas region grew at an accelerated 69 percent in 1995, with most of the investment growth in 1995 coming from U.S. companies connected with ASIC and logic products. Capital spending is decelerating in 1996 and into 1997, b u t Dataquest expects that investment in advanced technology, coupled with the earlier capacity u p t u r n from the advanced logic segment, will stabilize the region's spending later in 1997 and will lead the market's recovery in 1998 to 1999 as it did in 1993.

Dataquest is currently forecasting that the Americas region will be the second-fastest-growing region, at a 16.1 percent CAGR for 1995 through 2001,

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast Iff: driven by the recent low cost of capital and the need for foreign multinational and foundry manufacturers to build fabs in the United States to be closer to their customers.

The relatively strong growth in capital spending had been driven by the ongoing growth in PCs, telecommunications, and networking. This key driver has not disappeared, as these products have seen increasing use as tools to increase workplace productivity. Electronic products with increased semiconductor content have created enormous demand for microprocessors, microcontrollers, SRAM, programmable logic and memory, standard logic, and peripheral controllers. The U.S. companies dominate many of these market segments. These segments combined are expected to maintain fairly stable growth rates over the next few years, with PC growth slowing (however, still maintaining a 17 percent CAGR) and networking and telecommunications expanding. The near-term market for PCs has remained robust worldwide, despite the slower growth and penetration into the U.S. home market.

N e w products and services, such as personal communicators, interactive television, and video on demand provide the potential for enormous growth in semiconductor sales longer term, especially for highly integrated complex logic and signal-processing chips that will be the core engines of future systems.

Although the strategic strength of the core logic products enables a healthy and flourishing semiconductor production envirorunent, it is also one that is less volatile in capital spending. In the boom years of 1994 and

1995, the Americas region grew at somewhat lower than the market rates.

This trait will also enable the Americas market to grow in capital spending at faster than market rates (or remain more stable) in the slower years, such as 1997 and 1998. We believe companies will strategically invest in technologically advanced capacity to preserve competitive advantage.

Capital investment trends in the Americas region for 1996 have a definite split personality. While Intel is finishing u p expansion of Fab 11.2 in New

Mexico and ramping Fab 12 in Arizona, it has reduced spending from initial expectations because of increased yields. Micron Technology stopped investments in Lehi, Utah, but continues aggressive expansion and conversion to 200mm at all three fabs in Boise, Idaho. Memory-sensitive plant expansions such as Fujitsu's Gresham, Oregon, plant a n d Integrated Device Technology's Oregon fab have been delayed or have slower ramps, yet IBM has been very aggressive in its Burlington, Vermont, expansion of 16Mb DRAM production thus far. Logic investment has seen a slowdown as well, with LSI Logic's p u s h out of its Oregon fab, the delay of Motorola's North Carolina PowerPC fab, and the slow ramp-up of

Advanced Micro Devices' Fab 25 in Austin, Texas. Yet Cirrus Logic and

Lucent Technologies have increased spending dramatically to ramp fabs in the eastern United States, with Atmel expanding aggressively in

Colorado and Rockwell's emerging success creating opportunity for equipment companies in California. SGS-Thomson has remained aggressive in spending in the United States, placing firushing touches on its new

Arizona facility, as well as starting u p Fab 4 in CarroUton, Texas. Texas

Instruments remains aggressive, spending in capacity expansion for DSP chip capacity. Although it is not likely that either Samsung or TSMC will

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

16 Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide place equipment into their new U.S. fabs in 1996, capital spending on the shells is progressing.

Japan: DRAM Capacity Additions Stop, Investment in Teclinology Under Way

Japan's 49 percent increase in capital spending in 1995 is only 37 percent on a yen basis as Japanese companies look to invest outside Japan to optimize buying power. Japanese spending has essentially stopped growing and actually will decline about 2 percent in 1996. Because of the early cutback in spending, 1997 remains a flat capital spending year, as well, with only modest growth in 1998.

Some of the Japanese electronics giants that experienced good profit growth in 1995, driven by semiconductor operations, have seen those profits evaporate with the precipitous fall of DRAM prices. Although spending on capacity has essentially stopped, two other types of investment are likely to be important in Japan now through 1998.

First, Japanese companies will invest in any new technology and equipment targeted at the 0.25-micron production arena. This technology will not likely be in volume production until 1999, but the Japanese companies are expected to take advantage of this slowdown to understand and progress d o w n the learning curve on these new process technologies.

Second, the Japanese companies found that the shells built in 1990 and

1991 became an asset during the ramp in 1993 and 1994. By building a fab shell, equipped with a skeleton equipment set, they were positioned to more quickly ramp u p production w h e n the market turned u p . We see that same pattern repeating, so we would expect several new fabs to be started in 1996 and 1997, although at very low run rates. Once these fabs are in place, the Japanese companies can continue to review the market every six months, making course corrections in April and October, as they have been doing through the last cycle ramp.

I

Although new facilities by Japanese companies will come on line outside

Japan throughout the rest of this decade, DRAM investments inside Japan are really the only driving force today, although diversification has come to the forefront again in Japan. Japanese companies will continue to invest but will grow outside Japan faster than within Japan. We are therefore forecasting a below-average CAGR of 12.9 percent for the Japan region for

1995 through 2001.

One bright spot is that a PC boom could emerge in Japan over the next year or two, spawned by the networking infrastructure that is being built.

This would breathe new life into the Japanese semiconductor market and our forecast would be brightened a bit. We do not think that even a PC boom, however, would create a forecast different from several percentage points below the world average. The fundamentals of Japanese production capacity are still too heavily concentrated in DRAMs, with no clear future direction emerging as yet, which keeps us from being more optimistic about capital activities in Japan.

SEIVlM-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast 17

^

I

Europe Sustains Presence as a Growth Market

After a growth bubble of 46 percent in 1993, higher than expected, European spending "moderated" to a slower-than-market growth rate in 1994 after multinationals (Intel) substantially completed the majority of their expansions in 1993. The growth of 44 percent in 1994 is nonetheless extremely healthy, primarily being fueled by the European companies themselves—the ever-present SGS-Thomson, Philips expanding in

Nijmegen, the Netherlands, and Ericcson equipping its expansion.

Europe continued to attract capital in 1995, growing 65 percent. Large multinationals are still present, with Motorola upgrading the Scotland fab bought from Digital Equipment Corporation, the new IBM/Philips venture in Germany, Analog Devices' expansion in Ireland, Texas

Instruments' continued expansion in Italy, and the IBM/Siemens fab's continued ramp of 16Mb DRAMs in France. The key expansion is Siemens' new fab in Dresden, which was the key driver pulling Siemens into the top

10 in capital spending worldwide in 1996. Like the Uruted States, Europe is experiencing slowdowns this year. Although SGS-Thomson and

Siemens remain strong. Philips and the Japanese companies have pulled back investment in capacity significantly. We are looking for continued growth in 1996 of 22 percent as production continues to ramp from these and eight new fabs, most notably by GEC Plessey and TEMIC. However, in part because we do not believe that Siemens can sustain its current spending, we are calling for a 12 percent decline in spending in Europe for

1997 and a flat-to-down 1998 because we expect multinationals will ramp domestic memory fabs before Europe. Samsung has announced a fab to come on line in Europe but as yet is undecided about the exact location or timing.

Europe is viewed as a strategic location for production longer term to take better advantage of European and 16Mb DRAM growth in the future, driven by the PC production boom, without import tariffs. We therefore expect Europe to be a nearly average investment region in the long term, with a six-year CAGR of 15 percent.

I

Asia/Pacific Investing Focusing on Foundry as the DRAM Falls

The often erratic but sustained semiconductor capital spending growth in the Asia/Pacific region continued at the explosive rate of 112 percent in

1995. We expect a "moderated" growth of about 30 percent in 1996 as several new fab projects are built and equipped. However, the tide has turned in the DRAM area, and we are forecasting the Asia/Pacific region to be hit the hardest in 1997, with a nearly 15 decline in capital spending. Longer term, we expect Asia/Pacific to exhibit among the most aggressive growth in capital spending of any region. Da la quest forecasts a 1995-through-200l

CAGR of 20.3 percent.

Spending in 1995 and early 1996 came primarily from two areas, DRAMs and foundry capacity. The Korean conglomerates are continuing their relentless DRAM capacity expansion plans, although more moderately in

1996. We do expect these companies to succumb to the inevitable reality of overcapacity, with significant cutbacks for 1997.

SEMIVI-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

The real story of interest in 1995 and 1996 is the new Taiwan players.

Vanguard International brought on its new DRAM fab late last year, and

PowerChip Semiconductor and N a n Ya Plastics, among others, brought new DRAM capacity on line late last year and early 1996. All of these are targeted at 16Mb DRAM running 200mm wafers. Current players such as

TI/Acer and Mosel Vitelic were also increasing their spending with new projects. But the tide has turned quickly, likely accelerated by the fact that the Taiwanese stock market is very close to the U.S. stock market in its reaction to b a d news. Many companies in DRAM are now cutting back feverishly to save near-term profitability.

However, Taiwan chip companies TSMC, Macronix, and United

Microelectronics Corporation, along with Chartered Semiconductor in

Singapore a n d Submicron Technology in Thailand, continue with their plans to expand foundry capacity. InterCormect Technology, with its new foundry fab in Malaysia, is part of what we believe will be several new entries into this business. The reason for the continued interest in spending capital in this area comes from the fact that the core business is dependent on logic a n d PC unit demand rather than DRAM. Further, Dataquest estimates that only about 32 percent of the contracted manufacturing of semiconductors originates from fabless companies. The remainder is from integrated device manufacturers (IDMs) that wish to place the manufacturing of lower value-added products away from their own facilities in order to maximize resources and cost, that wish to reduce investment risks using foundries as an extension of their own capacity, or that wish to use the more advanced technology of foundries (in some cases) as a growth strategy.

The last few years have seen the flourishing of the dedicated foundry, mostly in Asia/Pacific. It is still believed that the largest concentration of foundry capacity in the world today, however, is in Japan, with companies like Rohm, Seiko-Epson, Sharp, a n d other large integrated companies. But with the investment trends in Asia/Pacific, the lead will likely change within the next five years.

However, the appetite for leading-edge foundries has caused another transformation. With the cost of capital increasing and with a higher level needed for leading-edge equipment, foundry companies have established longer-term contracts with customer companies, sometimes involving capital irrfusions toward production equipment and sometimes involving technology. In return, customer companies are looking for dedicated capacity allocations. Many joint venture have been announced in the last year, and w e expect this trend to continue. The foundry industry is now a strategic industry rather than simply a tactical one. With this transformation nearly complete, we are starting to see the dedicated investment to build new foundry capacity.

Who's Investing Where?

In our recently completed capital spending survey, Dataquest gathered information o n h o w money is being spent. Table 2-5 summarizes how companies based in different regions are spending their money abroad for

1995, and Table 2-6 summarizes this for 1996. About 79 percent of the money spent w e n t into domestic economies worldwide in 1995, and that

SEMIVI-WW-IVIT-9602 ©1996 Dataquest August 26,1996

Semiconductor Capital Spending Forecast 19 ratio increased slightly to 80 percent in 1996 as companies tend to cut back externally first.

Asia/Pacific comparues have historically placed all their investments domestically, but 1994 was the first year of diversification, which continued in 1995. Asia/Pacific companies spent about 4 percent of their money abroad in 1995, increasing to about 6 percent in 1996. We would expect this ratio to increase significantly over the next two to three years.

European companies have been the most aggressive capital exporters historically, placing only 59 percent of their investment inside Europe.

This figure grew slightly to 63 percent in 1996 and should expand in 1997 as European companies rein in spending.

Japanese companies are very close to the worldwide average, with about

17 percent domestic investment in 1995, rising to 80 percent in 1996.

Americas region companies are also high domestic spenders, with about

73 percent staying at home for both years.

The Americas and Japanese regions are net investors, while European and

Asia/Pacific regions are net beneficiaries of that investment. This parallels those regions being net exporters and net importers of semiconductors, respectively.

Table 2-5

Regional Investment Patterns of Semiconductor Manufacturers in 1995

(Millions of U.S. Dollars)

American Companies

Japanese Companies

European Companies

Asia/Pacific Companies

All Companies

Growth from 1994 (%)

Source: Dataquest (July 1996)

Worldwide

13,840.3

12,042.3

3,301.5

9,124.0

38,308.1

73.5

Americas

10,112.2

1,328.7

420.2

285.5

12,146.6

68.8

Japan

655.1

9,247.7

7.1

0

9,910.0

48.6

Europe

1,423.3

738.7

1,936.0

38.9

4,136.9

65.2

Asia/Pacific

1,649.6

727.3

938.1

8,799.6

12,114.6

111.8

Percentage of

World Spending

36.1

31.4

8.6

23.8

100.0

Table 2-6

Regional Investment Patterns of Semiconductor Manufacturers in 1996

(MilUons of U.S. Dollars)

American Companies

Japanese Companies

European Companies

Asia/Pacific Companies

All Companies

Growth from 1995 (%)

Source: Dataquest (July 1996)

Worldwide

17,261.3

11,201.5

3,802.0

13,079.2

45,344.0

18.4

Americas

12,598.0

1,051.5

359.7

818.8

14,828.1

22.1

Japan

742.4

8,949.2

0

0

9,691.7

-2.2

Europe

2,019.3

603.3

2,412.5

23.0

5,058.2

22.3

Asia/Pacific

1,901.5

597.4

1,029.8

12,237.4

15,766.1

30.1

Percentage of

World Spending

38.1

24.7

8.4

28.8

100.0

SEMI\/l-WW-MT-9602 ©1996 Dataquest

August 26,1996

20 Semiconductor Equipment, IVIanufacturing, and Materials Worldwide

Although all regions are spending in Asia/Pacific and all multinational regions are investing in Europe, only North American companies have the strategic vision to invest in Japan. Japanese companies are also investing on a worldwide basis. We believe this is one of the key elements necessary in a strategic plan for a semiconductor company to be competitive on a global basis.

Dataquest Perspective

The capital spending boom experienced in 1993 through the first half of

1996 is over, and the industry is now in what we would characterize as an

18-to-24-month pause, with investment in capacity initially declining and stabilizing as demand catches u p to supply. The marked downturn in the

DRAM investment cycle was triggered in part by the 1x16 configuration of the 16Mb DRAM achieving yields of about 60 percent to 65 percent, which occurred near the end of 1995.

In the second half of 1995, a complicating factor entered the picture—the larger-than-actual anticipated demand for memory created by the introduction of Windows 95, which, taken with the tight supply of memory last year, caused inventories to irrflate to enormous levels. This situation started correcting in the final quarter of last year, slightly in advance, but essentially on top of, the yield trigger for the 16Mb DRAM conversion timing. N o w we have a double-barreled driver (demand and supply) for the DRAM price slide and capital spending downturn. These events have increased the anticipated severity of the downturn.

The industry is now relying on the continued growth in personal computer unit sales, with a d d e d growth in telecommunications and networking products, to create a unit demand picture that will keep this s l o w d o w n short-lived from a historical perspective. The wafer fab capacity bubble has burst in all regions and for most semiconductor products, most notably DRAMs, mixed-signal devices, and analog devices. The 1995 spending growth was almost entirely driven by DRAM and microcomponent capacity purchases, 1996 was a year of transition, and 1997 is a year of investment in technology.

Based o n how capacity is migrating among device types, Dataquest believes that the first area of spending recovery will be advanced logic, as early as mid-1997. Equipment companies positioned for these markets will have a more moderate slowdown and perhaps can grow through this time if they have advanced technology. The microcontroller (MCU), analog, mixed-signal, and telecom chip capacity will be next to recover, but this will probably not occur until the end of 1997 or early 1998. The DRAM segment, the root cause of the problem and the very last to recover, is not expected to resume robust spending until mid-to-late 1998. The next major and broad investment cycle will have momentum by 1999.

We believe that capital spending may be influenced in late 1997 through

1999 positively with faciUty construction and equipment purchases for the world's first 300mm wafer fab. We have built this infrastructure investment into our model.

SEMM-WW-l\/IT-9602 ©1996 Dataquest August 26,1996

I

Chapter 3

Wafer Fab Equipment Forecast

This chapter presents data on worldwide spending by region for wafer fabrication equipment. Wafer fab equipment spending in a region includes spending by all semiconductor producers with plants in that region. Included are all classifications of equipment for front-end semiconductor operations.

Chapter Highlights

This chapter will discuss the following highlights:

• Wafer fab equipment spending growth in 1995 exceeded 1994, with

71 percent growth worldwide. However, overcapacity has gripped the industry, and wafer fab equipment growth will slow markedly to

17 percent in 1996 a n d suffer a decline in 1997.

• The only reasons that 1996 remains a double-digit growth year are that there were strong backlogs coming out of 1995 and because some companies increased spending early in the year. The second half of 1996 is expected to be lower than the first half.

• Segment growth in 1995 and the first half of 1996 was being led by

DRAM and capital spending-sensitive equipment, with steppers, implant, wafer inspection, and factory automation equipment exhibiting sigruficantly stronger-than-market growth. We believe the second half of 1996 and 1997 will concentrate on advanced technology equipment and some capacity additions in the advanced logic area.

• A purchasing behavior that will be prevalent over the next 12 to

18 months will be new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. These segments will be those related primarily to deep-UV lithography and inspection and 0.25-micron multilevel metallization process schemes.

• Dataquest's model that measures the net cumulative underinvestment or overinvestment indicates that by the end of 1995, the semiconductor manufacturing world was overinvested in wafer fab equipment to the tune of $7.0 billion, or 37 percent of the market. This is above the peaks exhibited in 1984 a n d 1989, so it is no surprise that excess capacity has emerged during 1996 in the DRAM market, where capacity has been a d d e d recently.

• After four strong expansion years in 1993 through 1996, equipment purchases in 1997 should decline, followed by a flat 1998. The next expansion should kick in by late 1998 or 1999, driven by 0.3-micron to

0.35-micron capacity expansion. The worldwide wafer fab equipment market is forecast to grow at 16.2 percent CAGR between 1995 and

2001, slightly above the semiconductor market growth.

• We have factored in an infrastructure investment in equipment for late

1997 through 1999, which will affect the forecast size of the markets positively. This additional investment will be for initial equipment to fill a couple of 300mm fabs to r u n silicon by 1999. The bulk of this 300mm equipment bubble occurs in 1998. However, our outlook for a significant 300mm equipment market will wait until 2001 to 2002.

SEMM-WW-MT-9602 ©1996 Dataquest 21

22 Semiconductor Equipment, Manufacturing, and Materials Worldwide

This chapter presents historical and forecast data on the worldwide wafer fabrication equipment market, by region and by key equipment segment.

In this midyear forecast for wafer fab equipment, w e have chosen to focus our forecast of equipment categories on specific segments and issues, iiamely:

• The annual investment theme for 1995 to 2000

• Steppers and automatic photoresist processing equipment (track)

• Dry etch and chemical mechanical polishing (CMP)

• Silicon epitaxy, chemical vapor deposition (CVD), and physical vapor deposition (PVD)

• Diffusion and RTF

• Ion implantation (medium current, high current, and high voltage)

• Segments that will fare best the next two years

These segments of the equipment market represent not only the majority of all wafer fab equipment expenditure in the world today, but also embody the key technological capability for advanced device production.

Highlights of some of the factors affecting individual equipment segment forecasts also are presented.

Equipment spending in a region refers to spending by all companies— both domestic and foreign—within the region. We note also that yearly exchange rate variations can have a significant effect on 1989-through-

1996 data appearing in the tables in this chapter. Appendix B details the exchange rates used in this document.

Table 3-1 provides historical market data, by geographic region for 1989 through 1995, and Table 3-2 shows forecast market data, by geographic region for 1995 through 2001. Table 3-3 shows historical data for key equipment segments for 1989 through 1995. Table 3-4 provides forecast data for key equipment segment for the years 1995 through 2001.

Table 3-1

W o r l d w i d e Wafer Fab E q u i p m e n t Market History, b y R e g i o n , 1989-1995

(Millions of U . S . Dollars)

Americas

Change(%)

Japan

1989

1,657

7.9

1990

1,589

-4.1

1991

1,519

-4.4

3,012

1992

1,576

3.8

2,098

Change(%)

Europe

2,813

23.9

721

2,996

6.5

764

0.5

641

-30.3

641

Change (%)

Asia/Pacific

8.7

820

6.0

522

-16.1

832

Change (%)

Total Wafer Fab Equipment

Change(%)

58.1

6,011

20.6

-36.4

5,871

-2.3

59.5

6,003

2.3

Note: Some columns may not add to totals shown because of rounding.

Source: Dataquest (July 1996)

0.0

783

-5.8

5,098

-15.1

1993

2,118

34.4

2,450

16.8

988

54.1

1,312

67.5

6,868

34.7

1994

3,190

50.6

3,661

49.4

1,370

38.6

2,567

95.7

10,787

57.1

1995

5,179

62.3

6,352

73.5

2,316

69.1

5,208

102.9

19,054

76.6

CAGR (%)

1989-1995

20.9

14.5

21.5

36.1

21.2

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Wafer Fab Equipment Forecast 23

Table 3-2

Worldwide Wafer Fab Equipment Market Forecast by Region, 1995-2001

(Millions of U.S. Dollars)

Americas

Change (%)

Japan

1995

5,179

62.3

6,352

1996

6,217

20.1

6,423

1997

5,453

-12.3

5,272

1998

6,163

13.0

5,635

Change (%)

Europe

Change (%)

Asia/Pacific

Change (%)

Total Wafer Fab Equipment

73.5

2,316

69.1

5,208

102.9

19,054

1.1

2,642

14.1

7,026

34.9

22,309

-17.9

2,428

-8.1

5,607

-20.2

18,760

Change (%) 76.6 17.1 -15.9

Note: Some columns may not add to totals shown because of rounding.

Source: Dataquest (July 1996)

6.9

2,486

2.4

5,616

0.2

19,900

6.1

1999

7,797

26.5

7,011

24.4

3,124

25.7

7,448

32.6

25,380

27.5

2000

10,530

35.1

9,679

38.1

4,253

36.1

11,757

57.8

36,219

42.7

2001

13,105

24.5

11,694

20.8

5,520

29.8

16,570

40.9

46,889

29.5

CAGR(%)

1995-2001

16.7

10.7

15.6

21.3

16.2

Table 3-3

Wafer Fab Equipment Revenue History by Equipment Segment, 1989-1995

(Millions of U.S. Dollars)

Equipment Segment

Worldwide Fab Equipment

Change(%)

1989

6,011

20.5

1990

5,871

-2.3

1991

6,003

2.2

1992

5,098

-15.1

1993

6,868

34.7

1994

10,787

57.1

1995

19,054

76.6

CAGR (%)

1989-1995

21.2

Steppers

Track

Maskmaking Lithography

Direct Write E-Beam

Other Lithography^

Total Lithography/Track

1,181

322

69

70

122

1,764

1,052

317

47

76

119

1,612

979

364

48

50

107

1,549

646

353

53

26

80

1,158

1,007

500

52

23

97

1,679

1,833

695

79

35

90

2,732

3332

1,413

82

38

80

4,944

Automated Wet Stations

Other Clean Process

Dry Strip

Dry Etch

Chemical Mechanical Polishing

Total Etch and Clean

243

134

121

670

NS

1,168

268

132

118

690

NS

1,208

291

143

119

717

11

1,281

286

103

123

682

20

1,213

285

198

138

1,083

44

1,748

468

213

213

1,592

64

2,548

928

389

369

2,842

197

4,724

25.0

19.5

20.4

27.2

NA

26.2

(Continued)

18.9

27.9

2.9

-9.7

-6.8

18.7

SEMM-WW-MT-9602 ©1996 Dataquest

August 26,1996

24

Semiconductor Equipment, iVIanufacturing, and Materials Worldwide

Table 3-3 (Continued)

Wafer Fab Equipment Revenue History by Equipment Segment, 1989-1995

(Millions of U.S. Dollars)

Equipment Segment

Tube CVD

Nontube Reactor CVD

Sputtering

Silicon Epitaxy

Other Deposition^

Total Deposition

1989

220

388

320

75

170

1,173

1990

259

457

359

68

153

1,296

1991

268

474

425

89

147

1,403

1992

213

437

446

84

119

1,300

1993

283

585

584

83

115

1,650

1994

442

885

1,012

114

101

2,553

1995

780

1,803

1,567

207

124

4,480

CAGR (%)

1989-1995

23.5

29.2

30.3

18.4

-5.1

25.0

Diffusion

RTF

Total Thermal Nondeposition

332

25

357

325

33

358

326

46

372

246

36

283

342

45

388

Medium-Current Implant

High-Current Implant

High-Voltage Implant

Total Ion Implantation

131

301

25

457

114

250

7

370

108

228

18

353

83

164

16

263

108

233

18

359

490

76

567

242

391

27

659

773

154

927

384

550

119

1,053

Optical Metrology

CD-SEM

Thin Film Measurement

Patterned Wafer Inspection

Auto Unpatterned Detection

Other Process ControP

Total Process Control

Factory Automation

Other Equipment

Total FA/Other Equipment

74

81

NS

116

37

369

676

59

88

NS

105

45

313

609

59

93

43

90

41

307

632

40

78

58

109

30

228

542

43

83

71

144

32

270

644

67

154

100

281

56

446

1,103

98

313

196

535

111

601

1,854

195

222

417

216

202

418

227

185

412

194

146

340

250

151

401

412

213

625

686

386

1,072

Total Wafer Fab Equipment 6,011

5,871

6,003

5,098 6,868 10,787

NS = Not surveyed

NA = Not applicable

1 Includes contact/proximity, projection aligners, and X-ray lithography

^includes evaporation, iVIOCVD, and iVIBE

^Includes auto review/classification, manual detection/review, and other process control equipment

Note: Some columns may not add to totals shown because of rounding.

Source: Dataquest (July 1996)

19,054

15.1

35.2

17.2

19.6

10.6

30.1

14.9

21.2

4.8

25.4

NA

29.1

20.1

8.5

18.3

23.3

9.6

17.0

SEMM-WW-MT-9602 ©1996 Dataquest

August 26,1996

Wafer Fab Equipment Forecast 25

Table 3-4

Wafer Fab Equipment Revenue Forecast by Equipment Segment, 1995-2001

(MilHons of U.S. Dollars)

Equipment Segment

Worldwide Fab Equipnient

Change (%)

1995

19,054

76.6

1996

22,309

17.1

1997

18,760

-15.9

1998

19,900

6.1

1999

25,380

27.5

2000

36,219

42.7

2001

46,889

29.5

CAGR (%)

1995-2001

16.2

Steppers

Track

Maskmaking Lithography-

Direct Write E-Beam

Other Lithography^

Total Lithography/Track

3,332

1,413

82

38

80

4,944

3,768

1,573

149

49

85

5,624

2,673

1,235

187

36

n

4,208

3,019

1,367

217

43

77

4,722

4,126

1,908

271

54

86

6,445

6,140

2,671

332

68

107

9,317

7,619

3,460

422

98

127

11,727

14.8

16.1

31.4

17.2

8.0

15.5

Automated Wet Stations

Other Clean Process

Dry Strip

Dry Etch

Chemical Mechanical Polishing

Total Etch and Clean

928

389

369

2,842

197

4,724

1,071

497

431

3,324

252

5,575

954

438

385

2,871

268

4,917

1,004

445

402

2,954

323

5,128

1,245

515

493

3,638

456

6,347

1,694

695

702

5,127

677

8,896

2,274

938

928

6,560

985

11,685

Tube CVD

Nontube Reactor CVD

Sputtering

Silicon Epitaxy

Other Deposition^

Total Deposition

780

1,803

1,567

207

124

4,480

872

2,182

1,798

277

118

5,247

746

1,955

1,573

172

89

4,635

788

2,068

1,638

299

85

4,878

974

2,556

2,046

389

96

6,061

1,358

3,690

2,888

524

118

8,579

1,711

5,017

3,775

694

136

11,333

Diffusion

RTP

Total Thermal Nondeposition

773

154

927

886

239

1,124

737

255

992

760

284

1,044

932

382

1,314

1,301

578

1,879

1,650

689

2,340

Medium-Current Implant

High-Current Implant

High-Voltage Implant

Total Ion Implantation

384

550

119

1,053

435

714

152

1,301

340

561

102

1,003

325

551

108

985

394

764

150

1,309

599

1,212

235

2,047

769

1,604

333

2,705

13.5

28.4

16.7

12.3

19.5

18.7

17.0

(Continued)

14.0

18.6

15.8

22.4

1.6

16.7

16.1

15.8

16.6

15.0

30.8

16.3

SEMM-WW-MT-9602 ©1996Dataquest

August 26,1996

26 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 3-4 (Continued)

Wafer Fab Equipment Revenue Forecast by Equipment Segment, 1995-2001

(Millions of U.S. Dollars)

Equipment Segment

Optical Metrology

CD-SEM

Thin Film Measurement

Patterned Wafer Inspection

Auto Unpatterned Detection

Other Process C o n t r o l

Total Process Control

1995

98

313

196

535

111

601

1,854

1996

103

339

232

609

145

776

2,204

1997

76

264

210

501

138

665

1,853

1998

81

307

225

532

132

666

1,942

1999

106

421

291

670

143

776

2,408

2000

146

574

364

998

175

1,080

3,337

2001

183

722

464

1,266

244

1,500

4,379

CAGR (%)

1995-2001

11.0

15.0

15.4

15.4

14.1

16.5

15.4

Factory Automation

Other Equipment

Total F A / O t h e r Equipment

686

386

1,072

834

399

1,234

793

359

1,152

798

404

1,201

1,023

473

1,496

1,498

667

2,164

1,876

844

2,720

18.2

13.9

16.8

Total Wafer Fab Equipment 19,054 22,309 18,760 19,900 25,380

'Includes contact/proximity, projection aligners, and X-ray lithography

^Includes evaporation, MOCVD, and MBE

^Includes auto review/classification, manual detection/review, and other process control equipment

Note: Some columns may not add to totals shown because of rounding.

Source: Dataquest (July 1996)

36,219 46,889

16.2

Annual Investment Themes for 1996 through 2001

Behind our equipment and segment forecasts are assumptions about how semiconductor producers will perform and invest. These are summarized in Table 3-5 for the years 1996 through 2001. The following areas are considered: the availability of profits for reinvestment, memory versus logic growth, technology shifts, and brick and mortar versus equipment purchases.

When Will Demand Expand to Meet Capacity? An Update to the

Overinvestment or Underinvestment Model

In Dataquest's forecasts for the last couple of years, we have shown a model that provided a measure of the net cumulative overinvestment or underinvestment in wafer fab equipment to support capacity needs in the industry. Because equipment purchases precede actual capacity on line by a number of months or quarters,, this model could be viewed as a gross leading indicator of capacity shortages and excesses. The results of this model are closer to a 1.5-year to three-year indicator of turning points in the equipment industry. The methodology of the net cumulative investment (NCI) model is linked to our longer-range forecast model.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Wafer Fab Equipment Forecast

f

Table 3-5

Annual D r i v i n g Forces and I n v e s t m e n t T h e m e s for Wafer Fab Equipment, 1996-2001

Logic Semiconductor Unit

Growth

Investment in Logic Capacity

Memory Semiconductor Unit

(Not Bit) Growth

1996

Moderate

Moderate

Moderate

Investment in Memory Capacity Moderate

Front-End Equipment versus

Facilities Loading of Capital

Balanced

Primary Technologies Invested 0.35-0.5 micron

Note: Scale = strong > solid > moderate > weak > dead

Source: Dataquest (August 1996)

1997

Moderate to Solid

Moderate

Weak

Dead

Facilities

0.3-0.5 micron

1998

Solid

Solid

Moderate

Weak

Facilities

0.3-0.4 micron

1999

SoHd

Solid

Solid

Moderate

Equipment

0.25-0.4 micron

2000

Strong

Strong

Strong

Strong

Equipment

0.25-0.4 micron

2001

SoHd

SoUd

Moderate

Solid

Balanced

0.2-0.35 micron

This methodology starts with a few key assumptions and baselines:

• Long-term growth rates for semiconductors and wafer fab equipment are correlated. In other words, semiconductor revenue and profits are needed before money can be spent on equipment and vice versa.

• Also, net cumulative investment equals zero over time ... meaning that in a noncyclical environment where annual growth rates are constant, investment and capacity are at equilibrium at all times. Of course, this industry cycles through overinvestment and underinvestment.

• The output is a tangible number and is in dollars of over or underinvestment at year's end. However, the more useful output of the model divides this gross dollar number by the wafer fab equipment market size. The result is a percentage of market figure that is repeatable in level from cycle to cycle.

• To take into consideration the long-term growth of the semiconductor and equipment industries, the model has a factor allowing the fundamentals of the industry to change over time.

A net positive or negative investment is calculated relative to the longterm growth baseline annually and then added to the prior year. The calculation resulted in a dollar value net cumulative overinvestment or underinvestment and has correlated well with historical patterns.

Figures 3-1 and 3-2 show the most recent results of the model, little changed from our previous forecast update. In absolute dollar terms, by the end of 1995 the industry was $7.0 billion overinvested, or 37 percent of the wafer fab equipment market, exceeding levels witnessed during the

1984 and 1989 peaks. These levels are occurring for two basic reasons.

First, PC unit demand continued to show annual growth in the range of the low twenties. About one-third of the semiconductor industry and over one-half of the capital spending on new capacity is to support this demand.

I

August 26,1996

SEMM-WW-MT-9602 ©1996 Dataquest

27

28 Semiconductor Equipment, IVIanufacturing, and Materials Worldwide

Figure 3-1

Net Cimiulative Overinvestment and Underinvestment of Semiconductor Wafer Fab ^

Equipment (Millions of U.S. Dollars) "

Millions of Dollars

8,000 IT

6,000

m

4,000

2,000

ii

I

-2,000

-4,000

^ ,

^^%^_m7%^

T I r^

^ . g ^ .

imr^

^

••

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

965716

Source: Dataquest (August 1996)

Figure 3-2

Net Cimiulative Overinvestment and Underinvestment of Semiconductor Wafer Fab

Equipment (Percentage of Wafer Fab Equipment Market)

Percent of Wafer Fab Equipment Market

40-ri

30

20

WL

1

10H

m

I I I " " " I ^^^T

.^3.

T r

M

-10

-20

-30

I

m

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

96S717

Source: Dataquest (August 1996)

©1996 Dataquest August 26,1996

SEMM-WW-MT-9602

Water Fafa Equipment Forecast 29

^

^

I

Second, the DRAM market had not yet converted to r u n the more siliconefficient 16Mb DRAM, placing this investment cycle about seven years behind the last cycle. DRAM bit demand generally runs at 50 percent annually, a n d DRAM manufacturing has depended on increasing unit densities (increasing the bits per square inch) to meet this demand.

Shrinks of existing generations alone bring only 15 percent to 25 percent armual bit per square inch efficiencies, while converting a fab running

4Mb DRAMs to 16Mb DRAMs would increase bits per square inch by two to three times. Because low yields of the wider-configuration DRAM held back the economic conversion of the market, top-line bit d e m a n d is translating to square inch demand (silicon) and the equipment to run it.

Because the equipment being installed is fully convertible to r u n 16Mb

DRAMs, w e can think of these fabs building "pent-up supply" in bits.

Once 16Mb DRAM yields (for the 1x16 configuration) exceeded 60 percent to 65 percent, it is more economical to r u n these lines, and DRAM prices eroded.

We have factored into the model a n investment in a couple of 300mm fabs starting in late 1997 through 1999, w i t h the bulk in 1998. This is considered an equipment "bubble demand" because the equipment will be shipped into a nonproductive fab (meaning no semiconductor revenue will be generated initially).

With our forecast for a momentum-style growth in 1996 and two pause years in 1997 a n d 1998, the model indicates a reacceleration of equipment spending starting in 1999.

When Will Demand Expand to Meet Capacity? A Fundamental Set of Analyses

The NCI model described above is only a tool to indicate possible future turning points a n d should not be relied on to actually forecast capacity supply versus demand absolutely. A more fundamental, basic approach is required—looking at square inches of silicon capacity. H o w many wafers can be processed, and by what type of process, is much more enlightening in measuring capacity than measuring it for a particular device type.

Capacity is very fluid; a stepper does not care whether the picture it takes is for a DRAM, SRAM, or logic device. But there are limits in transferring capacity. For example, logic processes have specialized process techniques that are not found in DRAMs and vice versa. SRAMs can use a DRAM equipment mix or a logic-oriented process scheme; the latter tends to have faster access times.

There are two major markets we can isolate in order to understand basic capacity supply and demand—the DRAM market and the foundry market. The latter is particularly interesting for two reasons. First, foundry capacity has tended to be more heavily oriented toward logic and ASIC processes, giving u s a second perspective o n capacity versus supply.

Second, the major customer base for the foundry is the fabless company, whose products tend to be those that are placed within the PC logic and graphics chipsets. Because PCs n o w account for about one-third of all semiconductor use, about two-thirds of all DRAM consumption, and are the main engine for the current semiconductor boom, looking at PCrelated capacity issues is important for understanding potential equipment market turning points.

SEMM-WW-MT-9602 ©1996 Dataqu est August 26,1996

30 Semiconductor Equipment, Manufacturing, and Materials Worldwide

The details of these analyses are provided in other reports, such as

Dataquest's DRAM Supply/Demand Quarterly Statistics a n d ongoing research on the foundry market. A summary of the basic results and impacts will be given here.

In any s u p p l y / d e m a n d trend, there is a cycle between oversupply and undersupply. Investment in capacity tends to b e in reaction to these situations and there is inevitably overshoot in both directions. Analysis is based on square inches of silicon and not on revenue, bit-demand, or unit demand. If demand for silicon area exceeds supply, the market is technically in undersupply. We will refer to the maximum undersupply point as the market "pinch point."

The following are the basic conclusions and impacts:

• The DRAM market's pinch point was during 1995, and, with the conversion to a more silicon-efficient (in terms of bits per square inch)

16Mb triggered, it caused the industry to step into oversupply quickly.

This condition is expected to last into early 1998.

• The foundry market is split. There is oversupply at 0.6 micron and higher, but the market is in undersupply at 0.5 micron and below. The market is expected to remain in undercapacity at 0.5 micron through early 1997.

• 1996 is a year of transition and spotty investment in capacity. DRAM prices crashed early in the year. The market begins to transition to the

16Mb during the year, a n d prices will continue to decline gradually into

1997. Foundry prices are expected to be firm for advanced technology but edge downward during the year.

• 1997 will be a year of contraction, with weak DRAM capacity investment but continued strong investment in technology with logic capacity growth coming back later in the year. But the foundry capacity expansions will likely lose steam through the year into 1998. We expect

DRAM investment to decline some 20 to 30 percent, with logic investment increasing 5 to 10 percent.

• 1998 will be a frustrating year. Even though DRAM capacity is likely to see its peak oversupply in 1997, significant reinvestment will not likely occur until late in 1998 or 1999 as companies delay investment a n d

Japanese companies finish 200mm conversions. As the foundry market approaches oversupply, logic investment will cool. Yet we expect that semiconductor d e m a n d will begin to accelerate, signaling a underlying strength in the market and anticipation of the eventual upturn.

• 1999 becomes the first growth year in the next b o o m cycle. DRAM investment is likely to pick u p again, while foundry investment will lag.

U.S. logic capacity will likely be strong and lead foundry investment out of the pause. By 2000, the next equipment boom will be well u n d e r way and likely to last into 2001.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Wafer Fab Equipment Forecast 31

Highlights of Key Equipment Segment Markets and Forecasts

Steppers and Track

From 1989, the peak year of stepper shipments at more than 950 units, the market tumbled to less than 400 tools shipped in 1992 before recovering.

During this DRAM-sensitive ramp, the industry experienced its first year exceeding 1,000 steppers shipped, indeed 1,228 steppers were shipped in

1995. The peak year will be tikis year, in which w e expect over 1,300 steppers to ship. Shifts in the product mix toward higher-priced i-line systems and wide field lenses have also driven u p average selling prices (ASPs).

This trend will be offset somewhat in 1996 by the weak yen to drive a revenue increase of 13 percent o n a dollar basis in 1996, slightly below the market growth. The stepper segment, being capacity-serisitive, is forecast to lag the market's performance in 1997 and 1998 and then lead market growth starting in 1999.

Stepper revenue is forecast to grow at a 14.8 percent CAGR, slightly below the market average for 1995 through 2001. Our forecast for stepper unit growth over the five-year forecast horizon remains modest, about 4.5 percent CAGR between 1995 a n d 2001.

With the adoption of phase-shift mask technology and off-axis illumination techruques, as well as conventional i-line tools with variable NA, i-line is clearly a viable technology d o w n to the 0.3-micron regime and will continue to dominate the overall stepper technology mix through the year

2000. Excimer/deep-UV steppers will begin to represent a more significant portion of the product mix from 1997 onward for use in below-

0.3-micron devices and ICs w i t h large chip areas such as advanced microprocessors that require large field size capability. Dataquest believes that field size pressures accompanied b y shrinking geometry will drive the industry toward step-and-scan (or step-and-stitch) technologies for the majority of excimer/deep-UV shipnients, begirming in 1997.

Track equipment is forecast to grow at a 16.1 percent CAGR between 1995 and 2001, essentially equivalent to the industry growth of 16.2 percent.

Although w e believe that the rapid shift in the product mix toward higher-priced systems was recently completed, we do expect another product shift to occur in the track market, associated with the ramp of deep-UV steppers, which require more sophisticated environmental control systems and will translate to higher ASPs.

Etch and Clean: Dry Etch and Chemical Mechanical Polishing

Dataquest began covering the chemical mechanical polishing (CMP) market in 1993. At this time, Dataquest includes the post-CMP clean system, usually sold in conjunction with a CMP tool, as part of the cleaning segment, and not in the CMP segment. (This will likely change when statistics are collected for 1996 because there are many systems sold today as integrated CMP and clean.) The year 1994 was a time of evaluating the technology and gaining experience with it in production before widespread adoption into production.

SEMM-WW-l\/IT-9602 ©1996 Dataquest August 26,1996

32 Semiconductor Equipment, Manufacturing, and Materials Worldwide

In 1995, that adoption started, with CMP systems growing 208 percent to

$197 million. Even though the application appears to be limited to devices with at least three levels of metal, which tends to exclude the DRAM market, the acceptance of the technology into logic and particularly the foundry market has been the key turning point. Based on the demand of the customer, however, and because of tihe introduction of more robust equipment, w e now believe most foundries offer CMP at 0.5 micron.

These systems are used to remove material from the surface of the wafer, resulting in a flat surface over the entire wafer. This global planarization, primarily of dielectric layers, is required to achieve Wgh yields in devices where three or more levels of metal are used. Today's advanced logic and

ASIC devices are fueling this market growth. Dataquest believes that this technology and market will become a major part of semiconductor manufacturing in the long run.

If we were to make a forecast based purely on technology driving the market, we would not be slowing the CMP market forecast in 1997 and 1998.

However, we believe that time will see some holding back of capital investment, a n d history has shown us that even advanced, emerging technologies are rarely spared in a capital slowdown. Nevertheless, CMP is our fastest-growing segment, with a 31 percent CAGR for 1995 through

2001.

Dry etch systems continue to exhibit strong revenue growth, with a CAGR of 15 percent forecast for the years 1995 through 2001. Unit shipments are expected to grow as greater multilevel intercormect process capacity is brought online, increasing the need for dielectric and metal capacity.

Relatively strong ASP growth will lend additional momentum to dry etch revenue growth as new high-density plasma systems for 0.35-micron applications enter the market and multichamber cluster tools continue to increase their presence. The success of CMP will hold etch below market growth, however, particularly in metal etch, as stringer removal becomes a nonissue.

Deposition: CVD, PVD, and Silicon Epitaxy

CVD equipment sales are predicted to grow at a 17.3 percent annualized rate from 1995 through 2001, above overall equipment growth. The steady growth in multilevel intercormect capacity will continue to generate demand for dielectric and metal CVD systems. ASP growth will also contribute to revenue growth as more highly integrated systems with improved productivity and particle control appear in the marketplace.

Advanced dielectric deposition systems utilizing high-density plasma

(HDP) sources have been introduced and will gain momentum for intermetal dielectric (IMD) applications for sub-0.5 micron processes. Metal

CVD will continue to exhibit strong growth, driven by blanket tungsten for contact and via plugs, CVD barrier metals such as CVD titanium nitride, and dichlorosilane (DCS) tungsten silicide for shrink 16Mb a n d

64Mb DRAMs. For these reasons, the forecast for nontube CVD systems outperforms tube furnaces.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Wafer Fab Equipment Forecast 33

Sputter deposition systems are forecast to grow at a n aimualized rate of

15.8 percent for 1995 through 2001. As in the case of dry etch and CVD equipment, continued expansion of multilevel interconnect process capacity is the primary driver behind sputter system growth. Rapid growth in average system ASP helped to drive total revenue growth before 1995, primarily from the quick and expanding dominance of Applied Materials in the market. With Applied now accounting for more than 50 percent of the market, the bulk of the ASP increases are behind us. Changes in system architecture, pioneered by the Applied Materials Endura system, will continue to yield improvements in film properties, equipment productivity, and defect density. This is a market segment that will be somewhat buffered from a slowdown in DRAM investment, as the fundamental growth in the number of metal layers in ASICs and logic devices drives a more stable outlook.

The shift from batch to single-wafer epitaxial systems has been the primary driver of epitaxial deposition systems, given the need for 200mm epitaxial wafer capacity for CMOS logic applications. However, this capacity is more expensive than wafer suppliers would like, so we expect the concept of "minibatches" to emerge as a viable production strategy, as it has in CVD. Moore Technologies is known to have such a product on the market. A strong automotive, power, and discrete device market has increased d e m a n d for the specialty batch units, and growth in this segment actually exceeded CMOS logic in 1995 and continues to be strong in

1996. There are clear indications that epitaxial layers will be required for some 64Mb DRAMs (see Chapter 4). This, an increased product mix of logic semiconductors at the 200mm wafer capacity, and sustained demand for discrete devices and power ICs will be the primary drivers for epitaxial deposition equipment growth. In fact, epitaxial reactors are expected to be a star performer with an average annual growth of 22 percent and will be somewhat insulated from the near-term slowdown.

Thermal Nondeposition Processes: Diffusion and RTP

Diffusion systems are expected to demonstrate a CAGR of 13.5 percent for 1995 through 2001. Newer vertical systems will be configurable as multitube clusters with integrated dry clean capability to compete with single-wafer cluster tools. Tube systems will also incorporate small batch capabilities to offer greater flexibility for custom and semicustom circuit manufacturers.

RTP will grow at an annualized rate of 28.4 percent for 1995 through 2001.

This market grew much faster than we anticipated, nearly doubling in

1994 and slightiy more than doubling in 1995. The growth in 1994 was primarily fueled by the growing acceptance of self-aligned silicide processes in logic process flows. The growth in 1995 comes from new offerings in the market from Applied Materials, CVC, and Mattson Technology, and from the expansion of the application into traditional tube diffusion steps. The real growth for this segment will come from transitioning of the thermal

"nondepositing" processes away from diffusion tubes and into singlewafer RTP systems for 300mm wafers. We have factored in a large complement of systems into initial 300mm facilities starting in late 1997, largely contributing to the higher-than-market growth. RTP systems are primarily used today for salicide armeals and some implant drive-ins and are largely driven by logic and ASIC capacity.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

34 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Dataquest believes that batch tube systems will continue to resist penetration by RTP in areas such as well drive, BPSG reflow, and thermal oxidation because of the demonstrated cost-of-ownership benefits in these areas, at least through 200mm wafers. For 300mm wafers, there will also be a strong desire on the part of the semiconductor manufacturer to continue to use the batch tube systems since these systems offer much better cost efficiencies.

Ion Implantation

Overall ion implantation system revenue is forecast to grow at a CAGR of

17 percent for the years 1995 through 2001. This market segment will continue to be one of the most volatile, because of the highly device-specific nature of the implant segments and because of the dependence on new fab capacity for unit growth. The fastest-growing segment is expected to be high-energy implantation, which is evoking intense interest because of its potential for process simplification and manufacturing cost reduction.

The first year of true production ramp is expected to occur in 1998 as

0.4-micron technologies become mainstream, although early adopters such as Samsung have placed high-voltage implant into 16Mb DRAMs.

New implant systems will continue to offer improvement in uniformity, particle control, charging, and wafer throughput. The number of implant steps requiring medium-current ion sources is expected to grow faster than high-dose implant steps, again driven by the higher worldwide semiconductor logic component, with the shallow junctions preferentially driving the trend toward medium-current implants.

However, our forecast does not reflect this trend, with medium-current implant sales lagging the market with only 12.3 percent CAGR. Why?

Recently Eaton Corporation, Applied Materials, and Genus Inc. have introduced expanded capability to existing systems or new systems that are targeted to compete in more than one segment. For example. Applied

Materials' 9500 systems now basically have a range to cover both highcurrent and medium-current capabilities, and Genus' new 1520 model expands the range effectively across high-voltage and medium-current.

Traditional batch medium-current systems are effectively being squeezed out of the market. This is occurring because equipment utilization rates for implanters tend to be among the lowest in the fab. Semiconductor manufacturers, in an effort to increase utilization and reduce cost, are tending to buy equipment with broader ranges.

There are medium-current applications that will still require dedicated medium-current systems, namely high-tilt (greater than 42 degree) implants and V-p adjustment implants. However, these will be better executed using implanters with single-wafer end stations rather than the traditional batch systems.

Dataquest is investigating the redefinition of ion implant segments in order to capture this market dynamic better.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Wafer Fab Equipment Forecast 35

The Segments That Will Fare Best in the Next Two Years

During the coming slowdown, there will be two kinds of purchasing behavior that equipment companies can take advantage of to buffer sales declines. The first behavior is tied to which types of capacity will be required early in the recovery cycle. If there is a heavier dependence on advanced logic or supply to the materials industry, these segments wiU fare better than DRAM-dependent or semiconductor capacity-dependent segments. Included in this category are the segments of nontube CVD, sputtering, silicon epitaxy, maskmaking lithography, and process control systems in the materials business and RTF.

The second purchasing behavior that will be prevalent over the next 12 to

18 months will be purchasing of new processes a n d equipment directed at solving issues for 0.35-micron a n d 0.25-micron manufacturing. By no means will this listing be complete, but we will highlight a few of these areas here.

The movement to deep-UV lithography will pick u p steam through this slowdown. The unit forecast for deep-UV steppers is for more than

200 systems to ship in 1998, u p from the 100 or so shipping today. There are several equipment segments that will benefit from this movement — those inspection and process control systems that m u s t operate in the deep-UV wavelength to detect defects and materials on the surface of the wafer, namely mask inspection and thin-film measurement equipment.

Thin-film measurement is key to intralevel and interlevel metal interconnect and storage capacitance applications. Thin-film measurement systems are used in line to monitor the in-etch, lithography (photoresist), deposition, and diffusion steps. Although this market was driven b y logic applications more than memory production, in the past several years,

DRAM manufacturers have begun integrating thin-film measurement stations into their process lines.

Chemical mechanical polishing systems are becoming more robust and able to meet tighter specifications. Japanese and Asia/Pacific companies are just beginning to investigate these systems. Although dielectric CMP is relatively common, metal CMP is still emerging. Above three levels of metal, it becomes more beneficial to implement a metal CMP step to the process.

Although high-density etch has been in use and has relatively high market penetration for polysilicon and metal etch, almost no high-density etchers are used in oxide or dielectric etch. The need exists for such a robust process, and Lam Research has just introduced a system that could help penetrate this market yet to be won.

Gap-fill dielectric CVD has been a highly competitive development arena.

Novellus Systems appears to have product m o m e n t u m with its Speed product line. We believe this product area offers a way for companies to effectively buffer the current slowdown.

Wafer cleaning systems can also offer an area in which a company's technology can shine. The recent armouncement by Steag AG in Germany of its vapor dryer was built into a successful strategy w h e n it sold these

SEMl\/l-WW-MT-9602 ©1996 Dataquest August 26,1996

36 Semiconductor Equipment, Manufacturing, and Materials Worldwide systems as part of an integrated automated wet clean station. Wafers are placed in the dryer system after a wet cleaning process to dry but do not spin. Instead, they are pulled from a bath of isopropyl alcohol (IPA) and water in a controlled fashion into an IPA vapor atmosphere. As the wafer is pulled out, the liquid sheets off the surface, leaving no water spots, which often hold killer residue defects. The concept of automating cleaning processes will be a key product issue over the next few years.

Dataquest Perspective

Wafer fab equipment spending is expected to grow 17 percent worldwide in 1996. The only reasons that 1996 remains a double-digit growth year are that there were strong backlogs coming out of 1995 and that there was increased spending by some companies early in the year.

After three and a half strong expansion years in 1993 through 1996, equipment purchases in 1997 should decline markedly, followed by an essentially flat 1998. Investment in DRAM capacity will be curtailed as producers elect to convert their 4Mb DRAM capacity to 16Mb, which adds bit capacity through the instant increase in bits per square inch. Also, many Japanese DRAM facilities now running 150mm wafers will convert to 200mm wafers, further delaying the need for new equipment. DRAMsensitive equipment technologies or capital-intensive segments will be affected more than logic-sensitive technologies. The next expansion should kick in by 1999, driven by 0.3-micron to 0.35-micron process capacity expansion.

Dataquest has factored in an infrastructure investment in equipment for late 1997 through 1999, which will affect the forecast size of the markets positively. This additional investment will be for irutial equipment to fill a couple of 300mm fabs to run silicon by 1999. However, our outlook for a significant 300mm equipment market will wait until well after 2000.

During the coming slowdown, there will be two kinds of purchasing behavior that equipment companies can take advantage of to buffer sales declines. The first behavior is tied to which types of capacity will be required early in the recovery cycle. If there is a heavier dependence on advanced logic or supply to the materials industry, these segments will fare better than DRAM-dependent or semiconductor capacity-dependent segments. Included in this category are the segments of nontube CVD, sputtering, metal etch, silicon epitaxy, maskmaking lithography, and process control systems in the materials business and RTF.

The second purchasing behavior that will be prevalent over the next 12 to

18 months will be purchasing new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. These segments will be those related primarily to deep-UV lithography and inspection and 0.25-micron multilevel metallization schemes.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

I

Chapter 4

Silicon Wafer Forecast

Dataquest's current forecast for regional silicon wafer demand reflects significant silicon wafer growth in 1995 and 1996, with slower growth in 1997 and 1998 as the industry migrates from the 4Mb DRAM to the more silicon-efficient 16Mb DRAM (in terms of bits per square inch). A resumption of accelerated growth starting in late 1998 into 1999 and beyond is in line with an increased PC unit, telecommunications equipment, and, therefore, semiconductor consumption forecast worldwide. Our latest forecast, along with highlights of some of the key factors affecting the regional markets, is presented below.

Silicon Forecast Tables

Tables in this chapter include Dataquest's most recent forecasts of regional unit silicon wafer consumption. Tables 4-1 through 4-5 detail unit consumption by region. Individual forecasts of major product segments such as prime, epitaxial, and test and monitor wafers are included. Tables 4-6 through 4-10 present regional forecasts for wafer size distribution.

>

>

The 200mm Wafer Ramps Up—Suppliers Have Responded with Capacity

Dataquest has been studying the subject of 200mm wafers and their ramp rate closely over the last two years, particularly in light of the massive number of announcements in fab capacity and requirements through the rest of the decade. We are expecting that more than 200 fabs processing

200mm wafers will be on line by the year 2000. Future demand for 200mm product wafers has been affected positively by the semiconductor industry's excess capacity. Why? Semiconductor comparues now have some time to convert 150mm fabs to the more cost-effective 200mm wafers. This has added incremental product wafer demand to the industry at 200mm wafers. A large factor in 200mm demand, the consumption of test wafers, is expected to grow more slowly than once thought as semiconductor manufacturers have responded to the tight supply of wafers by accelerating test wafer reduction programs. However, the industry is consuming just under the ratio of one test wafer for every product wafer. This compares with the industry average of about one test wafer per seven product wafers for sizes below 200mm. By the year 2000, we now expect this ratio to be one test wafer for over three product wafers.

Wafer suppliers have answered the need with several new wafer plants and billions of dollars of conunitted investment. We continue to believe that the ramp of 200mm wafers will be supply-constrained from time to time through the decade, although the faster-than-anticipated conversion to the 16Mb DRAM will mean adequate supply to meet demand in 1997 and early 1998.

In this forecast update, we have revised the forecast for 200mm wafers— again upward—to reflect not only the expected fab activity in each region but also the dramatic increase in the supply commitment being made by the silicon manufacturers since our last forecast update. Dataquest's cur-

SEMM-WW-MT-9602 ©1996 Dataquest 37

>

C=

( Q

1=

C/>

f - +

N3

0 5

U3

CD

CD

Crt

CO o

@

( O

( O

OJ

o

K-

(U

X3

Table 4-1

Forecast of Captive and Merchant Silicon* and Merchant Epitaxial Wafers by Region (Million

Worldwide Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

1992

2,096.7

1,848.2

248.5

1993

2,446.2

2,151.7

294.5

1994

2,907.6

2,520.3

387.3

1995

3,523.8

3,003.8

520.0

1996

4,102.2

3,515.6

586.6

1997

4,544.7

3,902.9

641.8

1998

4,943.8

4,210.1

733.7

1999

5,449.7

4,601.8

847.9

Americas Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

Japan Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

Europe Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

Asia/Pacific Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

1

"Includes prime, test, and monitor wafers

Source: Dataquest (July 1996)

650.9

522.8

128.1

972.8

885.8

87.0

235.0

211.1

23.9

238.0

228.5

9.5

716.9

560.7

156.2

1,134.3

1,045.2

89.1

289.6

253.4

36.2

305.4

292.4

13.0

829.0

634.2

194.8

1,277.9

1,159.5

118.4

353.8

297.7

56.1

446.9

428.9

18.0

943.1

696.4

246.7

1,483.4

1,319.7

163.7

457.7

373.6

84.1

639.6

614.1

25.5

1,046.4

770.8

275.6

1,675.0

1,492.5

182.5

514.5

415.4

99.1

866.3

836.9

29.4

1,149.2

851.2

298.0

1,803.6

1,602.7

200.9

562.6

451.6

111.0

1,029.3

997.4

31.9

1,257.2

926.9

330.3

1,949.3

1,712.7

236.6

606.5

480.0

126.5

1,130.8

1,090.5

40.3

1,400.5

1,024.7

375.8

2,111.9

1,841.3

270.6

658.9

511.9

147.0

1,278.4

1,223.9

54.5

> c:

<o c:

CO

»-+ tV3

cn

CO

CO

C7)

CO

C D

o>

O to

@

CO

C D

<y>

o

Pi

Table 4-2

Forecast Growth Rates of Captive and Merchant Silicon* and Merchant Epitaxial Wafers by R

Worldwide Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

1992

2.5

1.2

13.2

1993

16.7

16.4

18.5

1994

18.9

17.1

31.5

1995

21.2

19.2

34.3

1996

16.4

17.0

12.8

1997

10.8

11.0

9.4

1998

8.8

7.9

14.3

Americas Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

Japan Total Silicon + Epitaxial

Mercliant and Captive Silicon*

Epitaxial Silicon

Europe Total Silicon + Epitaxial

Merchant and Captive Silicon*

Epitaxial Silicon

Asia/Pacific Total Silicon + Epitaxial

Merchant and Captive SiHcon*

Epitaxial Silicon

'Includes prime, test, and monitor wafers

Source: Dataquest (July 1996)

10.0

3.2

52.1

-6.9

-5.9

-16.3

1.8

1.5

3.9

30.7

31.4

14.5

10.1

7.2

21.9

16.6

18.0

2.4

23.2

20.0

51.5

28.3

28.0

36.8

15.6

22.2

17.5

55.0

46.3

46.7

38.5

13.1

24.7

12.7

10.9

32.9

13.8

29.4

25.5

49.9

43.1

43.2

41.7

9.8

26.6

16.1

13.8

38.3

11.0

10.7

11.7

12.9

13.1

11.5

12.4

11.2

17.8

35.4

36.3

15.3

9.8

10.4

8.1

7.7

7A

10.1

9.3

8.7

12.0

18.8

19.2

8.5

9.4

8.9

10.8

8.1

6.9

17.8

7.8

6.3

14.0

9.9

9.3

26.3

>

c=

<a c/>

r-l-

tV3

CD

CO

CO

ty> to

CT>

O

N3

@

CO

CO

0 5

O

K.

0>

CD

CO

Table 4-3

Forecast of Captive and Merchant Silicon* Wafers by Region (Millions of Square Inches)

Worldwide Total SilicJOft*

Merchant Silicon

Captive Silicon

1992

1,848.2

1,728.2

120.0

1993

2,151.7

2,026.7

125.0

1994

2,520.3

2,373.3

147.0

1995

3,003.8

2,929.8

74.0

1996

3,515.6

3,441.6

74.0

1997

3,902.9

3,828.9

74.0

1998

4,210.1

4,136.1

74.0

1999

4,601.8

4,527.8

74.0

Americas Total Silicoft?

Merchant Silicon

Captive Silicon

Japan Total Silicon*

522.8

450.8

72.0

885.8

Merchant Silicon

Captive Silicon

Europe Total Silicon*

Merchant Silicon

848.8

37.0

211.1

206.1

5.0 Captive Silicon

Asia/Pacific Total Silicon*

Merchant Silicon

Captive Silicon

NA = Not applicable

"Includes prime, test, and monitor wafers

Source: Dataquest (July 1996)

228.5

222.5

6.0

560.7

486.7

74.0

1,045.2

1,008.2

37.0

253.4

248.4

5.0

292.4

283.4

9.0

634.2

547.2

87.0

1,159.5

1,117.5

42.0

297.7

291.7

6.0

428.9

416.9

12.0

696.4

647.4

49.0

1,319.7

1,297.7

22.0

373.6

370.6

3.0

614.1

614.1

0

770.8

721.8

49.0

1,492.5

1,470.5

22.0

415.4

412.4

3.0

836.9

836.9

0

851.2

802.2

49.0

1,602.7

1,580.7

22.0

451.6

448.6

3.0

997.4

997.4

0

926.9

877.9

49.0

1,712.7

1,690.7

22.0

480.0

477.0

3.0

1,090.5

1,090.5

0

1,024.7

975.7

49.0

1,841.3

1,819.3

22.0

511.9

508.9

3.0

1,223.9

1,223.9

0

> c:

( Q

CZ to

CD

CO

CO oi

CO

m

i :

CO a> o tV3

@

-~L

0 3

CO

CD

O

CO

»—t-

CO

CO

CO

Table 4-4

Forecast Growth Rates of Captive and Merchant Silicon* by Region(Percentage of MSI)

Worldwide Total Silicon*

Mercliant Silicon

Captive Silicon

1992

0.3

0

2.9

1993

16.4

17.3

4.2

1994

17.1

17.1

17.6

1995

19.2

23.4

-49.7

1996

17.0

17.5

0

1997

11.0

11.3

0

1998

7.9

8.0

0

Americas Total Silicon*

Merchant Silicon

Captive Silicon

Japan Total Silicon*

Merchant Silicon

Captive Silicon

Europe Total Silicorl*

Merchant Silicon

Captive Silicon

Asia/Pacific Total Silicon*

Merchant Silicon

Captive Silicon

NM = Not meaningful

NA = Not applicable

'Includes prime, test, and monitor wafers

Source: Dataquest (July 1996)

0.3

0

2.9

-5.2

-5.1

-7.5

11.6

11.9

0

25.9

22.6

NM

7.2

20.0

20.5

0

28.0

27.4

50.0

8.0

2.8

18.0

18.8

0

13.1

12.4

17.6

10.9

10.8

13.5

17.5

17.4

20.0

46.7

47.1

33.3

9.8

18.3

-43.7

13.8

16.1

-47.6

25.5

27.0

-50.0

43.2

47.3

-100.0

10.7

11.5

0

13.1

13.3

0

11.2

11.3

0

36.3

36.3

NA

10.4

11.1

0

19.2

19.2

NA

0

7.4

7.5

0

8.7

8.8

8.9

9.4

0

6.9

7.0

0

6.3

6.3

0

9.3

9.3

NA

> c

<Q c= w

05

CO

CO

CD

CO to

05

O ro

@

^-k

CO

CO

CJ)

o

Table 4-5

Forecast of Merchant Prime and Test/Monitor Wafers by Region

(Millions of Square Inches)

Worldwide Merchant Silicon

Growth Rate (%)

Prime

Test and Monitor

1992

1,728.2

1.0

1,382.5

345.7

1993

2,026.7

17.3

1,595.1

431.6

1994

2,373.3

17.1

1,841.6

531.7

1995

2,929.8

23.4

2,228.8

701.0

1996

3,441.6

17.5

2,517.1

924.5

1997

3,828.9

11.3

2,745.2

1,083.7

1998

4,136.1

8.0

2,966.1

1,170.0

1999

4,527.8

9.5

3,287.0

1,240.8

Americas Merchant Sili£0ii

Growth Rate {%)

Prime

Test and Monitor

Japan Merchant Silicon

Growth Rate {%)

Prime

Test and Monitor

Europe Merchant Silicon

Growth Rate (%)

Prime

Test and Monitor

Asia/Pacific Merchant SiSpMrii

Growth Rate (7o)

Prime

Test and Monitor

Source: Dataquest (July 1996)

450.8

3.3

360.6

90.2

848.8

-5.8

679.0

169.8

206.1

1.6

164.9

41.2

222.5

28.1

178.0

44.5

486.7

8.0

380.0

106.7

1,008.2

18.8

792.1

216.1

248.4

20.5

200.2

48.2

283.4

27.4

222.8

60.6

547.2

12.4

417.5

129.7

1,117.5

10.8

871.5

246.0

291.7

17.4

223.6

68.1

416.9

47.1

329.0

87.9

370.6

27.0

283.4

87.2

614.1

47.3

470.6

143.5

647.4

18.3

464.8

182.6

1,297.7

16.1

1,010.0

287.7

721.8

11.5

484.8

237.0

1,470.5

13.3

1,100.4

370.1

412.4

11.3

288.0

124.4

836.9

36.3

643.9

193.0

802.2

11.1

514.9

287.3

1,580.7

7.5

1,174.2

406.5

448.6

8.8

306.3

142.3

997.4

19.2

749.8

247.6

877.9

9.4

572.9

305.0

1,690.7

7.0

1,256.8

433.9

477.0

6.3

320.0

157.0

1,090.5

9.3

816.4

274.1

975.7

11.1

647.8

327.9

1,819.3

7.6

1,365.3

454.0

508.9

6.7

341.1

167.8

1,223.9

12.2

932.8

291.1

Silicon Wafer Forecast

I r H

O

O ev| o

s

P 4

* o

^

O

n

LO r H

ON

T ^

o

?8

^

r H

^

o

s

r H

-*

CN

OS

NO

CO r H

o

• * o o

N£>

CO

o

r H

i n

NO

CO

o

NO

• *

0 0

d

O

d o

^ _ l

^ . so

CN

• ^

r-<

o

NC

o

C v

^d

NO

r^

r H

o

NO

CO

0 0

• *

• *

• ^

d)

O

d o

O

i n

CN

d o

NO

CO

CJN

CJ^

CNj

CO

CO

CO

NO

ON

i n

S

I N

d

CO

<5^

CN

CN|

o

CN

NO

o

I N

CO

i n d

CO

oo

CO

CO

o ro

0 0

0 0

0 0

i n

• *

d

CN

d

r H

CN

• *

r H

NO

o

CO

'^

ON

o^

CN r H

CO

CO

I N r H

I N

CM

d i n

CN

d

• ^

ON

0 0 r H

i n o

NO

OS

0\

^•^

o \o o

NO

K

CM

CO r H

I N

I N

CO

0 0

d

• *

o

O

d o

•* o^.

"*

0 0

0 0 t N

O N r H

o

I N

o

NO

O )

ON

o o

"*

0 0

iri

CO

o

O

d o

'"' i n

• < #

0 0

d

T—(

I

I

t:

R

H->

cn

>H

01

MH

«

«

ON r H

i n

O N

O N

T-(

o

ON

o

( N

O

ir^

l O

I N

T—f

I N

CN

-*

NO

0 0

CN|

o

T—(

o

r-< r-H

1—1

CO r H

CN r H

CN

^ i n

• *

'^

O N r H

o o d o

r-M

CN

O

'"i^

•*

•<*<

NO r H

o

8

^-»

• *

CN

CO

< N

^r~*

CN

>*

O N

O N

T H

I—1

o

"^

r H

0 0

• ^

r~*

I N

o

Cv

• *

CSI

(N

r H

o

O

8 r H

0 0

o

ON

0 0 r H

Xi

G

O

^ . ^

4-> A

o^

QJ U

.a «

on (u

>-i b

0) (Q

^ O^

S ^

U 01

•rt bO

^ • 1 iS

QJ " O 4 )

-S; T! w

-^ o S3

o o r<

tr>

ON

O i

r-t

Z i

•H

is

• - I

D

4-1

•a

; j

•w

CO

« u

01

TJ

o

MH

o

. 5

^.

"frt

01

«

i

t n

O N

o>

I H

N

O i

O N f ^

« ^

2 c

; H

a;

6

( 0

s

a; x: u

l - l

«

r H

o

^

r H

O

NO

r-l

0 0

0 0

CN

o

I N

• ^

• *

NO

o

CNJ

o

r^

CN t N

0 0 r H

o

CO

CO

CO

CO

• *

I N

CN) rJH

c6

I^^

O

!< f^

7~-*

CN

CN

p

ON

r-l

0 0

CO

I N

CN

I N

NO

0 0

• * !

NO

i n

a^

o

r^

O

d o

""^

^ O

CN

CN

\6

r H

o d o

C N

CJN

P

CNT i n

CN

)H

01

•4-*

01

D

C D

a> ss

^ - 1

C

0> tn

<u

" o

C

1 — • f M

QJ

0^

cn

Oj

" u

C

•—=

c<^

O r-"

1

( N

I T l i

m

r H

1

E

CJ

CN

o o

CO

T 1

H-i

E2

55

s

1

1 rtS

1

x;

s

©1996Dataquest

CNI

d i n

• ^

I N

d

CO

CO r H

oc5

NO

i n

r H

-* o

• *

ON

I N r H

( N

ON

i n

CO

d

I N

-*

CN

CN

CO

i n i n

CO

CO

\ D

NO

• *

CO

CO

o

^

CN!

I N r H

O S

CN uri

• *

d

CM

iri

• *

0 0

I N

CO

O r H

•<t

CN

o

ON

i n

NO r H r H

NO

i n

ON

d i n iri

ON

t <

CO

CO

ON

CO r H

0 0

i n o o

C N

i n m

C N

CO

i n

ON

d

NO

i n

CO

i n

CO

I N

I N

CO

ON

ON

•5j<

CO

I N

o

NO

NO

CO r H

r~*

CN

uri p

r H

NO

i n

CN

CM

CO

o

I N

CO

o

CN

• ^

CM

CO

o p

r^

CM r H

I N

o i n p

r-^

CO

NO

CO

CN!

CO

CO

NO

CO

CN

CO

CO

CNJ r H

^

r^

CO

I N

p

I N

I N

T—f

CN r H

CM

o

ON

r-l

0 0

CO

I N

CNl

I N

NO

ocj

^

NO

i n

ON

o

^^

CO

d

ir^

r ^

CN

CjN

"*

' t o

01

JS

CJ

tn

u a an,

M

«

^ ^

^M

• ^ ^ i i c

V] O

cn

a i

S 5

'—'

• E ^

C ^-'

D

«J

^

^

• " H

CO

O r H

1

i n

CN

T-H

i

o i n

r H

s e

o

1

m

o

E-H

p ^ cn

0)

HH

01

e

n j

5

u

OJ

V U H

NS

K

ot

M l

( 0

>

<

CO o>

3-

1

August 26,1996

43

44

Semiconductor Equipment, Manufacturing, and Materials Worldwide

T H o

O

O

2

«s a^

W\ r H

0 0 ffs t s

o ro d

CN

0 0 t ^

0 0 i n

T - H

CO i n

si

'^

0 0

I - H

o d o

r H

^

i n t s ^ r H

NO

d

T - H

o

CO

d

T - H

as

I S

o<

i n

d

CO

CN

ON

• *

CN

T—i

O

d o

T - H

• ^

0 0 i n

'"'

T - H

CO

T - H

o

i r >

d

( N

d

r H

0 0

d

T - H vo

ON

( N

< N

0 0

>*

t v

d o d o

r H r H

O

^ v r H

"*

T ^

T ^

o

VD

d

CN r H r H i n

CN r H

0 0

d

CO

0 0

• ^

' J *

^ d o d o

r H t s i n

CN

^

M ;

CJN

o

I S

d

\D

CN

T - H

CO

• *

T-H o

CN

CO

•^ d

• *

o o d o

T - H

ON

•*

T - H

0 0

ON

o

I S

d

0 0 r H r H

p

0 0

v~i

d

( N

0 0

I S r H

CO

d

0 0

0 0 i n

N O

T-H

NO

o

i s

d

0 0

T - H

r^

r H

0 0

NO

I S

T - H

o

NO

T - H

CN

d

• *

ON

o

NO

o p

r - i

I S

T - < r H

o

0 0

T - H i n

T - H

ON

CO

T - H

I - H

d

0 0

ON

• *

0 0

O N i n

o

T - H

,^

NO

^

r H

CO

0 0

T - H

• ^

T - H

NO

^

r-f

T - H

o d

NO

NO

• *

NO

0 0 i n

o

T - H

^

ON NO

0 0

• *

CO

T - H i n

O N

o

NO r h

Tt<

CN t s i n

«

(/I

O)

M N

«

c o

•V4

«4-»

1-^

s -2

VO ff>

OS

r H

LO

T - l

'*

0 ^

as

r H

en

9 ^ a s r H t s

9\

r H

o

O N

d

• ^

'^

T - ^

CN

sd

r H

CN

NO

CO

CO

cs

CO

o

O

d o

^^

NO

P

T - H

p

r-<

T - H r H

d p

T—H t ^ t s

T - H

CN

ON

T - H

T - H

d

>*

CJN

T - H

CN

o o d o

T-H

CO

ON

0 0

CO

T - H

T - H

d

r - H r H

I T )

d

CN

CO r j

cs

0 0 r>4

• ^ f s l

CO

T - H

o o d o

T-H

O N

CN

0 0

NO

iri

T - H

d

CSj

T - H

O g:i i n

\£5

CN

c^

T - H i n

0 0

o

l - H

d

CO r H

CO v D

CN

ON

r—*

CO i n

NO

CO

ON

CO t

O

s

T-H

^^

I S

T - H

d

T-H

o d o

T-H

^_

O i n

NO

d

T - H

o

CO

T - H

• t

H r H

ON

0 0

0 0

CO

T - H

ON

NO

o

•*

CO

• *

• * i n i n

CO

d

CO

T - H

I S

CO r H i n

ON

0 0

CO r H

CN

•5)5

o

ON

( N

>*

O N

CN

iri

CO

d

CO r H

O

"*

T - H

c^

ON

o

CO

T - H

CN

CS

o

• *

d

• *

T - H

T - H i n

CN

d

CN

T - H

O

C O

^ H

o d

ON

d

T-H

CO

T - H

o

NO

NO

CO

o o

i n

CN

d

CN

^^

r H

• = ! < r H

ON

d

r H

I S

0 0 i n

d

,

NO i n

CO

CO

0 0

• ^

Qj "n

( 8 ^

*H k 2

• ^

I—I

«

t s

O

c<

t ^ r H

CN

T - H

CN

O

o<

T - H

0 0

CO

I S

CN

I s

NO

0 0

• *

NO i n

ON

o

I - H

s5

O oi

S

^ . 2

•BO

I"' (» a> 0)

.a-S

^ e

; - ( I — I

a» Q.

MH g;

« «

^ g^

bO

(Q

t&i

• M

C3

0) u h(

u

01

•i-t

01

6

(0

s

J 3

x;

u

V rfl^

u

e

•^

D

i

K

3

cr bo

« c

0) in

Hi j :

u

C fc;

0>

c-

en

CU

J S

6

6 o

r H

s

6

i n

CN r-"

<o

t n

( O

o

CN

o o

CO

SEMM-WW-MT-9602

ra

o

!

fD

o

H

j : :

4 - J

o

i H

©1996Dataquest

• *

T-H

CO

I S

o c<

t^.

T - H

CN r H

<N

O

ON

1 - H

0 0

CO

c^

<N t ^

NO

0 0

• *

NO i n

ON

Ul

•1-1

IH

«

H-J

CD

S H

V4H f «

^ ^

^ 2

C ( 0

• s ^

i <-l-l

^ o

CO

•£ S

l U

J S u

.2

O

1—H

•si

I X

D

V

T J

C

g o o

i n t s r H

o

i n

^-v

<n

OJ

x;

t j

_c

l-l

S

»*-^

"S

6

(C

S

I-*

t e

o

( N

6

B

a

CO

1

o

H

(C

) H

<u

>

<

CO

3

D -

CO

Ts

O

0)

2

O

CO

August 26,1996

I

2096-il/\|-MM-|/\ll/\l3S

966l.'92isn6nv

lS8nbBiBa966L® o

3

CD

^

a>

J3

1

( D

i

1

<

ftl

H

00

t

a>

3^

f-i

5

S'

3

S"

1-1

>—V

3*

^•^^

K-^

o'

?

Si

i

B-

CJO

CD

H

1

t o o o

3

3

^

m

C 3

3

3

H

^

t o

U l

^

o

H

1

Lj

n

rr re a i w

G

to

3"

§ ! •

S C

5' ^'

3 ET

Cfi — -

»

S.I

Si'

1-1 N<;

m* O"

^ ^

0/2

i-p

W

1-1

uT

CO o

NO

CJI

ON

i ; ^

OO

ON

^

t o

^

O J

0 0

1—1

NO o t o

1—1

t o

H-1

V I

V I b

V I w

H->

*-

U 1 o

S

4 ^

0 0

ON

1

o w

I—i

O N

V I

0 0 bo

I - '

o

ON to w o on

H ^

o\

CJI

* •

k) o

>—>

^

JNJ o bo h ^

NO t o

1—1

o

V ] t o

^

o

O l k)

ON

CJI c» bo o ro j ^ t o w io

h—1

NO bo

>-^

NO t o

CJI o

U l w

^

OS

CJI

CJI o t^

h - »

t o

ON

NO t o o a^

^ - J

^ J

ON

NJ w o on

U l

U 1

ON

NO

N> o

^

^

t o

0 0

ON t o o t o

1 - i

o

NO

N3

H - i

o

i-t

o

- J .

1

ON

NO

1—1

O N

ON

"H-1 w hpl.

1—I

s

b o

1—1

t o

V I

1—1

ON

•—1 h J

t o

NO

g

ST

H o

I - l -

Ul

NO v j

O J

1—1

V ]

0 0

1—1

ON

U l

k

tn

1—1

o b

1—1

o o b

1—1

4 i .

0 0 w

h-i

o o b o o b w o

Q

[ 3

1

o o o

*3 tNJ o o

. - 1

U l

H

1

U l

3 t o

_ 1

1—1

O

NO

U l

ON

1 ^

oo

ON

V I t o

V I

OJ

G O

^ J

U l

1 ^

ON

NO

O J

O N

V I

>^

ON

CJI o t o

O J t o t o

1—1

CJJ

OJ

4^

NO

V 3 t o

ON

>(>•

t o

1—1

ON

NO b

>l^

NO bo

N> vO i^

1-1

C 3 o

3

h—i

t o

•—1

VI

1 - ^

O J t o

1—1

>—1

U l

1—1

o

4^

NO

U l

V I b

V I

1—1

V ]

H ^

U l w

^^

(C sn t o

t [

n

•-a fD r5 fD

3

i-h

a

&)

3 rt

<n

O J

• - 1

4 ^

O

O

5S*»

•-^ ^ J_ re

zn

X I c

»

I-t

3

3" en

C g

S'

i

3

S ffl

M l A

|N9

)-*

O l

OQ HtS

ro n

^ T n

HQ SG

C N

u ro

1 ^ i _ j

n O

C^ as

S ST

N ^ l ^ *

^ ^

, f »

(T S*

^ §

25^

5 s

2 S.

&> M

13 VO

S^

E^t^

3 O

a

M »

ST

N^* f *

N^*

1—1

4^

O

h—4

i—1

O

H I

NO

NO t U

M

VO cn o a*

Ms ro

I-t

cn

3. en l O

V I

NO o

NO

O ON

V I

^ J

NO b t o t o t o

NO

i i

U l

ON

^

v q

1—1

w o

NO

^

CJO o hC^

) — 1

NO b

1—1

o io t o b o

V ]

V I

1—1

0 0 o

1—1

o o b o t o

ON t o

1 ^

ON

H - l

t o o b

ON

NO o bo o

M

NO

VO

•<l pi

V J

ON

;^

bo o

K - l

H J

^

CO l o

CJI

h — 1

0 0 bo

NO

NO

H ^

NO o

0 0

h—l

1—1

NO

NO

1—1

8 b o t o

NO

H-1

J ^

U l

V I

1—4

0 0

O J

ON t o p

V I o

I-t

NO

0 0

CJI

GO

CJI

>•p

I—1

)—L

Oi

NO

C>J

1 ^

>^ h J

0 0

U l

NO

V ]

h J

0 0 o

0 0 w

^to

" h - l

1—1

( O

1—1

o o b o

U l

(Ji

t o b rfa.

4 ^

ON

H ^

ON

V I

CJI

ON o bs o

M

NO

VO

CJI

\D

W

0 0

CJI

ON o ro

1—1

ON

I—1 t o

0 0

NO

1—1

0 0

NO

NO

V I

1—1

V I o

1—1

1—1

NO t o

O J o\

1—1

o b

l—i

b

C*J

C>J

INJ

1 ^

U l i-1

h ^

CJI io

CJI b o

U l o h J o a

O l vO

NO

NO

( O w o w

CJI

t ^

CJI o

1—1

0 0

>^

NO

ON

1—1

U l o

NO

NO t o

U l

NO

0 0

1—1

o o b

1—1

4 ^

C>J

NJ

0 0

»t^

V I l(i>

1—1

C>J

U l

4^

U l o

4 ^ o

I O o o

M

i

1SB08J0J J9iEM UOOIIIS

9V

Semiconductor Equipment, Manufacturing, and Materials Worldwide

46

O T-i (N •<S<

OS • ^ O O 0 0 vO

O 'O vO > s ON

• • 8 « ^

O I—I CN 0 0 • ^ i-H O t ^ I-H d -^ CN oc> d i n ' ^

CN s o

O 00 O O O VO tN

O tN t^ \ 0 OS i n

CM i n

O ^ l O s O c O L n o O O s s O

d f s ' c ^ i t N t v : o ^ o o

CN i n o ^

O ^^ ^H \ 0 sO 0 0

d -^^ cs) so !>;

rH O

(N vO

00

0^

0^

O CN CS 00 s o CN O O t^ 00 d O^ CTs O O

CO i n

8 ° ^'

O CN O CO OS s o

o o

d i-I • • •

CO CO

CO « * O

O (N SO 00 CO o OS g o

Tt CO SD s o o --; tN s o

O CN rH CO CN OS d iri CO IN ' ^

O t x 00 d «?

CN i n

a\

0^

-<i< OS o d CO

o

CO

o

OS i n

0 0

CO

OS

so

C^

CO

o in CN o o in •^ so in

o i n

CN

CO CO i n

o

CN i n

VJJ

iri

'A

IH

T-H o

•<a< o

i n t^ i n

Tf

0 0

so

ON

T—1 o

X I

•c

tn

X)

tfl

0

a> w

V

ci

N N M

0)

u

0)

<4H

«

^

OS fl rt

g.

CA)

V bO rt

^ a»

V

PH

^ o

c; a

V

H w

b

^«^^

V4

O)

V M

(S

s

cn

o

fs

D

4->

en

ffs

• F *

Si rH p

13

4 - en n u

r.

Ix

^

;»^

x>

C

o

4i>>

;3

-9

c o

a

• H rt

1ft

a\

t l

OS

OS r^

a\

ON

cs a\

ON

S d

I T

Cfi

^-»

a>

6

<C

Q

U c rH s o O ^^ OO 0 0 O O 00 ^ d d d 00 d OS i n

OS f s | r H TJ* , - (

i n c - x O o o o - ^ c N

d d

CN CN Tjt

CO

C N t v O i n c O C O O O O C N

d r - I c N o s c N - * d S i c o

CS4 Csl • * O ' ^ (N

CO i n OS i n c ^ T-H

O (N

s

O s o - ^

CO CO

•^ t ^ t ^ CN 00 t^ s o

1-H o ^ o CO so in

CO tN csi OS t>! 00 OS

^ r-i CM rr O o i n M

s ^ ^

I

CN

O

0 0

o o

i n

0 0

CO

r-l

CO

(N

o

rH

CO tN

C N

r—i

CSJ

I—1

CN

O

OS

T—1

00

CO

CN

C N

so

0 0 s o i n

CJs

o

fsl

r^

CN

• *

iM i n i - H - ^ p C S s O p O O - * d d s o ' i ' i n i - I i N " ~ :

T-i i n

CN t ^ CN i n i n CO o d d in -^ •* d

CO O i n

- OS

^

1

(N

2

IN

• *

en

<u

J :

u

C

Z)

a Q

3 _ en

%M

c o

•s

<2i

J 3

« S) o -g

B

B

o

ffl

5 m o o o

o

CN i n o o

CN CO i - i 1-1

C-) CO

2 « s2 ^

sJS

'P

60

(8

Uc

<U

>

<

•e

I

SEMM-WW-MT-9602 August 26,1996

©1996Dataquest

O <-- • * 0 0 - ^ CN d • ^ CN Cv OS o OS i n

CO ^.

CN s o

Silicon Water Forecast

rH

0

0

CM

0 i n

d

0

c^

•^

• *

SO

CO

OS

iri

i n s o

d p r H

IN rH

r H iri r H

I

0

0

0

P ^

o\

9i

T'

rH

OD

C^

a^

rH

t^

0^ rH

0

so d so en

T-H

tfi

CO

TJH

CO

0

so

i n

• * d

0

d

0 rH

0 0

r H

i n

t^

0 0

r H

0 CO

d

tN

-*

OS

iri

CO

CO

CO

IN i n i n

r H d

0

d

0

r H

0 0 tN

CN

r H

r r

CO

r H t—1

0

0 0

d

0 iri

0

t^

t o

iri

CO

^

r H

i n

0 0

d

0

T^

^_,

CO

r H r H

Os

Os

I—1 c5

OS

d

OS

ir>

10

06

OS

d

• ^

IN

CO

• *

0 0

d

0

r H

a

CN

0 0

06

T—i

en

t n

<w

CD

U

<« n

A

G

O

ve

0^ a^ r i

Ifl

o>

OV

1-1

• < *

OS

OS r H

1—1

0 tN

0

• *

0

CN

T — I

SO

r H

•^

K

00

d r H

IN

OS

CM

CN

0

CJ

0

0 0

Cs!

so

^

r H iri

^ tN

OS

• *

Os

CO

CO

• *

OS

(N

0

0

Os

^>;

T-t

0

0

d

0

r H

0

d

0 rH

0

8

r H so

SO

0 0

0

• < *

VO

• *

^

r H

CO

'^

t^

>*

CO

SO

• ^

^

9 o -b

M.52

2; D

O N 11-1

rH C

*>D cn

OS

OS

1-H

N

OS

OS iH

SO

0

0 0

0

OS

CN

SO i n

0

m

00

d

CM

CO

CN i n

IN

m

r H

CTt

CN

r H

• *

i n

•<# d r H

0 t v rsj

1

0

d

0

r H

0

d

0

r H

i n

S

CO

06

CN

0 0

CO

Csl

IN

d

CO

O »i

(0 ^

2 C

^

I—1

CO

IN

0

K

IN

r H

fsi

T H

CM

P

OS rH

0 0

CO

IN

Csl

IN

SO

06

Tf s o i n

OS

0

r^

o*

^

o &

^ • 1 t o

^s

•c >.

.H c

CO ' • '

«- Si

<£ «

> u*

i-<

01 a->

IV

6

(5

(3

>^

- O

CO

01

JZ

u

01

t-i cc l l

T'S

s

u

•M

I ll^

s

C

Of

u

1 ^

01

(U

&0

n j

(^1 tn

OJ

u

B

^^ cs

cc

a;

JZ

u

C

• • H

m

CJ

i~l

0

1

tn

tN«

r H

G c i n

r H

^

1

C

0 tN

1

0

CO

1

^

«

Sn

S^

0

«

H

x:

- 1 - ^

s

0 <N

r H

CO

• ^

0

• ^

IN

?i

r H d

Csl

0

r H d

i n

r-i

i n

r H

i n

so

0 CO

r-i

i n

• ^

rH

• *

0

OS

r H

m

IN

SO

0

d

• * so

-* i n

• ^ so

0

• * r H

• *

• *

0

• ^

i n

iri r H so

• ^

r-^

r H

0

d

0

d

"*

0 0

CO

so

CN

d

CO

r H

s o

• ^

CN

• ^

tN

• *

r—l

0

csi

r-^

0 0

IN

CO

-*

CN

sd

CO

d

CO

r H

0

iri

s o

• ^

-rf

iri

rH

CN

Os

0 0 0

iri

CO i n

0

so

• < * ' d

• * r H

CO

iri

OS

>*

IN

r H

0

SO

0 0 0

CN

CO

0

0 0

iri

• ^ d

"^

v^

T ^ iri

IN

nJ

i n

d

r^

Os

CO

0 rH

SO

CN

CTs i n

iri

s o

d

• * r H

• *

• ^

CM

Tji

^^

06

SO

r^

0 CO

d

tN

Os

CN

iri

s o

d

CO r-i

CO

CO

CO

CO

0 0

iri

tN

d

0 OS 0

r^

iri so d

OS

r H

CN

CN

Os

CN tN

• ^

r-t

d

1

OS tN

r H

i n

0 0

• ^

'^

r H

CO

CN

0

IN

IN

r H

CN r~< rsi

0

OS

r H

0 0

CO cv! r j

IN

SO

00

• *

s o i n

Os

0

r H

en

f-i

jB cn

01

S'^

••E i

-^1

. — « j i : c

5 a

• ^ s eft

<D

x:

u

C

D tn

01

C

C

T ™

0

1 i n

!N

r-i

C

c c^

i n

r H

1

0

CI

1

0

0 t o

CJ

i~t

01

H»*

01

m

s S

1

6 tc s

O)

^J-p ra n;

CO

a>

a>

3

- 5 iS

0

tD ^

1^

I

47

August 26,1996

SEMI\/l-WW-MT-9602 ©1996Dataquest

48 Semiconductor Equipment, Manufacturing, and Materials Worldwide rent forecast for the year 2000 represents a level now about equal to what the wafer manufacturing industry has committed to supply to date, and our forecast in general now is more purely a demand forecast. Although w e are stopping short of calling a shortage of 200mm wafers to the point of restricting ramp-up plans, buyers of wafers will experience firm-torising prices again by 1999 and may be placed on allocation from time to time.

In a recent Dataquest Perspective ("A Fresh Look at 200mm Wafer Supply and Demand: Still Tight," SEMM-WW-DP-9603, July 22,1996) w e have detailed the updated analysis for demand and supply of 200mm wafers.

Polysilicon to Remain Short from Now through 1997 and Tight in 1998

The industry has been consuming more polysilicon than it has produced since 1994, and we expect that to continue through 1997, depleting industry reserves later this year. However, several new plant commitments have been made, and polysilicon production will again exceed supply starting in 1998. The polysilicon market is expected to remain tight through 1998, but the threat of shortage after 1997 seems to have disappeared through our forecast horizon. More details are included in a recent Dataquest Perspective ("Polysilicon Supply/Demand Update,"

SEMM-WW-DP-9604, July 22,1996).

What about 300mm Wafers?

N o w that the wafer size has been settled and the time horizon for the first

300m plant(s) has been proposed, we initiated a forecast for consumption of 300mm wafers. The level is still very low before the year 2000; however, the recent goal of a fab to come on line by 1998 or 1999 means that 300mm wafers will be made and significant activity will be occurring in R&D.

Although no company has yet stated a firm commitment for a fab, there has been significant activity on the development front. Most of the activity recently has been centered in Japan, and NEC and Texas Instruments have recently been noticeably more aggressive. A new consortium of Japanese companies, Selete (Semiconductor Leading Edge Technologies Inc.), has been formed there to spearhead the 300mm technology development.

Motorola and Samsung continue to be among the companies mentioned in conjunction with a possible 300mm fab, as well.

Silicon wafer manufacturers have noticeably ramped commitment, as at least five different companies have aimounced or have committed to pilot lines for 300mm wafer production. The earliest of these starts production in late 1997, with the latest coming on line during 1999.

Although no semiconductor company has yet to commit to purchasing equipment for a formally announced fab, we would expect at least two pilot fabs trying to come on line by 1999. We should hear announcements during 1996 (in fact, one may be very close in Japan), and we would expect initial equipment to ship by the end of 1997.

We believe the first commercially productive plant will be started around

2001 to 2003 (after the feasibility noted above), with serious volume rampu p in the years 2003 to 2005. This would be consistent with shrink

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Silicon Wafer Forecast 49

0.25-micron technology being primarily produced on 300mm wafers. This means that 200mm wafers represent at least a two-technology-generation wafer size, and fabs being built today m.ay have longer lives than history would indicate.

Further details a n d issues regarding the move toward 300mm wafers will be included in a Dataquest Perspective document to be published in the next couple of months.

Epitaxial Wafer Trends: Are Tliere DRAMs in tlie Future?

Sales of merchant epitaxial wafers by the wafer suppliers accounted for

520 million square inches (MSI) in 1995. About 65 percent of these wafers were used for CMOS logic applications, while only 2 percent were used for DRAM products. The remainder were shipped into the p o w e r / d i s crete device segments. By the year 2000, Dataquest expects that fully

16 percent of merchant epitaxial silicon will be used for DRAMs, primarily driven by two factors. The CMOS logic application remains dominant at

54 percent.

First, trench capacitor designs benefit from the use of epitaxial silicon. It is believed that b o t h IBM a n d Siemens are currently using epitaxial silicon, while Toshiba has been using hi-wafers (wafers that include a hydrogen bake surface treatment to reduce surface contamination) to improve the performance of its trench capacitor design. We believe that for lategeneration shrink versions of 16Mb and for 64Mb DRAMs from companies that employ the trench design (which include the companies noted, plus Texas Instruments and its affiliated fabs and perhaps including

Hitachi), epitaxial silicon will have to be used.

Second, even in the stacked capacitor design, some companies (primarily in Japan) are finding that epitaxial silicon can be used to enhance yield by eliminating the very localized silicon "pit" defect that has recently become important at the 0.25-micron level technology in larger wafers (200mm and 300mm). Fujitsu is believed to be seriously evaluating and nearly committed to the use of epitaxial silicon, and several other Japanese companies, including Matsushita and Oki, also have epitaxial silicon under consideration for DRAMs.

Dataquest will w at ch this area of device construction technology very closely over the next several years, as there is a distinct possibility that this need for epitaxial silicon will be a single-generation need. At the

0.18-micron generation, the use of other capacitor dielectrics and electrode materials may enable companies to return to planar or simple stack structures, perhaps eliminating the driving need for the epitaxial layer.

SEI\/ll\/l-WW-l\/IT-9602 ©1996 Dataquest August 26,1996

50 Semiconductor Equipment, Manufacturing, and Materiais Worldwide

Highlights of the Americas Region Silicon Wafer Market and Forecast

Silicon consumption in Americas grew 14 percent in 1995 to 943 million square inches, and we expect a milder, 11 percent growth in 1996. Demand for microprocessor and other logic chips has been and will continue to be the key driver behind increased silicon demand in Americas, and epitaxial wafer demand will be focused on CMOS logic.

Merchant epitaxial wafer consumption increased 27 percent in 1995 to

247 MSI, driven in large part by microprocessor manufacturers, such as

Intel and AMD, which build their microprocessors on epitaxial wafers.

Strength in the automotive and discrete segments of the chip market have also been much stronger than expected. By 2000, epitaxial silicon will account for 27 percent of the square inches consumed in the Americas region—the highest concentration in any region.

Dataquest's longer-term forecast for Americas silicon consumption remains firm, primarily because w e believe that the United States will attract a larger share of foreign multinational fabs, such as those recently announced by Hyundai, TSMC, Samsung, and others to be announced.

We are projecting that total silicon MSI will grow at a 10.6 percent CAGR for 1995 through 2001.

Highlights of the Japanese Silicon Wafer Market and Forecast

Our Japan silicon consumption forecast has slightly increased from our last update, with the silicon market growing 16 percent to 1,483 MSI in

1995, with continued moderate growth in 1996 and softening in 1997 and

1998 as the industry migrates to 16Mb DRAMs.

Unlike Americas, with its sizable CMOS epitaxial wafer market, Japan's merchant epitaxial wafer market is more focused on discrete and bipolar applications today. Therefore, a recovery in the economy will have more of an effect on the growth of Japan's epitaxial wafer market, and early indications are that it has kicked into gear. Epitaxial demand grew 38 percent in 1995 after a 33 percent increase in 1994. Another growth year is expected in Japan in 1996, although it will be slower as companies use this time of excess capacity to begin to pull epitaxial deposition back inside.

Dataquest remains moderately coriservative with its longer-term growth scenario for silicon wafer demand in Japan as the country continues to work through the evolution of its semiconductor production infrastructure (seen as being too heavily dependent on the DRAM). Investment patterns during 1995 suggested that Japanese semiconductor manufacturers are willing to come to the table and invest, preserving their stake in the memory business against the Korean companies. The desired shift of the

Japanese product mix to higher-value-added semiconductors had apparently been put aside until the current memory cycle subsided but has again come to the forefront. We are estimating that silicon demand will grow at a 9.8 percent CAGR for 1995 through 2001, the slowest growth of all regions.

Epitaxial silicon deposition will experience strong growth later in the decade as companies begin ramping the use of epitaxial layers into the

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

Silicon Wafer Forecast 51

^

I

DRAM. As a result, we are calling for a CAGR of 17.8 percent for epitaxial wafer consumption in the coming six years, above the worldwide average of 15.1 percent.

Highlights of the European Silicon Wafer Market and Forecast

Demand for silicon wafers in Europe, as well as wafer fabrication equipment, remains heavily dependent on the fab activities of foreign semiconductor firms. With increased presence from European companies, the outlook for silicon consumption has brightened.

The European market grew 29 percent in 1995 to 458 MSI. European silicon demand will moderate significantly in 1996 through 1998. Siemens

Dresden and other DRAM production, and U.S. multinationals Intel,

Motorola, and Texas Instruments will continue to r a m p to answer the demand in Europe, helping silicon consumption grow another 12 percent in 1996. This is lower growth than anticipated six months ago, however, as ramp schedules for fabs in Europe to support the telecommunications sector have been reduced recently. The longer-term picture for Europe remain moderately healthy, with a 10.4 percent CAGR through the year

2001.

Epitaxial wafer demand in the region increased dramatically in 1995 to

84 MSI, growing 50 percent. This has been a direct result of production ramps by Intel in Ireland, Philips, IBM in France, and Siemens in

Germany. In the near term, epitaxial wafer demand increases will come from predominantly European producers, primarily from the p o w e r / discrete segment. Later in the decade, the DRAM capacity will ramp and the migration to epitaxial silicon in DRAMs from multinationals with fabs in Europe will drive the epitaxial market to a CAGR over 4 percent above polished bare wafer consumption through the decade.

i

Highlights of the Asia/Pacific Silicon Wafer Market and Forecast

Silicon consumption grew at a 43 percent pace in 1995, the largest growth rate of any region in the world. Production is expanding at a fierce pace and not expected to ease through 1997, as many large 200mm fab projects are in various stages of construction and start-up. As the Asia/Pacific region collected nearly the largest proportion of 1995 capital, the trend for high silicon consumption growth will continue unabated with the exception of 1998, w h e n we expect that the region will be caught in a period after foundry projects have just finished ramping u p b u t before the next significant DRAM capacity expansion occurs.

The phenomenal growth in silicon consumption Asia/Pacific in 1995 was tied directly the manufacturing activities of the Korean DRAM producers.

Growth will be seen here, b u t increased production in Taiwan, Singapore,

Thailand, and, most recently, announced Malaysian fabs will also contribute to the growth. As these predominantly foundry fabs come on line in

1996 and beyond, they will provide some regional consumption stability as memory-related silicon consumption cools in 1997 a n d 1998. Taiwan, with its many new DRAM producers coming on line in 1996, will cause silicon consumption to grow in Asia/Pacific by nearly 35 percent again in

1996. Asia/Pacific remains the fastest-growing silicon consumer, with a five-year CAGR forecast of 18.2 percent.

SEI\/IM-WW-MT-9602 ©1996 Dataquest August 26,1996

52 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Epitaxial wafer consumption will grow at the fastest rate in the A s i a /

Pacific region, as well, with a CAGR of 22 percent. However, this comes from a low base of consumption that will remain low. Although the foundries will require a significant portion of their capacity to consume epitaxial wafers, the DRAM fabs in the region do not use epitaxial wafers.

With the exception of a few joint-venture fabs, the epitaxial consumption in Asia/Pacific DRAM capacity will remain quite low compared with other regions. Epitaxial wafer consumption will remain below 5 percent of total Asia/Pacific MSI by the year 2000 (compared with 16 percent worldwide).

Silicon Wafer Revenue Forecast

Dataquest has been tracking silicon wafer revenue and market share since

1985 but has always provided forecast information in terms of square inch area and unit wafer size distributions. With the armouncement of the initial public offering of MEMC Electronic Materials earlier last year, we have initiated an industry revenue forecast.

Table 4-11 contains the revenue forecast for silicon wafers worldwide. Our methodology takes the wafer size distribution forecast in Table 4-6 and multiplies this matrix by our view of wafer price trends and test wafer consumption trends (by wafer size) in the forecast horizon. Dataquest does not publish a specific wafer price forecast, but does make comments o n trend outlooks from time to time (see the Dataquest Perspective

"A Fresh Look at 200mm Wafer Supply and Demand: Still Tight,"

SEMM-WW-DP-9603, July 22,1996).

In our analysis, w e have concluded that the revenue forecast would resemble the semiconductor industry more closely than the capital spending markets. The concept of semiconductor revenue per square inch is m o r e closely tied to silicon consumption than raw wafer capacity of the industry, although this metric is cyclical. The six-year CAGR of 15.7 percent is about equal to the semiconductor forecast of 15.5 percent. This is consistent with the model that semiconductor manufacturers will attempt to control the costs associated with manufacturing, which includes using silicon more efficiently in the future. This would tend to keep the growth rate of silicon revenue slightiy below that of the semiconductor revenue.

Yet our forecast does not reflect this, primarily because over the next several years the industry will be migrating to 200mm wafers, which have a silicon wafer revenue per square inch significantly higher than the smaller w^afer sizes. Therefore, the mix shift will tend to increase the growth rates in revenue, bringing the forecast more in line with the semiconductor growth rates.

Table 4-11

W o r l d w i d e Merchant S i l i c o n Wafer R e v e n u e Forecast, 1993-2001

(Includes P o l i s h e d , V i r g i n Test, and Epitaxial Silicon) ( M i l l i o n s of U.S. Dollars)

1993

1994

Growth (%) 19.1 28.6

Source: Dataquest (July 1996)

1995

6,298

37.4

1996 1997 1998 1999 2000

13359

20.0

CAGR (%)

2001 1995-2001

15,120 15.7

13.2

SEMM-WW-MT-9602

©1996 Dataquest August 26,1996

Silicon Wafer Forecast 53

Dataquest Perspective

The silicon market, driven by a strong long-term picture for semiconductor unit demand in general will grow faster over the next six years than the last six years. As the industry transforms into a 200mm baseline, the outlook for silicon wafer manufacturers becomes brighter. Silicon manufacturers have answered the call for 200mm capacity with significantly increased capital outlays. We believe silicon manufacturers' ramp plans in

200mm have been strategically and smartly measured because the overcapacity situations of 1985 are being remembered, and we are not expecting that scenario to be repeated.

Activity in 300mm wafer development has accelerated, particularly in

Japan. Although no semiconductor company has yet to commit to purchasing equipment for a formally announced fab, we would expect at least two pilot fabs with the goal of coming on line by 1999.

By the year 2000, Dataquest expects that fully 16 percent of merchant epitaxial silicon will be used for DRAMs, up from about 2 percent in 1995. We believe that for late-generation shrink versions of 16Mb and for 64Mb

DRAMs from companies that employ the trench design, epitaxial silicon will be needed. Also, other companies may decide to use epitaxial layers to solve specific yield issues at 0.25 micron, although it may be only a single-generation solution.

The silicon has become recognized again as being strategic in the semiconductor manufacturing infrastructure. Will this continue? We believe it will, as long as silicon suppliers continue to concentrate on value-add processes and techniques, as the equipment manufacturers have done, and as long as they adequately and smartly plan capacity additions.

SEI\/ll\/l-WW-MT-9602 ©1996 Dataquest August 26,1996

^ Chapter 5

^ Semiconductor Consumption Forecast

This chapter presents data on the worldwide semiconductor market by region. The regional semiconductor market, or regional semiconductor consumption, deals with where chips are consumed; this contrasts with regional semiconductor production, which deals with where chips are manufactured. The data presented here is for the merchant market and do not include the value of chips made by captive semiconductor manufacturers for internal use.

This is an excerpt from the semiconductor five-year forecast published by Dataquest (Semiconductor Five-Year Forecast TrendsSpring 1996,

SEMM-WW-MT-9601, May 13,1996). Further details regarding this forecast can be found in that publication.

Yearly exchange rate variations can have a significant effect on the 1989 through 1996 data in the following tables. For more information about the exchange rates used and their effects, refer to Appendix B.

Semiconductor Consumption

Table 5-1 shows revenue and growth from semiconductor shipments for the years 1989 through 1995 broken down by region. Table 5-2 shows revenue and growth from semiconductor shipments for the years 1995 through 2001 broken down by region.

Table 5-1

Worldwide Semiconductor Consumption History by Region—Merchant Semiconductor

Companies Only (Millions of U.S. Dollars)

CAGR(%)

1989-1995

18.9 Americas

Percentage Growth

Japan

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (July 1996)

54,339

6.8

1989

17,070

7.7

21,491

3.5

9,498

11.9

6,280

9.2

1990

16,540

-3.1

20,257

-5.7

10,415

9.7

7,333

16.8

54,545

0.4

1991

16,990

2.7

22,496

11.1

11,014

5.8

9,194

25.4

59,694

9.4

1992

20,430

20.2

20,579

-8.5

12,218

10.9

12,034

30.9

65,261

9.3

1993

27,926

36.7

24,645

19.8

15,461

26.5

17,486

45.3

85,518

31.0

1994

35,939

28.7

31,010

25.8

20,819

34.7

22,812

30.5

110,580

29.3

1995

48,349

34.5

42,164

36.0

28,341

36.1

32,417

42.1

151,271

36.8

11.9

20.0

31.5

18.6

SEMM-WW-MT-9602 ©1996 Dataquest

55

56

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 5-2

Worldwide Semiconductor Consumption Forecast by Region—Merchant

Semiconductor Companies Only (Millions of U.S. Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Europe

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (July 1996)

1995

48349

34.5

42,164

36.0

28,341

36.1

32,417

42.1

151,271

36.8

1996

52,478

8.5

41,244

-2.2

31,479

11.1

37,411

15.4

162,612

7.5

1997

60,217

14.7

45,286

9.8

35,734

13.5

43,004

15.0

184,241

13.3

1998

70,352

16.8

51,144

12.9

41,079

15.0

51,258

19.2

213,833

16.1

1999

85,481

21.5

60,212

17.7

48,433

17.9

62,121

21.2

256,247

19.8

2000

104,579

22.3

71,693

19.1

56,828

17.3

76,736

23.5

309,836

20.9

2001

122,044

16.7

79,723

11.2

65,295

14.9

91,162

18.8

358,224

15.6

CAGR (%)

1995-2001

16.7

11.2

14.9

18.8

15.5

I

SEMM-WW-MT-9602

©1996 Dataquest August 26,1996

i

I

Chapter 6

Semiconductor Production Forecast

This chapter presents data on the w^orldwide semiconductor production by region. Semiconductor production is defined by the place where the wafers are fabricated, and regional semiconductor production includes all production in the region, including merchant and captive producers and all foreign producers. For instance, Americas region semiconductor production includes Digital Equipment Corporation and Delco fabs as well as

Japanese company and European company fabs in North America.

Yearly exchange rate variatioris can have a significant effect on the data up to 1996 in the following tables. For more information about the exchange rates used and their effects, refer to Appendix B.

The semiconductor industry has a global manufacturing infrastructure.

Production of semiconductors is constantly shifting among regions as new capital flows toward areas of relative lower capital cost and higher growth of consumption. Dataquest reviews some of the trends, and potential impacts, for the future.

Historical Semiconductor Production

Table 6-1 shows the historical semiconductor production for 1989 through

1995 by region. Dataquest follows a methodology that employs our fab database, estimating the memory, microcomponent, logic, and analog/ power/discrete production components separately and estimating net production among regions for foundry activity. This approach provides insight into production trends for front-end manufacturing.

Because of the reclassification of the MOS portion of IBM Microelectronics' business as merchant, the captive production figures changed dramatically in 1993. However, IBM's bipolar production, which is consumed internally, is still classified as captive by Dataquest.

Captive Semiconductor Production

Semiconductor production from captive manufacturers is estimated to be

$1.81 billion in 1995, down from just under $2 billion in 1994. IBM has restructured and entered the merchant semiconductor market as of 1993.

Dataquest has reclassified IBM's MOS sem^iconductor production to merchant, but the bipolar products (exclusively used internally) are still reported as captive. This part of IBM's business will be converted to MOS over the next three to five years, and this resulted in a lower figure for captive production in 1995 and future years.

Many captive producers may consider the move to merchant to take better advantage of the worldwide growth of semiconductors, leveraging their investments in plant and equipment for higher return in a larger end-user base. Still others may elect to take advantage of the now evolving and maturing foundry business, electing to contract out their manufacturing rather than invest in expensive new facilities for their relatively small production base. We have not, however, included any such movement

SEMM-WW-MT-9602 ©1996 Dataquest 57

58

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 6-1

Worldwide Semiconductor Production History by Region—Merchant and Captive

Semiconductor Company Sales (Millions of U.S. Dollars)

Total Americas

Merchant

Captive

Percentage Growth

Percentage Woridwide

Total Japan

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Europe

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Asia/Pacific

Merchant

Captive

Percentage Growth

Percentage Worldwide

Worldwide

Percentage Growth

Merchant

Percentage Growth

Captive

Percentage Growth

NA = Not applicable

Source: Dataquest (July 1996)

408

6.7

48.2

6,451

5,782

669

10.2

1989

22,232

18,464

3,768

8.3

37.6

28,527

28,119

10.9

1,974

1,974

NA

5.7

3.3

59,184

7.6

54,339

6.8

4,845

17.4

1990

24,202

4.0

59,328

0.2

54,545

0.4

4,783

-1.3

20,453

3,749

8.9

40.8

26,384

25,977

407

-7.5

44.5

6,350

5,723

627

-1.6

10.7

2,392

2,392

NA

21.2

64,453

8.6

59,693

9.4

4,760

-0.5

6,396

583

9.9

10.8

3,097

3,097

NA

29.5

4.8

1991

26,039

22,275

3,764

7.6

40.4

28,338

27,925

413

7.4

44.0

6,979

1993

32,643

30,942

1,701

11.8

37.3

35,515

35,515

0

25.6

40.6

11,741

11,421

320

36.7

13.4

7,636

7,636

NA

73.9

8.7

87,535

24.2

85,514

31.0

2,021

-61.1

23.1

12.2

4,391

4,391

NA

41.8

6.2

70,460

9.3

359

-0.2

40.1

8,589

7,957

632

65,260

9.3

5,200

9.2

1992

29,207

24,998

4,209

12.2

41.5

28,273

27,914

15,243

220

31.7

13.7

12,095

12,095

NA

58.4

10.8

112,464

28.5

110,514

1994

39,617

37,887

1,730

21.4

35.2

45,289

45,289

0

27.5

40.3

15,463

29.2

1,950

-3.5

33.9

13.5

21,732

21,732

NA

79.7

14.2

153,084

36.1

151,271

36.9

1,813

-7.0

1995

49,535

47,882

1,653

25.0

32.4

61,106

61,106

0

34.9

39.9

20,711

20,551

160

CAGR (%)

1989-1995

14.3

17.2

-12.8

13.5

13.8

-100.0

21.5

23.5

-21.2

49.2

49.2

17.2

18.6

-15.1 toward merchant or fabless business in our captive production forecast.

The four largest captive producers account for nearly 70 percent of the

$1.81 billion in 1995. These producers are IBM G^ipolar only), Delco

Electronics, Digital, and Northern Telecom.

Restatement of Regional Semiconductor Production from 1992: Foundry

Dataquest recently completed a detailed characterization of the foundry industry. As a result of this work, we have determined that a higher proportion of Americas region semiconductor production is being outsourced to Asia, and particularly Japan, than previously thought. We therefore felt an obligation to restate the regional production statistics from 1992.

SEMM-WW-MT-9602 ©1996 Dataquest

August 26,1996

Semiconductor Production Forecast 59

Table 6-2 shows Dataquest's estimates for the net import and export of foundry services among the regions. Clearly, the foundry industry is real and represents a key change in the semiconductor industry manufacturing infrastructure. We continue to expect U.S. companies in particular, both fabless companies and IDMs, to continue to expand their use of foundries in Asia/Pacific and Japan for semiconductor manufacturing.

T a b l e 6-2

N e t Regional Semiconductor Production b y Foundry, D e d i c a t e d a n d I D M Foundries I n c l u d e d ( M i l l i o n s of U.S. Dollars)

Americas

Japan

Europe

Asia/Pacific

1992

-491

250

-75

316

1993

-1312

584

-181

909

1994

-1,781

689

-301

1,393

Note: Positive number reflects net export, negative number reflects net Import.

Source: Dataquest (July 1996)

1995

-2,325

885

-300

1,740

^

I

The Move toward Asia Continues; European Growth Rests Temporarily

Some may find that the production trends of the last three years may contain two or three surprises. Of no surprise is the strong growth in A s i a /

Pacific production, over 14 percent of worldwide production in 1995. The strengtjh in Asia/Pacific DRAM producers and the emergence of the foundry market have been, and will continue to be in the long term, the key drivers of that growth. Expected regional capital spending and electronic equipment production will certainly maintain this production trend.

European production has expanded from just under 11 percent of the semiconductors produced in 1991 to 13.5 percent in 1995. This is also remarkable in that the last four years have been good overall growth years, resulting in the region's production nearly tripling in four years.

W h y the move to Europe? With the region's economies recovering a n d the

PC boom continuing, Europe has attracted PC production, particularly in the United Kingdom. Semiconductor production has moved along with the PC, with Intel and DRAM producers worldwide taking part. Also, the acceleration of telecommunications-related semiconductor production benefits the European comparues. Dataquest believes that, although multinationals will continue to invest heavily in Europe, the trend is in a holding pattern now in part because of the concentration of spending growth in Asia/Pacific and the recent pause in telecommunications equipment production.

There has been a sigiuficant relative decline in the percentage of the world's production being done in the United States. Over the last three years, Americas region production decreased from about 42 percent to

32 percent of the world's production overall. This is despite the fact that

U.S. companies have increased their share of the worldwide semiconductor market in the same period. Several factors are at work here. First,

N o r t h American multinational companies have been investing heavily overseas. North America has been a net exporter of capital for several

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

60 Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide years now, as foreign companies have yet to balance the scales with investments inside the United States. This trend should stabilize over the next several years, as Japanese, Taiwanese, and Korean companies have started to accelerate their investment in the United States.

Second, although U.S. companies are recognized as technology leaders, they have recently begun calling on foreign producers to manufacture their products in the foundry market. Fabless companies have been the key driver of the market to this point, but starting in 1994, w e saw a major shift in the integrated device manufacturer (or IDM, a merchant supplier of semiconductors that has a fab) to increase use of foundries. This imbalance in the concentration of foundry capacity is starting to make U.S. semiconductor companies a little nervous and has actually inipressed upon key foundry suppliers the importance of beginning to build production in the United States. In fact, TSMC recently announced plans to build a major fab in the Uruted States, and w e believe that other foundry companies are likely to follow. Clearly, users of foundry have stated a preference for close access to the fab. Any foundry provider who has capacity in the

United States is likely to have a more stable customer base than those that do not.

And third, although Japanese a n d European companies have invested somewhat outside their own countries, these companies have remained

"patriots of the domestic economy" and have kept the vast majority of investment within the region, with perhaps the exceptions of NEC and

SGS-Thomson. This, along with the strong DRAM market over the last three years, has stabilized the Japanese production proportion over the last two years at approximately a 40 percent share of the production market. However, as w e now fully realize, the DRAM market is cyclical, and

Japanese foundries will feel pressure from Asia/Pacific producers, so we expect a resumption of the gradual decay in the base of production in

Japan through the rest of the decade.

Semiconductor Production Trends: The Accelerating Shift to Asia/Pacific

Table 6-3 shows forecast semiconductor production by region for 1995 through 2001. The major trend is the growth of the Asia/Pacific region, mostly at the expense of Japan. Companies like TSMC and United

Microelectronics Corporation in Taiwan, Chartered Semiconductor in

Singapore, a n d SubMicron Technology in Thailand will continue to build capacity faster than the worldwide market. Further, new DRAM companies (Vanguard, PowerChip, and N a n Ya) have sprung u p in Taiwan, which is likely to further erode Japanese DRAM share. By 2001, Data quest believes that Asia/Pacific will expand to over 18 percent on a revenue basis.

The Americas region will remain steady on a percentage basis, as the lower cost of capital and clear leadership in technology a n d innovative design motivate companies to invest in the United States. Also, A s i a /

Pacific companies have begun a regional production diversification program, with several major fabs earmarked for the United States. European production share is expected to expand slightly, with a product mix shifting to contain a higher memory component, driven b y tihe need for proximity to PC production, as more DRAM capacity is a d d e d by large

SEMI\/l-WW-MT-9602 ©1996 Dataquest August 26,1996

Semiconductor Production Forecast 61

Table 6-3

Worldwide Semiconductor Production Forecast by Region—Merchant and Captive

Semiconductor Company Sales (Millions of U.S. Dollars)

Total Americas

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Japan

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Europe

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Asia/Pacific

Merchant

Captive

Percentage Growth

Percentage Worldwide

Worldwide

Percentage Growth

Merchant

Percentage Growth

Captive

Percentage Growth

NA = Not applicable

Source: Dataquest (July 1996)

1995

49,535

47,882

1,653

25.0

32.4

61,106

61,106

0

34.9

39.9

20,711

20,551

160

33.9

13.5

21,732

21,732

NA

79.7

14.2

153,084

36.1

151,271

36.9

1,813

-7.0

37.6

25,689

25,609

80

14.7

13.8

28,005

28,005

NA

18.8

15.1

185,840

13.1

184,241

13.3

1,599

-3.2

1997

62,319

60,800

1,519

15.3

33.5

69,827

69,827

NA

8.7

39.1

22,390

22,278

112

8.1

13.6

23,579

23,579

NA

8.5

1996

54,064

52,524

1,540

9.1

32.9

64,232

64,232

NA

5.1

14.4

164,264

7.3

162,612

7.5

1,652

-8.9

36.3

29,771

29,723

48

15.9

13.8

33,999

33,999

NA

21.4

15.8

215,397

15.9

213,833

16.1

1,564

-2.2

1998

73,364

71,848

1,516

17.7

34.1

78,263

78,263

NA

12.1

35.8

35,905

35,875

30

20.6

13.9

43,049

43,049

NA

26.6

16.7

257,838

19.7

256,247

19.8

1,591

1.7

1999

86,635

85,074

1,561

18.1

33.6

92,249

92,249

NA

17.9

N A

27.4

17.6

311,468

20.8

309,836

20.9

1,632

2.6

35.3

43,397

43,377

20

20.9

13.9

54,841

54,841

2000

103,238

101,626

1,612

19.2

33.1

109,992

109,992

N A

19.2

34.9

50,880

50,868

12

17.2

14.1

66,271

66,271

NA

20.8

18.4

359,900

2001

117,012

115,348

1,664

13.3

32.5

125,737

125,737

NA

14.3

CAGR(%)

1995-2001

15.4

15.8

0.1

12.8

12.8

15.5

358,224

15.6

1,676

2.7

16.2

16.3

-35.1

20.4

20.4

NA.

15.3

15.5

-1.3

internationals very late in the decade. Japan's share of production is likely to continue to erode as the cost of capital remains high and as Japanese companies increasingly invest in capacity overseas.

Dataquest Perspective

Where the PC goes, so go semiconductors. This is true from the perspective of the business forecast as well as the production line. Europe and

Asia/Pacific, with very large capital spending upticks over the last several years—and expected to continue that trend—will continue to gain share in world production over the next several years.

SEMM-WW-MT-9602

©1996 Dataquest

August 26,1996

62 Semiconductor Equipment, Manufacturing, and Materials Worldwide

The shifts and currents in semiconductor production trends mean that equipment a n d material suppliers will absolutely need a global presence in every sense of the word to remain competitive in the market. Product supply can no longer depend on local trends, as all major semiconductor companies have m a d e it clear they are investing on a worldwide basis.

However, local service and support are required to maintain customer satisfaction.

Taiwan is clearly the new major production growth area. We would expect Malaysia a n d Thailand to be the next major growth countries in three to five years. Evidence of this includes recent join-venture fab announcements by Texas Instruments and others. Silicon plants are now being strategically placed, such as Shin-Etsu Handotai's Malaysian plant and recently announced joint venture in Taiwan, Komatsu's joint venture with Formosa Plastics in Taiwan, and MEMC's joint ventures in both

Korea (Posco-Hiils), Taiwan (Taisil), and Malaysia (MEMC-Kulim).

Further, the concept of contract manufacturing in semiconductors is clearly here to stay. Equipment and material suppliers could find themselves selling their technical products to an international team from several companies, including the manufacturer and the designer. However, the emergence of the dedicated foundry company, taking ownership of the process and manufacturing flow, will tend to centralize this activity.

Dataquest has started a research service. Semiconductor Contract

Manufacturing Worldwide, to continue to explore the key trends in contract manufacturing and foundries, including technology trends and s u p p l y / d e m a n d balance, through the decade.

SEMM-WW-MT-9602 ©1996 Dataquest August 26,1996

I

Appendix A

Economic Assumptions, Second Quarter 1996

With the reorganization of The Dun & Bradstreet Corporation and its split into several independent companies, as well as the recent acquisition of

Dataquest by the Gartner Group, we have temporarily lost access to regional macroeconomic outlooks and country-specific economic indexes and forecasts. The most recent version we have available is from February

1996, which is somewhat dated. We have just m a d e arrangements to reestablish this service and do expect to be able to publish a summary of world economic assumptions for our next report near the end of 1996.

Unfortunately, we did not have this service established in time for the publishing deadline of this report.

If you would like to receive a copy of the February 1996 economic assumptions report, please contact your account representative or Dataquest's

Inquiry Center.

I

I

SEMM-WW-MT-9602 ©1996 Dataquest 63

^ Appendix B

' Exchange Rates

Dataquest does not forecast exchange rates per se; however, we do forecast semiconductor-related markets in several regions of the world, and we use the U.S. dollar as a common currency for intermarket comparisons and aggregation. In general, in the forecast period, Dataquest assumes that the actual exchange rate of the full month prior to the month in which the forecast-input assumptions are set will apply throughout all future months of the forecast interval. For the forecasts presented here:

• Actual monthly exchange rates were used for all months in the historical interval up to May 1996.

• The May 1996 exchange rate was then assumed to hold for June through

December 1996 and throughout 1997 to 2001.

Dataquest uses an average annual exchange rate to convert annual revenue from local currency values to U.S. dollar values. Table B-1 outlines the rates used in the forecasts presented here.

Table B-1

Exchange Rates per U.S. Dollar

Country

Australia (Dollar)

Austria (Schilling)

Belgium (Franc)

Canada (Dollar)

China (Renminbi)

D e n m a r k (Krone)

European Union (ECU)

Finland (Markka)

France (Franc)

Germany (Mark)

Great Britain (Pound)

Greece (Drachma)

H o n g Kong (Dollar)

India (Rupee)

Ireland (Punt)

Italy (Lira)

Japan (Yen)

Malaysia (Ringgit)

Mexico (Peso)

Netherlands (Guilder)

N e w Zealand (DoUar)

1995

1.35

10.06

29.42

NA

NA

5.59

0.77

4.37

4.97

1.43

0.63

231.34

7.74

32.38

0.62

1,628.21

93.90

2.51

6.41

1.60

1.52

1996*

1.29

10.52

30.72

1.36

8.35

5.88

0.80

4.68

5.09

1.50

0.66

241.89

7.73

34.65

0.64

1,567.31

107.05

2.52

7.48

1.67

1.47

U.S.DoUar

Appreciation

1995-1996 (%)*

-4.75

4.54

4.43

N M

N M

5.08

3.72

7.10

2.36

4.54

3.97

4.56

-0.02

7.00

2.44

-3.74

14.00

0.46

16.77

4.31

-3.51

(Continued)

SEMM-WW-MT-9602 ©1996 Dataquest

65

66

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Table B-1 (Continued)

Exchange Rates per U.S. Dollar

Country

Norway (Krone)

Portugal (Escudo)

Singapore (Dollar)

South Africa (Rand)

South Korea (Won)

Spain (Peseta)

Sri Lanka (Rupee)

Sweden (Krona)

Switzerland (Franc)

Taiwan (Dollar)

Thailand (Baht)

'Estimated

NA = Not tracked until 1996

NM = Not meaningful

Source: Dataquest (June 1996)

1995

6.33

149.77

1.43

NA

770.57

124.40

NA

7.14

1.18

26.48

24.91

1996*

6.47

153.99

1.41

4.10

781.03

125.07

54.07

6.73

1.21

27.25

25.28

U.S.DoUar

Appreciation

1995-1996 (%)*

2.31

2.82

-1.13

NM

1.36

0.54

NM

-5.67

2.67

2.92

1.49

SEMM-WW-MT-9602 ©1996 Dataquest

August 26,1996

i

For More Information...

Clark Fuhs, Director/Principal Analyst (408) 468-8375

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individials in the subject companies, but is not guaranteed as to accuracy or completer\ess.

Dataquest

It does not contain material provided to us in confidence by our clients. Reproduction or disclosure in whole or in part to other parties shall be made upon the written and express consent of Dataquest.

©1996 Dataquest—Reproduction Prohibited

A Gartner Group Company Dataquest is a registered trademark of A.C. Nielsen Company

I

DATAQUEST WORLDWIDE OFFICES

NORTH AMERICA

Worldwide Headquarters

251 River Oaks Parkway-

San Jose, California 95134-1913

United States

Phone: 1-408-468-8000

Facsimile: 1-408-954-1780

EUROPE

European Headquarters

Tamesis, The Glanty

Egham, Surrey TW20 9AW

United Kingdom

Phone:+44 1784 431 611

Facsimile: +44 1784 488 980

East Coast Headquarters

Nine Technology Drive

RO. Box 5093

Westborough, Massachusetts 01581-5093

United States

Phone: 1-508-871-5555

Facsimile: 1-508-871-6262

Dataquest France

Immeuble Defense Bergdres

345, avenue Georges Clemenceau

TSA 40002

92882 - Nanterre CTC Cedex 9

France

Phone: +33 1 41 35 13 00

Facsimile: +33 1 41 35 13 13

Dataquest Global Events

3990 Westerly Place, Suite 100

Newport Beach, California 92660

United States

Phone: 1-714-476-9117

Facsimile: 1-714-476-9969

Dataquest Germany

Kronstadter Strasse 9

81677 Munchen

Germany

Phone: +49 89 93 09 09 0

Facsimile: +49 89 93 03 27 7

Sales Offices:

Washington, DC (Federal)

New York, NY (Financial)

Dallas, TX

LATIN AMERICA

Research Affiliates and Sales Offices:

Buenos Aires, Argentina

Sao Paulo, Brazil

Santiago, Chile

Mexico City, Mexico

Sales Offices:

Brussels, Belgium

Kfar Saba, Israel

Milan, Italy

Randburg, South Africa

Madrid, Spain

JAPAN

Japan Headquarters

Aobadai Hills 4-7-7

Aobadai

Meguro-ku, Tokyo 153

Japan

Phone: 81-3-3481-3670

Facsimile: 81-3-3481-3644

ASIA/PACIFIC

Asia/Pacific Headquarters

7/F China Underwriters Centre

88 Gloucester Road

Wan Chai

Hong Kong

Phone: 852-2824-6168

Facsimile: 852-2824-6138

Dataquest Korea

Suite 2407, Trade Tower

159 Samsung-dong, Kangnam-gu

Seoul 135-729

Korea

Phone: 822-551-1331

Facsimile: 822-551-1330

Dataquest Taiwan

llF-2, No. 188, Section 5

Nan King East Road

Taipei

Taiwan, R.O.C.

Phone: 8862-756-0389 i

FacsimOe: 8862-756-2663 \

Dataquest Singapore

105 Cedl Street

#06-01 /02 The Octagon

Singapore 069534

Phone: 65-227-1213

Facsimile: 65-227-4607

Dataquest Thailand

12/F, Varussa Building

29 Soi Chidlom

Ploenchit Road

Patumwan, Bangkok 10330

Thailand

Phone: 662-655-0577

Facsimile: 662-655-0576

Research Affiliates and Sales Offic

Melbourne, Australia

Beijing, China

Dataquest

A Gartner Group Company

©1996 Dataquest

DataQuest

Semiconductor Five-Year Forecast

Trends—Spring 1996

Market Trends

Program: Semiconductor Equipment, Manufacturing, and Materials Worldwide

Product Code: SEMM-WW-MT-9601

Publication Date: May 13,1996

Filing: Market Trends

Five-Year Forecast

Trends—Spring 1996

Market Trends

Program: Semiconductor Equipment, Manufacturing, and i\/laterjals Worldwide

Product Code: SEMM-WW-MT-9601

Publication Date: May 13,1996

Filing: Market Trends

Semiconductor Five-Year Forecast Trends—Spring 1996

Table of Contents

Page

1. Introduction and Assumptions 1

Forecast Summary 1

Forecast Highlights 2

Exchange Rates 3

2. Worldv^ide Forecast by Product Family 5

Worldwide Forecast Data 5

3. Worldwide Semiconductor Forecast by Region 9

4. Americas Forecast by Product Family 11

5. Japan Forecast by Product Family 15

6. Europe Forecast by Product Family 19

7. Asia/Pacific Forecast by Product Family 23

8. Forecast by Product 27

Microcomponent ICs 27

Memory ICs 28

Logic ICs 29

Analogies 29

Total Monolithic ICs 29

Discrete Devices 30

Optical Semiconductors 31

9. Forecast by Technology 33

Digital MOS and Bipolar IC Forecast 33

Appendix A—^Japanese Revenue History and Forecast in Yen 35

Appendix B—European Revenue History and Forecast in ECU 39

Appendix C—Definitions 43

Appendix D—Historical Exchange Rates 47

SEMM-WW-I\/IT-9601 ©1996Dataquest IVlaylS, 1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

List of Figures

2-1 Market Share by Product, 1995 and 2000 8

3-1 Semiconductor History and Forecast by Region 9

3-2 Regional Consumption as a Percentage of Total 10

4-1 Product Comparison, Americas Market, 1995 and 2000 14

5-1 Product Comparison, Japanese Market, 1995 and 2000 18

6-1 Product Comparison, European Market, 1995 and 2000 22

7-1 Product Comparison, Asia/Pacific Market, 1995 and 2000 26

8-1 Worldwide Semiconductor Forecast by Product 27

9-1 MOS versus Bipolar Forecast 33

A-1 Comparison of Revenue Shipment Growth in the Japan

Region—Dollars versus Yen 35

B-1 Comparison of Revenue Shipment Growth in European

Region—Dollars versus ECU 39

i

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

i

Semiconductor Five-Year Forecast Trends—Spring 1996

. List of Tables

Table Page

1-1 Changes in 1996 Forecast 2

2-1 Worldwide Semiconductor Growth by Product Type 5

2-2 Worldwide Semiconductor Market, Six-Year Revenue History,

1990-1995 6

2-3 Worldwide Semiconductor Market, Five-Year Revenue Forecast,

1995-2000 6

2-4 Worldwide Semiconductor Market, Historic Revenue Growth,

1990-1995 7

2-5 Worldwide Semiconductor Market, Forecast Five-Year Revenue

Growth 7

3-1 Total Semiconductor Consumption by Region, Five-Year Revenue

Forecast, 1995-2000 10

3-2 Total Semiconductor Growth Forecast by Region 10

4-1 Americas Semiconductor Market, Six-Year Revenue History,

1990-1995 12

4-2 Americas Semiconductor Market, Five-Year Revenue Forecast

1995-2000 12

4-3 Americas Semiconductor Market, Historic Revenue Growth,

1990-1995 13

4-4 Americas Semiconductor Market, Forecast Five-Year Revenue

Growth 13

5-1 Japanese Semiconductor Market, Six-Year Revenue History,

1990-1995 16

5-2 Japanese Semiconductor Market, Five-Year Revenue Forecast,

1995-2000 16

5-3 Japanese Semiconductor Market, Historic Revenue Growth,

1990-1995 17

5-4 Japanese Semiconductor Market, Forecast Five-Year Revenue

Growth 17

6-1 European Semiconductor Market, Six-Year Revenue History,

1990-1995 20

6-2 European Semiconductor Market, Five-Year Revenue Forecast,

1995-2000 20

6-3 European Semiconductor Market, Historic Revenue Growth,

1990-1995 21

6-4 European Semiconductor Market, Forecast Five-Year Revenue

Growth 21

7-1 Asia/Pacific Semiconductor Market, Six-Year Revenue History,

1990-1995 24

7-2 Asia/Pacific Semiconductor Market, Five-Year Revenue Forecast,

1995-2000 24

7-3 Asia/Pacific Semiconductor Market, Historic Revenue Growth,

1990-1995 25

7-4 Asia/Pacific Semiconductor Market, Forecast Five-Year Revenue

Growth 25

SEI\/IM-WW-MT-9601 ©1996Dataquest i\/lay 13,1996

iy Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

List of Tables (Continued),

Table Page

8-1 Microcomponent IC Market, Five-Year Revenue Forecast,

1995-2000 28

8-2 Memory IC Market by Region, Five-Year Revenue Forecast,

1995-2000 28

8-3 Logic IC Market by Region, Five-Year Revenue Forecast,

1995-2000 29

8-4 Analog IC Market by Region, Five-Year Revenue Forecast,

1995-2000 30

8-5 Total Monolithic IC Market by Region, Five-Year Revenue

Forecast, 1995-2000 30

8-6 Discrete Device Market by Region, Five-Year Revenue Forecast,

1995-2000 30

8-7 Optical Semiconductor Market by Region, Five-Year Revenue

Forecast, 1995-2000 31

9-1 Semiconductor Market by Process Technology, Six-Year Revenue

History, 1990-1995 34

9-2 Semiconductor Market by Process Technology, Five-Year

Revenue Forecast, 1995-2000 34

A-1 Japanese Semiconductor Market, Six-Year Yen Revenue History,

1990-1995 36

A-2 Japanese Semiconductor Market, Five-Year Yen Revenue

Forecast, 1995-2000 36

A-3 Japanese Semiconductor Market, Yen Revenue Growth,

1990-1995 37

A-4 Japanese Semiconductor Market, Forecast Five-Year Yen

Revenue Growth, 1995-2000 37

B-1 European Semiconductor Market, Six-Year ECU Revenue

History, 1990-1995 40

B-2 European Semiconductor Market, Five-Year ECU Revenue

Forecast, 1995-2000 40

B-3 European Semiconductor Market, Historic Revenue Growth,

1990-1995 41

B-4 European Semiconductor Market, Forecast Five-Year ECU

Revenue Growth, 1995-2000 41

D-1 Exchange Rates 47

SEi\/lM-WW-MT-9601 ©1996 Dataquest May 13,1996

Chapter 1

9 Introduction and Assumptions,

Dataquest Semiconductor Group analysts provide a semiconductor device revenue forecast twice a year, in April and October. These revenue forecasts, which cover a five-year horizon, comprise forecasts for the major product families and the four main geographic semiconductor-consuming regions. This document, completed in April 1996, is the latest of these forecasts. Although revenue is subject to the vagaries of exchange rate variations, it is the most useful means to consolidate the forecasts of widely differing products and the most meaningful measure of markets and companies. Unit forecasts, which underlie the microcomponent and memory

IC forecasts, are doUarized to arrive at the revenue forecast presented here.

Average annual exchange rates are used for revenue history, and the most recent "average" exchange rate is extended into the five-year forecast horizon. Dataquest does not forecast exchange rates.

>

The forecast is presented in two local currencies in Appendixes A and B, in yen for the Japanese market forecast in Appendix A and in ECU for the

European market forecast in Appendix B. The Americas market and the

Asia/Pacific-ROW market are forecast only in U.S. dollars.

In 1996, the "North America" market has been expanded to include the total North and South America region and will be known as the "Americas" region from this point forward. This matches the divisions found in

Dataquest's 1995 market share data.

Forecast Summary

The PC market, now the dominant market for semiconductors, grew nearly 26 percent in 1995. Semiconductors grew by 37 percent as demand continued to outstrip supply and DRAM average selling prices (ASPs) continued strong, at $25 per megabyte. DRAM revenue growth, which was 66 percent in 1993 and 60 percent in 1994, reached 81 percent in 1995.

The brakes on this growth were applied early in 1996 as ASPs tumbled.

The declining DRAM ASPs lead a number of factors that have aligned to take our 1996 forecast down to a surprising 7.6 percent growth. Beside

DRAM, some other factors causing our 1996 forecast to drop under 8 percent are excess inventories, slowing markets, and a stronger yen. Inventory problems occurred as the fourth quarter PC market was well below expectations, leaving the first half of 1996 struggling with an inventory correction. Triggered by this correction, DRAM prices tumbled with prices per megabyte going from $25 in 1995 to under $15 early in 1996. Although we had anticipated DRAM price erosion in 1996, this price erosion occurred far sooner and faster than we had forecast last fall.

It is important to recognize that these corrections do not signal an evaporating market. Although the semiconductor end markets have slowed, they are still healthy. Dataquest's PC unit forecast for 1996 is still at 19 percent worldwide. If these problems were not severely impacting revenue, we would still be forecasting growth between 15 percent to 22 percent.

Table 1-1 shows the impact of the major downside factors on our 1996 forecast.

SEMM-WW-MT-9601 ©1996 Dataquest

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 1-1

Changes in 1996 Forecast (Percent)

DRAM Revenue Growth (%)

Non-DRAM Product Growth (%)

Yen/Dollar Exchange Rate

Total Growth in 1996 (%)

'Excludes change in yen/dollar exchange rate

Source: Dataquest (May 1996)

October 1995

Forecast

33

18

93.90

22.1

This Forecast

14*

107.05

7.6

Change to Dollar

Growth (%)

-32

-4

-12

-14.5

Change to 1996

Worldwide

Forecast (%)

-8

-3

-3

-14.5

Both DRAM and Japan represent about one-fourth of the total semiconductor market, so their impact on the worldwide 1996 forecast shows up proportionately in the right column. If the 1996 yen-dollar exchange rate does not differ from 1995, the 3 percent change to the worldwide forecast would bring it back to double digits. If DRAM prices rebound more than expected, the growth could move the forecast up into the "normal" 15 percent range. This forecast is highly leveraged off of the fortunes of these two items.

Forecast Highlights

The following are the highlights of this forecast:

• Growth in 1996 drops under 8 percent after 37 percent growth in 1995.

• The PC market slows in 1996 to 19 percent unit growth versus 26 percent in 1995.

• The MPU market slows along with PC market. Price reductions bring

96 growth down to 17 percent.

• The DRAM price per bit will decline nearly 50 percent in 1996. Even with a high rate of bit growth, revenue growth will be nonexistent.

• Non-DRAM products will grow by 14 percent in 1996, growth consistent with historical rates.

• The Asia/Pacific regional market will exceed Japan in 1998 and will grow to 25 percent of the world market in 2000.

• The Americas forecast has decreased. Even with a 17 percent 1995 through 2000 compound armual growth rate (CAGR), the Americas will lose 1 percent of the world market (to 33.7 percent) by 1999 as Americas growth slows.

• Like the Americas, the European market's growth has been revised downward to a 17 percent CAGR from 1995 through 2000. Nonetheless, the European market share will remain at 18 percent over the forecast period.

We expect the semiconductor market to pass the $300 billion mark in 2000, as the adjustments seen in 1996 will not greatly impact the long-term growth of the market.

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

Introduction and Assumptions

Exchange Rates

The following exchange rates are used for the 1994 through 1999 forecast:

• ¥107.05 per dollar

• ECU 0.774 per dollar

The following chapters will discuss the forecast by product and region in more detail.

SEMM-WW-MT-9601 ©1996 Dataqu est May 13,1996

Chapter 2

Worldwide Forecast by Product Family

The growth by product in 1995 as well as the past five-year CAGR and forecast 1995-through-2000 CAGR is shown in Table 2-1.

Memory ICs will show much slower growth as the five-year compounded growth rate of 16 percent brings memory IC growth back in line. Microcomponent growth, as well, will slow as the Americas market grows more slowly and prices stabilize. Logic ICs and analog ICs are settling into

14 percent growth rates, growtih more consistent with the growfii of electronic equipment markets. Discrete devices have gained greater growth potential with the lead of power and radio frequency (RF) transistors.

Despite the growth potential of logic ICs, analog ICs, discrete devices, and optical semiconductors and the slowdown of memory IC growth, microcomponent and memory ICs will continue to increase their share of the semiconductor market at the expense of these other categories.

The tables on the following pages provide the complete five-year forecast by product type for the worldwide semiconductor market.

Worldwide Forecast Data

Tables 2-2 through 2-5 provide the five-year forecast by product type for the worldwide semiconductor market.

Table 2-1

Worldwide Semiconductor Growth by Product Type (Revenue in Millions of Dollars)

Microcomponents

Memory Total

Logic/ASIC Total

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995 Revenue

34,513

55,421

22,961

17,607

130,502

1,935

132,437

14,023

4,811

151,271

1994-1995

Growth (%)

30.7

64.4

22.0

15.4

38.5

16.2

38.2

30.3

23.7

36.9

CAGR (%)

1990-1995

Actual

29.2

34.6

13.5

14.8

24.8

8.5

24.4

12.8

14.8

22.6

CAGR (%)

1995-2000

Forecast

17.6

16.5

13.8

14.7

16.1

1.4

15.9

11.6

12.1

15.4

SEMM-WW-MT-9601 ©1996 Dataquest

Semiconductor Equipment, IVIanufacturing, and iVIaterials Worldwide

Table 2-2

Worldwide Semiconductor Market, Six-Year Revenue History, 1990-1995 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Moiiolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical

Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1990

9,584

12,559

431

12,128

12,182

3,742

8,440

8,845

43,170

1,289

44,459

7,674

2,412

54,545

1991

11,774

13,197

356

12,841

12,972

3,272

9,700

9,517

47,460

1,395

48,855

8,035

2,804

59,694

1992

14,359

15,626

318

15,308

12,918

2,875

10,043

10,180

53,083

1,335

54,418

8,155

2,688

65,261

1993

19,947

23,550

244

23,306

15,956

2,835

13,121

12,513

71,966

1,463

73,429

9,083

3,006

85,518

1994

26,408

33,704

199

33,505

18,821

2,713

16,108

15,263

94,196

1,665

95,861

10,763

3,889

110,513

1995

34,513

55,421

160

55,261

22,961

2,337

20,624

17,607

130,502

1,935

132,437

14,023

4,811

151,271

CAGR (%)

1990-1995

29.2

34.6

-18.0

35.4

13.5

-9.0

19.6

14.8

24.8

8.5

24.4

12.8

14.8

22.6

Table 2-3

Worldwide Semiconductor Market, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical

Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

34,513

55,421

160

55,261

22,961

2337

20,624

17,607

130,502

1,935

132,437

14,023

4,811

151,271

1996

39,945

55,749

119

55,630

24,910

2,012

22,898

19,562

140,166

1,947

142,113

15,300

5,199

162,612

1997

46,524

64,213

108

64,105

27,692

1,644

26,048

21,698

160,127

2,009

162,136

16,517

5,588

184,241

1998

54,885

75,098

93

75,005

31,906

1,415

30,491

25,147

187,036

2,030

189,066

18,481

6,286

213,833

1999

65,532

93,666

79

93,587

37,212

1,219

35,993

29,531

225,941

2,055

227,996

21,044

7,207

256,247

2000

77,645

118,680

71

118,609

43,748

1,066

42,682

34,911

274,984

2,075

277,059

24,251

8,526

CAGR (%)

1995-2000

17.6

16.5

-15.0

16.5

13.8

-14.5

15.7

14.7

16.1

1.4

15.9

11.6

12.1

309,836

15.4

SEIVIM-WW-MT-9601 ©1996 Dataquest IVIay13,1996

Worldwide Forecast by Product Family

Table 2-4

Worldwide Semiconductor Market, Historic Revenue Growth, 1990-1995 (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

6.5

13.5

-0.2

-5.8

-0.3

4.8

0.2

0.4

1990

22.7

-20.8

-6.3

-21.3

3.4

-2.9

1991

22.9

5.1

-17.4

5.9

6.5

-12.6

14.9

7.6

9.9

8.2

9.9

4.7

16.3

9.4

1992

22.0

18.4

-10.7

19.2

-0.4

-12.1

3.5

7.0

11.8

-4.3

11.4

1.5

-4.1

9.3

1993

38.9

50.7

-23.3

52.2

23.5

-1.4

30.6

22.9

35.6

9.6

34.9

11.4

11.8

31.0

1994

32.4

43.1

-18.4

43.8

18.0

-4.3

22.8

22.0

30.9

13.8

30.5

18.5

29.4

29.2

15.4

38.5

16.2

38.2

30.3

23.7

36.9

1995

30.7

64.4

-19.6

64.9

22.0

-13.9

28.0

CAGR (%)

1990-1995

29.2

34.6

-18.0

35.4

13.5

-9.0

19.6

14.8

24.8

8.5

24.4

12.8

14.8

22.6

Table 2-5

Worldwide Semiconductor Market, Forecast Five-Year Revenue Growth (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

15.4

38.5

16.2

38.2

30.3

23.7

36.9

1995

30.7

64.4

-19.6

64.9

22.0

-13.9

28.0 11.0

11.1

7.5

0.6

7.4

9.1

8.1

7.6

1996

16.1

0.6

-25.6

0.7

8.5

-13.9

13.8

10.9

14.2

3.2

14.1

8.0

7.5

13.3

1997

16.5

15.2

-9.2

15.2

11.2

-18.3

16.8

1.0

16.6

11.9

12.5

16.1

1998

18.0

17.0

-13.9

17.0

15.2

-13.9

17.1

15.9

1999

19.4

24.7

-15.1

24.8

16.6

-13.9

18.0

17.4

20.8

1.2

20.6

13.9

14.7

19.8

-12.6

18.6

18.2

21.7

1.0

21.5

15.2

18.3

20.9

2000

18.5

26.7

-10.1

26.7

17.6

CAGR (%)

1995-2000

16.1

1.4

15.9

11.6

12.1

15.4

17.6

16.5

-15.0

16.5

13.8

-14.5

15.7

14.7

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

Semiconductor Equipment, Manufacturing, and l\/laterials Worldwide

The impact of these varying rates of growth by product is shown in

Figure 2-1. In 1995, the PC-driven combination of microcomponent and memory ICs gained market share rapidly, going from 54 percent of the worldwide market in 1994 to almost 60 percent of the market in 1995. In

1992, memory and microcomponent ICs combined to share only 46 percent of the semiconductor market. This gain in market share driven by PC growth will slow. As the figure shows, memories and microcomponents will only gain a 3 percent share in the coming five-year period, after gaining 12 percent in the past five years. All other semiconductor categories, logic ICs, analog ICs, discrete devices and optical semiconductors, will lose market share, but at a slower pace than in the past.

Figure 2-1

Market Share by Product, 1995 and 2000

i

1995

Optical Semiconductors (3.2%)

Hybrid IC (1.3%)

2OO0

Optics^ Semiconductors (2.8%)

Hybrid IC (0.7%)

i

Total = $151 Billion

Source: Dataquest (May 1996)

Total = $310 Billion

9639CM

SEMIVI-WW-MT-9601

©1996 Dataquest

May 13,1996

i

\

I

I

Chapter 3

Worldwide Semiconductor Forecast by Region

The worldwide revenue forecast is broken into the four constituent regional revenue shipment forecasts in Figure 3-1. A significant feature of this figure is the passing of Japan by Asia/Pacific in revenue by 1998.

The 1993-through-1995 period showed remarkable corisistency in the growth of all regions; the three-year compounded growth rates for the

Americas, Japan, Europe, and Asia/Pacific regions were 33 percent,

27 percent, 32 percent, and 39 percent, respectively. In the coming five years, these growth rates will drop by half, and regional differences wUl become more pronounced. Although we have forecast differing growth rates by region, the forecast still does not suggest a major downturn in the coming five years, only a period of adjustment. The negative growth shown for Japan in 1996 is because of an expected dollar devaluation; the growth would be nearly 12 percent in yen.

The regional revenue data for the five-year semiconductor forecast is listed in Table 3-1 and the annual growth by region in Table 3-2.

The effect of this forecast on the share of the total market by region is provided in Figure 3-2, where the lower anticipated growth for the Japanese market results in a continuing decline of the Japanese market share of the total market. The decline in the Japanese market is neatly mirrored by the rise in the Asia/Pacific market; these changes are tightly related with the shift of Japanese manufacturing to Asia/Pacific sites eriiancing the growth of Asian markets.

Figure 3-1

Semiconductor History and Forecast by Region

Billions of Dollars

100-

8 0 -

6 0 -

4 0 -

20-j

Japan

""1 1 1 [ 1

1993 1:99^ 1995 1996 1997 1998

Source: Dataquest (May 1996)

SEMM-WW-MT-9601 ©1996 Dataquest

^ ^ ^ ^ , ^ ^ ^ ^ = ^

^^^^Z" "^i^^-^—-^^*^'''

1

1999 2000

962905

10

Semiconductor Equipment, ManufacturinQ, and Materials Worldwide

Table 3-1

Total Semiconductor Consumption by Region, Five-Year Revenue Forecast, 1995-2000

(Revenue in Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Semiconductor Total

Source: Dataquest (May 1996)

1995

4 8 3 9

42,164

28,341

32,417

151,271

1996

52,478

41,244

31,479

37,411

162,612

1997

60,217

45,286

35,734

43,004

184,241

1998

70,352

51,144

41,079

51,258

213,833

1999

85,481

60,212

48,433

62,121

256,247

2000

104,579

71,693

56,828

76,736

309,836

CAGR (%)

1995-2000

16.7

11.2

14.9

18.8

15.4

Table 3-2

Total Semiconductor Growth Forecast by Region (Percentage Revenue Growth over

Preceding Year)

Americas

Japan

Europe

Asia/Pacific

Semiconductor Total

Source: Dataquest (May 1996)

1995

35.2

36.0

35.6

42.0

36.9

1996

8.8

-2.2

11.1

15.4

7.6

1997

14.7

9.8

13.5

15.0

13.3

1998

16.8

12.9

15.0

19.2

16.1

1999

21.5

17.7

17.9

21.2

19.8

2000

22.3

19.1

17.3

23.5

20.9

CAGR (%)

1995-2000

16.7

11.2

14.9

18.8

15.4

Figure 3-2

Regional Consumption as a Percentage of Total

Percentage of World Dollar Shipments

45-

Americas

Japan

Europe

• Asia/Pacific

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Source: Dataquest (May 1996)

SEMM-WW-MT-9601 ©1996 Dataquest

May 13,1996

Chapter 4

Americas Forecast by Product Family.

The five-year forecast for the Americas market (the more inclusive successor to the North America region) is based on the following assumptions:

• The Americas PC market is slowing. Windows 95 didn't materialize as

the strong driver of growth. Many businesses are waiting for VN^ndows

NT before the next big hardware/software u p g r a d e cycle. This slowing of PC demand in the business community coupled with a saturation of the home PC market has left the forecast unit growth in 1996 at 13 percent. The lowered growth expectation has impacted all PC-related business (more than 50 percent of the Americas semiconductor market).

• Pentium processors pushed u p microprocessor (MPU) revenue strongly in 1995. With Intel's Pentium price reductions, a slowing Americas market, and no looming Pentium Pro changeover in 1996, MPU market growth is expected to drop to about half of 1995's 24 percent growth.

• High-ASP semiconductors, such as x86 processors and single in-line memory modules (SIMMs), will continue to be strongly consumed in the Americas and added to PCs or motherboards manufactured in the

Asia/Pacific region.

• Price reductions in Pentium processors and free-faUing DRAM prices will accelerate the consumption of higher-performance MPUs and larger DRAM configurations. The same money will buy twice the PC in

1996; a prospect that may develop new customers but that also runs the risk of alienating home PC consumers w h o may tire of the treadmill nature of PC buying and six-month obsolescence.

• Because of the strong computer market, microcomponent and memory

ICs grew from 61 percent of semiconductor revenue in the Americas market in 1994 to 68 percent in 1995, a somewhat uimatural spurt of growth that will not be repeated in 1996. We expect this share to drop to

67 percent in 1996, because memory IC revenue growth will lag all other major device families. By the year 2000, microconiponent and memory

ICs will account for 70 percent of semiconductor revenue in the Americas, a slow ramp from 1995's 68 percent.

• DRAM price-per-bit declines of 40 percent to 50 percent will be offset by increased bit demand, but this will barely keep DRAM revenue growth positive in 1996.

• Discrete device growth (22 percent in 1994 and 30 percent in 1995) increasingly comes from the use of power MOS field-effect transistors

(MOSFETs) and insulated gate bipolar transistors (IGBTs) in switching power supplies and peripheral drivers and the increasing use of RF devices. MOSFETs and IGBTs showed 37 percent and 59 percent growth in 1995, respectively. These devices will continue to post double-digit growth in 1996.

Tables 4-1 through 4-4 provide details of the Americas semiconductor market.

SEMM-WW-MT-9601 ©1996Dataquest 11

12

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 4-1

Americas Semiconductor Market, Six-Year Revenue History, 1990-1995 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source; Dataquest (May 1996)

14,371

245

14,616

1,611

313

16,540

1990

3381

4,485

160

4,325

4,101

1,417

2,684

2,404

15,024

245

15,269

1,389

332

16,990

1991

3,916

4,641

131

4,510

4,070

1,200

2,870

2,397

1992

5,282

5,837

130

5,707

4,287

1,102

3,185

2,689

18,095

309

18,404

1,603

423

20,430

3304

25,341

288

25,629

1,811

486

27,926

1993

7,620

8,868

83

8,785

5,549

1,090

4,459

1994

9,839

12,535

66

12,469

6,323

901

5,422

3,820

32,517

347

32,864

2,212

697

35,773

1995

12,421

20,530

55

20,475

7,528

741

6,787

3,995

44,474

378

44,852

2,870

627

48,349

CAGR (%)

1990-1995

29.7

35.6

-19.2

36.5

12.9

-12.2

20.4

10.7

25.3

9.1

25.1

12.2

14.9

23.9

Table 4-2

Americas Semiconductor Market, Five-Year Revenue Forecast 1995-2000 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

12,421

20330

55

20,475

7328

741

6,787

3,995

44,474

378

44,852

2370

627

48,349

1996

14,073

21,145

40

21,105

8,400

675

7,725

4,575

48,193

345

48,538

3,225

715

52,478

1997

16,182

24,297

33

24,264

9,581

576

9,005

5315

55,375

352

55,727

3,650

840

60,217

1998

18,779

28,647

27

28,620

11,074

504

10,570

6,232

64,732

370

65,102

4,190

1,060

70352

1999

21,950

36,710

20

36,690

12,887

422

12,465

7329

78376

385

79,261

4 3 9 5

1325

85,481

2000

25,575

47307

17

47,290

15,320

360

14,960

8,652

96354

400

97,254

5,700

1,625

104379

CAGR (%)

1995-2000

15.5

18.2

-20.9

18.2

153

-13.4

17.1

16.7

163

1.1

16.7

14.7

21.0

16.7

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

Americas Forecast by Product Family

13

Table 4-3

Americas Semiconductor Market, Historic Revenue Growth, 1990-1995 (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

8.0

-3.2

-3.5

-3.2

-1.7

-4.9

-3.1

1990

20.9

-24.6

-11.1

-25.1

5.8

-2.6

10.9

1991

15.8

3.5

-18.1

4.3

-0.8

-15.3

6.9

-0.3

4.5

0

4.5

-13.8

6.1

2.7

11.0

12.2

20.4

26.1

20.5

15.4

27.4

20.2

1992

34.9

25.8

-0.8

26.5

5.3

-8.2

1993

44.3

51.9

-36.2

53.9

29.4

-1.1

40.0

22.9

40.0

-6.8

39.3

13.0

14.9

36.7

1994

29.1

41.4

-20.5

41.9

13.9

-17.3

21.6

15.6

28.3

20.5

28.2

22.1

43.4

28.1

1995

26.2

63.8

-16.7

64.2

19.1

-17.8

25.2

4.6

36.8

8.9

36.5

29.7

-10.0

35.2

CAGR (%)

1990-1995

29.7

35.6

-19.2

36.5

12.9

-12.2

20.4

10.7

25.3

9.1

25.1

12.2

14.9

23.9

Table 4-4

Americas Semiconductor Market, Forecast Five-Year Revenue Growth (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Arialog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

-17.8

25.2

4.6

36.8

8.9

36.5

29.7

-10.0

35.2

1995

26.2

63.8

-16.7

64.2

19.1

14.5

8.6

-8.7

8.5

12.4

14.0

8.8

1996

14.2

3.0

-27.3

3.1

11.6

-8.9

13.8 16.6

16.2

14.9

2.0

14.8

13.2

17.5

14.7

1997

15.0

14.9

-17.5

15.0

14.1

-14.7

16.9

5.1

16.8

14.8

26.2

16.8

1998

16.0

17.9

-18.2

18.0

15.6

-12.5

17.4

17.3

1999

16.9

28.1

-25.9

28.2

16.4

-16.3

17.9

17.6

21.9

4.1

21.7

16.8

25.0

21.5

20.0

18.1

22.8

3.9

22.7

16.4

22.6

22.3

2000

16.5

28.9

-15.0

28.9

18.9

-14.7

CAGR (%)

1995-2000

15.5

18.2

-20.9

18.2

15.3

-13.4

17.1

16.7

16.8

1.1

16.7

14.7

21.0

16.7

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

Semiconductor Equipment, Manufacturing, and Materials Worldwide

14

The effect of the Americas forecast on the relative consumption by product is shown in Figure 4-1. With 68 percent of the revenue in the Americas stemming from microcomponents and memories, the Americas market is highly dependent on the health of data processing. Witii lowered growth expectations in these two product types, we expect that market shares will hold more constant than in the past, with microcomponents actually losing 1 percent over the next five years. Logic ICs will lose a 1 percent share as bipolar logic declines.

Figure 4-1

Product Comparison, Americas Market, 1995 a n d 2000

1995

Optical Semiconductors (1.3%)

Discrete (5.9%)

Hybrid IC (0.8%)

Analog/Mixed

(8.3%)

2000

Hybrid 10 (0.4%)

Analog/Mixed

(8.3%)

Optical Semiconductors (1.6%)

Discrete (5.5%)

Total = $48.2 Billion

Source: Dataquest (May 1996)

Total = $104.6 Billion

9629017

SEMM-WW-MT-9601 ©1996 Dataquest

May 13,1996

Chapter 5

Japan Forecast by Product Family,

The five-year semiconductor forecast for the Japanese market is based on the following assumptions:

• Growth had been accelerating in Japan after the disastrous revenue decline in 1992, but this three-year growth period is slowing. A 36 percent dollar growth in 1995 will be followed by a 2 percent decline in

1996. Although the past three years have had dollar revenue enhancements because of the yen-to-dollar depreciation, the 1996 forecast includes a dollar appreciation of 14 percent, turning a 12 percent yenbased growth in 1996 to a 2 percent decline in dollars.

• The Japanese market is fundamentally sound. PC growth in Japan will drop from the 58 percent seen in 1995 to above 30 percent. The continued migration of electronic equipment manufacturing to Asia/Pacific sites is a factor that will reduce revenue growth over the forecast period, but this migration has been somewhat stunted by constraints in growing Asian infrastructures. The result of this migration is that the Japanese market will drop from 28 percent of worldwide shipments in 1994 to slightly over 23 percent in the year 2000, a lower loss of share than our past forecasts, because depreciation is expected to slow the rate of migration.

• Microcomponents will show Japan's strongest product growth in 1996 as the MPU category is dominated by the dollar-based x86 devices from

Intel. The other product categories, more strongly supplied by domestic suppliers in yen-based revenue, are impacted by the devalued dollar exchange rate. The weak 1996/1995 growth of MCU, analog, optical semiconductor, and discrete is due to sluggish consumer equipment production.

• Microcomponents show the strongest five-year compounded growth in

Japan. The 21 percent compounded growth forecast for PC shipments in

Japan provides comparable growth for the MPU category.

• MCU, analog IC, and optical semiconductor growth in Japan will be reduced by tiie offshore production shift of consumer electronics, the biggest application for these devices in Japan.

• MOS memory revenue will decline by 16 percent in dollars (negative

4 percent in yen). Even with strong PC growlii, the bit growth will be insufficient to bring revenue into positive growth as ASPs drop by half,

DRAM consumption in Japan has been pumped up by robust growth in

SIMM production, which will be impacted by any possible slowdown in worldwide PC shipments. Prices are weakening for other memory products like SRAM, flash, and MROMs.

• Optical semiconductors showed a 32 percent growth in 1995, a considerable increase over the 21 percent seen in 1994. An explosive increase in the consumption of optically oriented computer peripherals such as

CD-ROM players, scanners, and laser/LED printers has helped to fuel this growth over the past two years. This growth will be blunted in 1996 and beyond as this multimedia frenzy slows. It is not expected that new growth opportunities in DVD will be seen in the optical semiconductor category until the end of the forecast period (the year 2000).

SEMM-WW-MT-9601 ©1996 Dataquest 15

16

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Tables 5-1 through 5-4 provide details on the Japanese semiconductor market

Table 5-1

Japanese Semiconductor Market, Six-Year Revenue History, 1990-1995 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

3,490

2,723

15,018

776

15,794

2,969

1,494

20,257

1990

2,974

4,390

194

4,196

4,931

1,441

3,094

16,417

860

17,277

3,432

1,787

22,496

1991

3,579

4,393

165

4,228

5,351

1,277

4,074

1992

3,269

4,175

138

4,037

4,849

1,016

3,833

2,903

15,196

750

15,946

3,077

1,556

20,579

1993

3,987

5,697

127

5,570

5,712

1,001

4,711

3,278

18,674

820

19,494

3,423

1,728

24,645

1994

5,603

7,344

98

7,246

7,111

1,118

5,993

4,048

24,106

889

24,995

3,916

2,097

31,008

1995

7,829

12,337

82

12,255

8,772

988

7,784

4,744

33,682

1,034

34,716

4,681

2,767

42,164

CAGR (%)

1990-1995

21.4

23.0

-15.8

23.9

12.2

-7.3

17.4

11.7

17.5

5.9

17.1

9.5

13.1

15.8

Table 5-2

Japanese Semiconductor Market, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

7,829

12,337

82

12,255

8,772

988

7,784

4,744

33,682

1,034

34,716

4,681

2,767

42,164

1996

8,558

10,409

63

10,346

8,934

824

8,110

4,801

32,702

1,045

33,747

4,708

2,789

41,244

1997

10,130

11,853

60

11,793

9,663

637

9,026

4,846

36,492

1,075

37,567

4,845

2,874

45,286

1998

12,001

13,481

55

13,426

10,919

546

10373

5331

41,732

1,075

42,807

5,232

3,105

51,144

1999

14,286

17,163

49

17,114

12,501

486

12,015

5,922

49,872

1,075

50,947

5,813

3,452

60,212

2000

16,934

22,162

45

22,117

14,180

437

13,743

6,612

59,888

1,075

60,963

6,641

4,089

71,693

CAGR (%)

1995-2000

16.7

12.4

-11.3

12.5

10.1

-15.1

12.0

6.9

12.2

0.8

11.9

7.2

8.1

11.2

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

Japan Forecast by Product Family

17

Table 5-3

Japanese Semiconductor Market, Historic Revenue Growth, 1990-1995 (Percentage

Revenue Growth over Preceding Year)

MicTocomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Moiiolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1990

11.7

-24.6

1.6

-25.5

2.7

-1.1

4.3

-0.4

-6.2

-in

-6.3

-3.6

-3.7

-5.7

1991

20.3

0.1

-14.9

0.8

8.5

-11.4

16.7

13.6

9.3

10.8

9.4

15.6

19.6

11.1

1992

-8.7

-5.0

-16.4

-4.5

-9.4

-20.4

-5.9

-6.2

-7.4

-12.8

-in

-10.3

-12.9

-8.5

1993

22.0

36.5

-8.0

38.0

17.8

-1.5

22.9

12.9

22.9

9.3

22.3

11.2

11.1

19.8

1994

40.5

28.9

-22.8

30.1

24.5

11.7

27.2

23.5

29.1

8.4

28.2

14.4

21.4

25.8

1995

39.7

68.0

-16.3

69.1

23.4

-11.6

29.9

17.2

39.7

16.3

38.9

19.5

32.0

36.0

CAGR (%)

1990-1995

21.4

23.0

-15.8

23.9

12.2

-7.3

17.4

11.7

17.5

5.9

17.1

9.5

13.1

15.8

Table 5-4

Japanese Semiconductor Market, Forecast Five-Year Revenue Growth (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

29.9

17.2

39.7

16.3

38.9

19.5

32.0

36.0

1995

39.7

68.0

-16.3

69.1

23.4

-11.6

1996

9.3

-15.6

-23.2

-15.6

1.8

-16.6

4.2

1.2

-2.9

1.1

-2.8

0.6

0.8

-2.2

1997

18.4

13.9

-4.8

14.0

8.2

-22.7

11.3

0.9

11.6

2.9

11.3

2.9

3.0

9.8

14.9

10.0

14.4

0

13.9

8.0

8.0

12.9

1998

18.5

13.7

-8.3

13.8

13.0

-14.3

15.8

11.1

19.5

0

19.0

11.1

11.2

17.7

1999

19.0

27.3

-10.9

27.5

14.5

-11.0

2000

18.5

29.1

-8.2

29.2

13.4

-10.1

14.4

11.7

20.1

0

19.7

14.2

18.5

19.1

CAGR (%)

1995-2000

16.7

12.4

-11.3

12.5

10.1

-15.1

12.0

6.9

12.2

0.8

11.9

7.2

8.1

11.2

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

18

Semiconductor Equipment, IVIanufacturing, and Materials Worldwide

Figure 5-1 illustrates the effect of the Japanese market forecast on the relative consumption by product. The figure highlights three main trends.

First, microcomponents are expected to track PC growth in Japan. Second, memory IC price erosion will hold memory growth down over the forecast period. Third, the non-DRAM, non-MPU devices will decline in market share as these devices increasingly move toward offshore equipment production. With a memory and microcomponent market share that is 20 percent less than that of the Americas, the Japanese market has had less dependence on the PC. Personal computers will make strong gains in the

Japanese market in the coming years.

Figure 5-1

Product Comparison, Japanese Market, 1995 and 2000

i

1995

Optical Semiconductors (6.6%)

Hybrtd IC (2.5%)

2000

Optical Semiconductors (5,7%)

Hybrid IC (1.5%)

Total = $42.2 Billion

Source: Dataquest (May 1996)

Total = $71.7 Billion

S629D8

i

SEMI\/I-WW-MT-9601

©1996 Dataquest May 13,1996

i

Chapter 6

Europe Forecast by Product Family,

The five-year semiconductor forecast for the European market, shown on the following pages, is based on these assumptions:

• With two consecutive years of growth exceeding 35 percent, the European market has shown considerably more strength than we had expected. This growth, based on the PC and personal communications booms, is expected to moderate in 1996.

• The European PC market, which grew by 25 percent in 1995, is stQl expected to do more than 20 percent in 1996. Declining prices for

DRAM will limit the semiconductor ride on this boom, however.

• DRAM revenue will be flat in 1996. Double-digit growth will return in

1997, although at a compounded rate below 20 percent. The more stable prices seen in 1997 will result in DRAM revenue growth consistent with

PC unit growth (17 percent).

• MCU growth continued strongly into 1995, with revenue growth exceeding 40 percent. ASP erosion and a slowing of demand will limit revenue growth in 1996, and beyond, to less than 20 percent.

• Shortages in discrete products erJianced the market in 1994 and 1995 as

ASPs were kept high. In 1995, a 45 percent annual growth more than doubled the 19 percent seen in 1994. Discrete growth will drop into lower growth in 1996 (10 percent) and beyond (11 percent CAGR, 1995 through 2000).

Tables 6-1 through 6-4 provide details on the European semiconductor market.

Figure 6-1 illustrates the consumption by product changes for the European market over the forecast period. Unlike past years of memory and microcomponent market incursion, the product mix remains fairly consistent over the forecast period. By the year 2000, microcomponents and memory ICs are expected to account for 62 percent of semiconductor shipment revenue, up slightly from the 60 percent of 1995. The growth in microcomponents derives from all segments of the microcomponent category, the MPUs and microperipherals (MPRs) in computers and the microcontroller (MCU) and digital signal processor (DSP) ICs used in communications and consumer products.

SEMM-WW-MT-9601 ©1996Dataquest 19

20

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 6-1

European Semiconductor Market, Six-Year Revenue History, 1990-1995 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

2,169

7,958

157

8,115

1,895

405

10,415

1990

1,802

2,105

55

2,050

1,882

510

1,372

1991

2,082

2,172

43

2,129

2,085

443

1,642

2,184

8,523

178

8,701

1,828

485

11,014

1992

2,723

2,698

38

2,660

2,137

388

1,749

2,249

9,807

151

9,958

1,826

434

12,218

1993

4,037

4,067

27

4,040

2,299

363

1,936

2,736

13,139

179

13,318

1,769

374

15,461

1994

5,408

6,602

28

6,574

2,659

329

2,330

3,370

18,039

178

18,217

2,108

575

20,900

1995

7,009

9,990

19

9,971

3,243

291

2,952

4,127

24,369

239

24,608

3,053

680

28,341

CAGR(%)

1990-1995

31.2

36.5

-19.2

37.2

11.5

-10.6

16.6

13.7

25.1

8.8

24.8

10.0

10.9

22.2

Table 6-2

European Semiconductor Market, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

7,009

9,990

19

9,971

3,243

291

2,952

4,127

24,369

239

24,608

3,053

680

28,341

1996

8,503

10,321

13

10,308

3,621

251

3,370

4,630

27,075

249

27,324

3,367

788

31,479

1997

9,775

11,962

13

11,949

3,974

203

3,771

5,306

31,017

245

31,262

3,603

869

35,734

1998

11365

13,917

10

13,907

4,540

174

4,366

6,049

35,871

248

36,119

3,985

975

41,079

1999

13,454

16,728

9

16,719

5,250

151

5,099

7,149

42,581

258

42,839

4,491

1,103

48,433

2000

15,437

19,919

8

19,911

6,175

134

6,041

8,580

50,111

263

50,374

5,178

1,276

56,828

CAGR (%)

1995-2000

17.1

14.8

-15.9

14.8

13.7

-14.4

15.4

15.8

15.5

1.9

15.4

11.1

13.4

14.9

SEMM-WW-MT-9601

©1996 Dataquest

May 13,1996

Europe Forecast by Product Family

21

Table 6-3

European Semiconductor Market, Historic Revenue Growth, 1990-1995

(Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1990

25.0

-15.4

-22.5

-15.2

-3.4

-8.3

-1.4

39.4

7.0

15.4

7.2

20.4

14.4

9.7

-13.1

19.7

0.7

7.1

13.4

7.2

1991

15.5

3.2

-21.8

3.9

10.8

-3.5

19.8

5.8

1992

30.8

24.2

-11.6

24.9

2.5

-12.4

6.5

3.0

15.1

-15.2

14.4

-0.1

-10.5

10.9

1993

48.3

50.7

-28.9

51.9

7.6

-6.4

10.7

21.7

34.0

18.5

33.7

-3.1

-13.8

26.5

-9.4

20.4

23.2

37.3

-0.6

36.8

19.2

53.7

35.2

1994

34.0

62.3

3.7

62.7

15.7

1995

29.6

51.3

-32.1

51.7

22.0

-11.6

26.7

22.5

35.1

34.3

35.1

44.8

18.3

35.6

CAGR (%)

1990-1995

31.2

36.5

-19.2

37.2

11.5

-10.6

16.6

13.7

25.1

8.8

24.8

10.0

10.9

22.2

Table 6-4

European Semiconductor Market, Forecast Five-Year Revenue Growth (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

29.6

51.3

-32.1

51.7

22.0

-11.6

26.7

22.5

35.1

34.3

35.1

44.8

18.3

35.6

1996

21.3

3.3

-31.6

3.4

11.7

-13.7

14.2

12.2

11.1

4.2

11.0

10.3

15.9

11.1

14.6

14.6

-1.6

14.4

7.0

10.3

13.5

1997

15.0

15.9

0

15.9

9.7

-19.1

11.9

1998

16.3

16.3

-23.1

16.4

14.2

-14.3

15.8

14.0

15.6

1.2

15.5

10.6

12.2

15.0

1999

18.4

20.2

-10.0

20.2

15.6

-13.2

16.8

18.2

18.7

4.0

18.6

12.7

13.1

17.9

2000

14.7

19.1

-11.1

19.1

17.6

-11.3

18.5

20.0

17.7

1.9

17.6

15.3

15.7

17.3

CAGR (%)

1995-2000

17.1

14.8

-15.9

14.8

13.7

-14.4

15.4

15.8

15.5

1.9

15.4

11.1

13.4

14.9

SEMI\/1-WW-MT-9601

©1996 Dataquest May 13,1996

22

Semiconductor Equipment, IVIanufacturing, and l\yiaterials Worldwide

Figiure 6-1

Product Comparison, Eiu-opean Market, 1995 and 2000

1995

Optical Semiconductors (2.4%)

Hybrid IC (0.8%)

2000

Optical Semiconductors (2.2%)

Hybrid IC (0.5%)

Total = $28.3 Billion

Source: Dataquest (May 1996)

Total = $56.8 Billion

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

Chapter 7

Asia/Pacific Forecast by Product FamilnL

The five-year forecast for the Asia/Pacific region shown on the following pages is based on the following assumptions:

• The PC business was slower than expected in 1995 and will slow again in 1996. DRAM, SRAM, and MPU growth have been reduced in 1996.

The combined MOS memory growth will decline to 10 percent after

74 percent in 1995. Microcomponent growth will drop from 30 percent in 1995 to 21 percent in 1996.

• Decreasing ASPs, down 40 percent over those of 1995, will offset much of the bit growth in Asia/Pacific, reducing DRAM revenue growth below 10 percent in 1996.

• SRAM ASPs are declining, but not falling. SRAM revenue growth will drop to less than half of tive 60 percent growth seen in 1995.

• Asia/Pacific's microprocessor market is almost totally dominated by x86 architectures. With ever-shortening PC life cycles, most PC products are shipped without the MPU on-board—92 percent of motherboards,

83 percent of desktop PCs, and 80 percent of notebook computers from

Taiwan are shipped this way. This trend will continue over the forecast period.

• China and the southern Asia/Pacific regions have shown sfrong growth in telecom and consumer equipment. High-end telecommunications equipment is being built in China, and the shipment of pagers and cellular phones continues to expand.

• A strengthening yen continues to drive elecfronic equipment production out of Japan and into the Asia/Pacific region. This production shift enhances the Asia/Pacific growth that comes with the growth of its own consuming markets. As a semiconductor consuming region, Asia/

Pacific will pass Japan in 1998.

Tables 7-1 through 7-4 provide details on the Asia/Pacific semiconductor market.

Figure 7-1 shows the impact of the five-year product forecast on the relative shares of the total Asia/Pacific market. The combined memory-microcomponent IC share increased from 56 percent of the market in 1994 to

61 percent in 1995. Like the three other geographical regions, Asia/Pacific will see little gain in the memory-microcomponent market (to 64 percent in 2000) as prices correct and the PC market slows. Analog and logic ICs, less affected by price erosion, will maintain market position.

SEMM-WW-MT-9601 ©1996Dataquest 23

24

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 7-1

Asia/Pacific Semiconductor Market, Six-Year Revenue History, 1990-1995 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1990

1,427

1,579

22

1,557

1,268

374

894

1,549

5,823

111

5,934

1,199

200

7 3 3 3

1,842

7,496

112

7,608

1,386

200

9,194

1991

2,197

1,991

17

1,974

1,466

352

1,114

1992

3,085

2,916

12

2,904

1,645

369

1,276

2,339

9,985

125

10,110

1,649

275

12,034

3,195

14,812

176

14,988

2,080

418

17,486

1993

4,303

4,918

7

4,911

2,396

381

2,015

4,025

19,534

251

19,785

2,527

520

22,832

1994

5,558

7,223

7

7,216

2,728

365

2,363

1995

7,254

12,564

4

12,560

3,418

317

3,101

4,741

27,977

284

28,261

3,419

737

32,417

CAGR (%)

1990-1995

38.4

51.4

-28.9

51.8

21.9

-3.3

28.2

25.1

36.9

20.7

36.6

23.3

29.8

34.6

Table 7-2

Asia/Pacific Semiconductor Market, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1995

7,254

12,564

4

12,560

3,418

317

3,101

4,741

27,977

284

28,261

3,419

737

32,417

1996

8,811

13,874

3

13,871

3,955

262

3,693

5,556

32,196

308

32,504

4,000

907

37,411

1997

10,437

16,101

2

16,099

4,474

228

4,246

6,231

37,243

337

37,580

4,419

1,005

43,004

1998

12,740

19,053

1

19,052

5,373

191

5,182

7,535

44,701

337

45,038

5,074

1,146

51,258

1999

15342

23,065

1

23,064

6,574

160

6,414

9,131

54,612

337

54,949

5,845

1,327

62,121

2000

19,699

29,292

1

29,291

8,073

135

7,938

11,067

68,131

337

68,468

6,732

1,536

76,736

CAGR(%)

1995-2000

22.1

18.4

-24.2

18.5

18.8

-15.7

20.7

18.5

19.5

3.5

19.4

14.5

15.8

18.8

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

Asia/Pacific Forecast by Product Family

25

Table 7-3

Asia/Pacific Semiconductor Market, Historic Revenue Growth, 1990-1995 (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

1990

57.2

-1.6

22.2

-1.9

9.7

-3.1

16.1

21.5

17.8

-19.0

16.8

16.7

16.3

16.8

1991

54.0

26.1

-22.7

26.8

15.6

-5.9

24.6

18.9

28.7

0.9

28.2

15.6

0

25.4

1992

40.4

46.5

-29.4

47.1

12.2

4.8

14.5

27.0

33.2

11.6

32.9

19.0

37.5

30.9

1993

39.5

68.7

-41.7

69.1

45.7

3.3

57.9

36.6

48.3

40.8

48.2

26.1

52.0

45.3

26.0

31.9

42.6

32.0

21.5

24.4

30.6

1994

29.2

46.9

0

46.9

13.9

-4.2

17.3

17.8

43.2

13.1

42.8

35.3

41.7

42.0

1995

30.5

73.9

-42.9

74.1

25.3

-13.2

31.2

CAGR (%)

1990-1995

38.4

51.4

-28.9

51.8

21.9

-3.3

28.2

25.1

36.9

20.7

36.6

23.3

29.8

34.6

Table 7-4

Asia/Pacific Semiconductor Market, Forecast Five-Year Revenue Growth (Percentage

Revenue Growth over Preceding Year)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Source: Dataquest (May 1996)

17.8

43.2

13.1

42.8

35.3

41.7

42.0

1995

30.5

73.9

-42.9

74.1

25.3

-13.2

31.2

1996

21.5

10.4

-25.0

10.4

15.7

-17.4

19.1

17.2

15.1

8.5

15.0

17.0

23.1

15.4

15.0

12.1

15.7

9.4

15.6

10.5

10.8

15.0

1997

18.5

16.1

-33.3

16.1

13.1

-13.0

22.0

20.9

20.0

0

19.8

14.8

14.0

19.2

1998

22.1

18.3

-50.0

18.3

20.1

-16.2

23.8

21.2

22.2

0

22.0

15.2

15.8

21.2

1999

24.3

21.1

0

21.1

22.4

-16.2

23.8

21.2

24.8

0

24.6

15.2

15.7

23.5

2000

24.3

27.0

0

27.0

22.8

-15.6

CAGR (%)

1995-2000

22.1

18.4

-24.2

18.5

18.8

-15.7

20.7

18.5

19.5

3.5

19.4

14.5

15.8

18.8

SEMM-WW-MT-9601

©1996 Dataquest

May 13,1996

26

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Figure 7-1

Product Comparison, Asia/Pacific Market, 1995 and 2000

1995

Optical Semiconductors (2.3%)

Hybrid IC ( 0 . 9 % ) ^ ^.^^^^^

2000

Optical Semiconductors (2.0%)

Hybrid IC (0,4%}__^^._J-p—^^.^^

Ivy^Discrete I

/ ' ^ ( 1 0 . 5 % ) \

/ Analog/Mixed ^ V '

I (14.6%) ^ N

Microcomponent \

(22.4%) \

V^DIscrete 1

/ ^ 8 . 8 % ) \

Microcomponent \

/ Analog/Mixed X i

(25.7%) \

/ (14.4%) \

Vtogic/ASIC /

\ (10.5%) /

Memory m

(38.8%) #

\ Logic/ASIC y/^

\ ( 1 0 . 5 % ! /

\

Memory M

(38.2%) M

Total = $32.4 Billion

Source: Dataquest (l^ay 1996)

Total = $76.7 Billion

96Z910

SEMM-WW-MT-9601 ©1996 Dataquest

May 13,1996

^

I

I

Chapter 8

Forecast by Product

Chapter 2 provided a brief discussion of the semiconductor product families. This diapter focuses on the individual products and summarizes the regional splits for each product category.

Figure 8-1 graphs the worldwide forecast by category for the forecast period. A major change to the forecast is the flat memory growth seen in

1996—a departure from last year's forecast, where we expected decelerating growth, but growth nevertheless. After this three-year correction period, memory revenue will again outpace microcomponent revenue growth, widening the gap toward the end of the forecast period.

Each of these major product categories is discussed in the following sections, and a regional forecast table is provided.

Microcomponent ICs

After six consecutive years of growth exceeding 20 percent, microcomponent growth is slowing. Growth will drop below 20 percent in 1996 and remain in the high teens for the duration of the forecast. The PC market is slowing somewhat and will post growth under 20 percent worldwide over the coming five years. Microprocessor ASPs wiU not rise as rapidly as in the recent past, and the slowing growth of PCs and multimedia peripherals will limit microperipheral growth. Communications and digital entertainment will keep DSP growtti above 20 percent compounded.

Microcontrollers continue to find new homes in every conceivable electronic product and will help hold the microcomponent CAGR near 18 percent. Table 8-1 shows the microcomponent growth by region for the coming five years, with some product detail presented below.

Figure 8-1

Worldwide Semiconductor Forecast by Product

Billions of Dollars

120

1 0 0 -

Optoelectronic

Discrete

Hybrid iC

Bipolar Digital

MOS Memory

MOS

Microcomponent

MOS Logic

Analog

1993 1994 1995 1996 1997 1998 1999 2000

Source: Dataquest (May 1996)

SEMM-WW-MT-9601

©1996 Dataquest

362311

27

28

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 8-1

Microcomponent IC Market, Five-Year Revenue Forecast, 1995-2000 (Revenue i n

Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Microcomponent IC Total

Source: Dataquest (May 1996)

1995

12,421

7,829

7,009

7,254

34,513

1996

14,073

8,558

8,503

8,811

39,945

1997

16,182

10,130

9,775

10,437

46,524

1998

18,779

12,001

11,365

12,740

54,885

1999

21,950

14,286

13,454

15,842

65,532

2000

25,575

16,934

15,437

19,699

77,645

CAGR (%)

1995-2000

15.5

16.7

17.1

22.1

17.6

Memory iCs

By accounting for 28 percent of total semiconductor revenue, DRAM has had an enormous effect on total semiconductor growth. With an 82 percent

DRAM revenue growth in 1995, the semiconductor market grew by

37 percent; excluding DRAM revenue growth, all other semiconductor products showed a combined growth of 25 percent. In 1996, Dataquest anticipates no DRAM growth, a problem that will limit total semiconductor growth to 8 percent even as non-DRAM devices will grow by 14 percent on average.

Memory IC demand will continue unabated in 1996. The only difference is that we are in oversupply and prices have declined precipitously. The year

1996 marks the end of the DRAM shortage and the return to the "normal" declining price-per-bit scenario. Bit growth is expected to be substantial but not sufficient to counter the large price-per-bit declines that have dropped price-per-megabyte below $14. A compounded revenue growth rate of 17 percent for DRAM over the 1995-through-2000 period, although a drop from the past five years, will allow DRAM to grow faster than the semiconductor market and account for more than 30 percent of the semiconductor revenue in the year 2000. The five-year compounded growth for memory ICs has dropped to 16.5 percent. Despite a drop in revenue in

Japan in 1996, all four regions will show double-digit CAGRs over the forecast period. Table 8-2 shows the memory IC forecast by region.

Table 8-2

Memory IC Market by Region, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Memory IC Total

Source: Dataquest (May 1996)

1995

20,530

12,337

9,990

12,564

55,421

1996

21,145

10,409

10,321

13,874

55,749

1997

24,297

11,853

11,962

16,101

64,213

1998

28,647

13,481

13,917

19,053

75,098

1999

36,710

17,163

16,728

23,065

93,666

2000

47,307

22,162

19,919

29,292

118,680

CAGR (%)

1995-2000

18.2

12.4

14.8

18.4

16.5

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

Forecast by Product 29

Logic ICs

Logic ICs include a broad and dissimilar set of products. These products can be cut by standard or ASIC, bipolar or MOS. A traditional cut used in this forecast is that of process technology—bipolar logic and MOS logic, which more or less track an "old versus new" division. After a two-year respite from rapidly declining revenue in 1993 and 1995, bipolar logic returned to a 14 percent decline in 1995, a "normal" rate that we expect to continue into 1996 and beyond.

MOS logic showed a 28 percent growth in revenue in 1995 after 27 percent in 1994. We expect growth to drop to eleven percent in 1996 and then settle into a long-term 16 percent growth rate driven by the still strong MOS programmable logic device (PLD), MOS gate array, and MOS cell-based products. MOS ASIC is the major driver of the MOS logic category. The total logic data combines both bipolar and MOS logic, giving an aggregate growth of 22 percent for 1995 and a five-year CAGR of 14 percent over the forecast period. Table 8-3 gives the combined logic forecast.

Analog ICs

Consumer entertainment products, being largely audio and video, are intrinsically analog in nature and have typically consumed about 40 percent of all analog ICs. The big declines seen in 1992 in the consumer market, especially in Japan and Europe, severely impacted the growth of analog ICs, resulting in a growth of only 6 percent. Since 1992, analog ICs have shown a consistent 23 percent annual growth. In 1995, we saw a drop from this trend, with a 15 percent growth.

Analog ICs show a very equal distribution among the four regions, with the Americas having the smallest share at 23 percent and Japan the largest at 27 percent. This distribution is changing as consumer equipment manufacturing increasingly migrates to Asia/Pacific sites. The increasing presence of analog ICs in computer and communications applications is stabilizing growth in the Americas and Europe. Table 8-4 shows the analog

IC growth rate by region over the forecast period.

Total Monolithic ICs

The combination of microcomponent, memory, logic, and analog ICs gives the total monolithic IC market. The five-year forecast for this summary category is shown in Table 8-5.

Table 8-3

Logic IC Market by Region, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Logic IC Total

Source: Dataquest (May 1996)

1995

7,528

8,772

3,243

3,418

22,961

1996

8,400

8,934

3,621

3,955

24,910

1997

9,581

9,663

3,974

4,474

27,692

1998

11,074

10,919

4,540

5,373

31,906

1999

12,887

12,501

5,250

6,574

37,212

2000

15,320

14,180

6,175

8,073

43,748

CAGR (%)

1995-2000

15.3

10.1

13.7

18.8

13.8

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

30

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 8-4

Analog IC Market by Region, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Americas

Japan

Eiirope

Asia/Pacific

Analog IC Total

Source: Dataquest (May 1996)

1995

3,995

4,744

4,127

4,741

17,607

1996

4,575

4,801

4,630

5,556

19,562

1997

5,315

4,846

5,306

6,231

21,698

1998

6,232

5,331

6,049

7,535

25,147

1999

7329

5,922

7,149

9,131

29,531

2000

8,652

6,612

8,580

11,067

34,911

CAGR(%)

1995-2000

16.7

6.9

15.8

18.5

14.7

Table 8-5

Total Monolithic IC Market by Region, Five-Year Revenue Forecast, 1995-2000 (Revenue in Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Monolithic IC Total

Source: Dataquest (May 1996)

1995

44,474

33,682

24,369

27,977

130,502

1996

48,193

32,702

27,075

32,196

140,166

1997

55,375

36,492

31,017

37,243

160,127

1998

64,732

41,732

35,871

44,701

187,036

1999

78,876

49,872

42,581

54,612

225,941

2000

96,854

59,888

50,111

68,131

274,984

CAGR(%)

1995-2000

16.8

12.2

15.5

19.5

16.1

Discrete Devices

Discrete devices showed a 30 percent revenue growth in 1995. Although the discrete device category has been losing market share because of tiie relentless integration of components, this category remains viable because power and RF devices are not readily integrated. Power transistors represent about one-third of discrete revenue and are expected to lead the discrete growth with a 14 percent CAGR. Table 8-6 gives the discrete forecast by region. The growing use of power discrete devices in power control and communications applications in the Americas has brought the compounded Americas growth rate back into double digits.

Table 8-6

Discrete Device Market by Region, Five-Year Revenue Forecast, 1995-2000 (Revenue in

Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Discrete Devices Total

Source: Dataquest (May 1996)

1995

2,870

4,681

3,053

3,419

14,023

1996

3,225

4,708

3,367

4,000

15,300

1997

3,650

4,845

3,603

4,419

16,517

1998

4,190

5,232

3,985

5,074

18,481

1999

4,895

5,813

4,491

5,845

21,044

2000

5,700

6,641

5,178

6,732

24,251

CAGR (%)

1995-2000

14.7

7.2

11.1

14.5

11.6

SEMM-WW-MT-9601

©1996 Dataquest May 13,1996

Forecast by Product 31

^

I

Optical Semiconductors

Even more than analog ICs or discrete devices, optical semiconductors find their primary market in consumer entertainment products. With scanners and copiers using charge-coupled devices (CCDs), CD-ROMs using laser diodes, and optical-fiber data links using semiconductor receivers and transmitters, the data processing market is showing an increasing impact on the optical semiconductor market. This impact was seen as a

24 percent revenue growth in 1995. Growth in 1996 is anticipated to be

8 percent as the computer peripherals and consumer markets slow. Laser diodes have continued to lead the growth in this category; 1996 shows a

36 percent revenue growth for this product type. The optical semiconductor forecast by region is given in Table 8-7.

Table 8-7

Optical Semiconductor Market by Region, Five-Year Revenue Forecast, 1995-2000

(Revenue in Millions of Dollars)

Americas

Japan

Europe

Asia/Pacific

Optical Semiconductors

Total

Source: Dataquest (May 1996)

1995

627

2,767

680

737

4,811

1996

715

2,789

788

907

5,199

1997

840

2,874

869

1,005

5,588

1998

1,060

3,105

975

1,146

6,286

1999

1,325

3,452

1,103

1,327

7,207

2000

, CAGR (%)

1995-2000

1,625

4,089

1,276

1,536

8,526

21.0

8.1:

13.4

15.8

12.1

I

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

I

I

Chapter 9

Forecast by Technology

Digital IVIOS and Bipolar IC Forecast

The five-year IC forecast includes the process categories of MOS digital and bipolar digital ICs. This process split is still important for the logic IC category but is of decreasing importance for the memory IC category. For microcomponent ICs, the bipolar subsegment has become fairly irrelevant and is no longer reported or forecast separately.

The forecast data for digital ICs, by process, is plotted in Figure 9-1. The graph shows that the bipolar portion of the digital IC market is declining at a 12 percent CAGR over the forecast period. By the year 2000, bipolar digital ICs will have declined to less than 0.5 percent of the total digital IC market.

Tables 9-1 and 9-2 show the five-year history and forecast, respectively, for the bipolar and MOS portions of the three main digital IC categories. It can be seen that, as a memory IC process technology, bipolar has been in a rapid slide that is slowing as revenue becomes insignificant. Bipolar logic

ICs accounted for 14 percent of logic IC revenue in 1994. By 2000, it is expected that bipolar logic will represent less than 3 percent of the logic IC revenue.

Figure 9-1

MOS versus Bipolar Forecast

Billions of Dollars

— - — • — '

MOS

_ _ — "

100-

_ _ ^ ^

1 0 -

Bipolar

1 -

1 1 1

1 1

1

19

90 1991 1992 1993

1994 1995 1996 1997

1

1998

Source: Dataquest (May 1996)

1 ~1

30

962912

33

SEMM-WW-MT-9601 ©1996 Dataquest

34

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 9-1

Semiconductor Market by Process Technology, Six-Year Revenue History, 1990-

1995

(Revenue in Millions of Dollars)

Bipolar Total

Bipolar Memory

Bipolar Logic

MOS Total

MOS Micro

MOS Memory

MOS Logic

Total Digital IC

Source: Dataquest (May 1996)

1990

4,173

431

3,742

30,152

9,584

12,128

8,440

34,325

1991

3,628

356

3,272

34,315

11,774

12,841

9,700

37,943

1992

3,193

318

2,875

39,710

14,359

15,308

10,043

42,903

1993

3,079

244

2,835

56,374

19,947

23,306

13,121

59,453

1994

2,912

199

2,713

76,021

26408

33,505

16,108

78,933

1995

2,497

160

2,337

110,298

34,513

55,261

20,624

112,895

CAGR(%)

1990-1995

-9.8

-18.0

-9.0

29.6

29.2

35.4

19.6

26.9

Table 9-2

Semiconductor Market by Process Technology, Five-Year Revenue Forecast, 1995-2000

(Revenue in Millions of Dollars)

Bipolar Total

Bipolar Memory

Bipolar Logic

MOS Total

MOS Micro

MOS Memory

MOS Logic

Total Digital IC

Source: Dataquest (May 1996)

1995

2,497

160

2,337

110,298

34,513

55,261

20,624

112,895

1996

2,131

119

2,012

118,473

39,945

55,630

22,898

120,604

1997

1,752

108

1,644

136,677

46,524

64,105

26,048

138,429

1998

1,508

93

1,415

160,381

54,885

75,005

30,491

161,889

1999

1,298

79

1,219

195,112

65,532

93,587

35,993

196,410

2000

1,137

71

1,066

238,936

77,645

118,609

42,682

240,073

CAGR (%)

1995-2000

-14.6

-15.0

-14.5

16.7

17.6

16.5

15.7

16.3

SEMM-WW-MT-9601 ©1996 Dataquest May 13,1996

^

I

Appendix A

Japanese Revenue History and Forecast in Yen.

Revenue growth in shipments to the Japan region differs according to whether the dollar or yen is used as the currency basis. As the dollar has tjrpically weakened against the yen, Japanese growth has often been inflated by this exchange rate change. Figure A-1 shows the annual growth in each of these two currencies over both the historical 1988through-1995 period and the forecast 1996-through-2000 period. Because

Dataquest does not forecast exchange rates, the forecast growth rates are the same.

The following tables show the yen-based revenue shipment data for the

Japan region. Tables A-1 and A-2 provide the Japanese revenue history and forecast, respectively, in yen. The historical exchange rates are shown at the bottom of these tables. Tables A-3 and A-4 show the annual growth associated with the year-to-year revenue growth. The rate of dollar appreciation against the yen for the period from 1990 through 1995 is shown at the bottom of Table A-3. Over tiie past five years, the dollar has declined in value, inflating the revenue growtibi of the Japanese market in dollars.

Figure A-1

Comparison of Revenue Shipment Growth in the Japan Region—Dollars versus Yen

Annual Growth (%)

50-

^ Growth in Dollars

- ^ Growth in Yen

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

962913

Source: Dataquest (May 1996)

t

SEMM-WW-MT-9601

©1996 Dataquest

35

36

Semiconductor Equipment, IVIanufacturing, and i\/laterials Worldwide

Table A-1

Japanese Semiconductor Market, Six-Year Yen Revenue History, 1990-1995 (Revenue in

Billions of Yen)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Yen/U.S.$ Exchange Rate

Source: Dataquest (May 1996)

208

503

392

2,163

112

2,274

1990

428

632

28

604

710

428

215

2,917

144.00

2,233

117

2,350

467

243

3,059

136.00

1991

487

597

22

575

728

174

554

421

128

485

367

1,922

95

2,016

389

197

2,602

126.45

1992

413

528

17

510

613

524

365

2,077

91

2,168

381

192

2,741

111.20

1993

443

634

14

619

635

111 114

610

412

2,454

91

2,545

399

213

3,157

101.81

1994

570

748

10

738

724

1995

735

1,158

8

1,151

824

93

731

445

3,163

97

3,260

440

260

3,959

93.90

CAGR (%)

1990-1995

11.4

12.9

-22.7

13.8

3.0

-14.9

7.8

2.6

7.9

-2.8

7.5

0.6

3.8

6.3

Table A-2

Japanese Semiconductor Market, Five-Year Yen Revenue Forecast, 1995-2000 (Revenue in Billions of Yen)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Artalog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

Yen/U.S.$ Exchange Rate

Source: Dataquest (May 1996)

1995

735

1,158

8

1,151

824

93

731

445

3,163

97

3,260

440

260

3,959

93.90

88

868-

514

3,501

112

3,613

504

1996

916

1,114

7

1,108

956

299

4,415

107.05

1997

1,084

1,269

6

1,262

1,034

68

966

519

3,906

115

4,022

519

308

4,848

107.05

1998

1,285

1,443

6

1,437

1,169

58

1,110

571

4,467

115

4,582

560

332

5,475

107.05

1,286

634

5 3 3 9

115

5,454

622

370

6,446

107.05

1999

1,529

1,837

5

1,832

1338

52

2000

1,813

2,372

5

2,368

1,518

47

1,471

708

6,411

115

6,526

711

438

7,675

107.05

CAGR (%)

1995-2000

19.8

15.4

-9.0

15.5

13.0

-12.8

15.0

9.7

15.2

3.5

14.9

10.1

11.0

14.2

SEMM-WW-l\/iT-9601 ©1996 Dataquest

May 13,1996

Japanese Revenue History and Forecast in Yen

37

Table A-3

Japanese Semiconductor Market, Yen Revenue Growth, 1990-1995 (Percentage Revenue

Growth in Yen)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

U.S.$ Appreciation versus Yen

Source: Dataquest (May 1996)

8.9

3.9

-2.2

-3.7

-2.2

0.6

0.5

-1.6

4.35

1990

16.6

-21.3

6.0

-22.2

7.2

3.2

7.3

3.2

4.7

3.3

9.2

13.0

4.9

-5.56

1991

13.7

-5.5

-19.7

-4.8

2.5

-16.3

10.2

-26.0

-12.5

-12.8

-13.9

-18.9

-14.2

1992

-15.1

-11.6

-22.2

-11.2

-15.7

-16.6

-19.0

-14.9

-7.02

-3.9

7.5

-2.2

-2.3

5.3

-12.06

1993

7.3

20.0

-19.1

21.3

3.6

-13.4

8.1

-0.7

8.1

1994

28.7

18.0

-29.4

19.1

14.0

2.3

16.5

13.1

18.2

-0.7

17.4

4.7

11.1

15.2

-8.44

1995

28.9

54.9

-22.8

56.0

13.8

-18.5

19.8

8.1

28.9

7.3

28.1

10.2

21.7

25.4

7.77

CAGR (%)

1990-1995

11.4

12.9

-22.7

13.8

3.0

-14.9

7.8

2.6

7.9

-2.8

7.5

0.6

3.8

6.3

-8.20

Table A-4

Japanese Semiconductor Market, Forecast Five-Year Yen Revenue Growtti, 1995-2000

(Percentage Revenue Growth in Yen)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Arialog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

1995

28.9

54.9

-22.8

56.0

13.8

-18.5

19.8

8.1

28.9

7.3

28.1

10.2

1996

24.6

-3.8

-12.4

-3.8

16.1

-4.9

18.8

15.4

10.7

15.2

10.8

14.7

11.3

0.9

11.6

2.9

11.3

2.9

1997

18.4

13.9

-4.8

14.0

8.2

-22.7

14.9

10.0

14.4

0

13.9

8.0

1998

18.5

13.7

-8.3

13.8

13.0

-14.3

15.8

11.1

19.5

0

19.0

11.1

1999

19.0

27.3

-10.9

27.5

14.5

-11.0

2000

18.5

29.1

-8.2

29.2

13.4

-10.1

14.4

11.7

20.1

0

19.7

14.2

CAGR (%)

1995-2000

19.8

15.4

-9.0

15.5

13.0

-12.8

15.0

9.7

15.2

3.5

14.9

10.1

Optical Semiconductors

Total Semiconductor

U.S.$ Appreciation versus Yen

Source: Dataquest (May 1996)

21.7

25.4

-7.77

14.9

11.5

14.00

3.0

9.8

0

8.0

12.9

0

11.2

17.7

0

18.5

19.1

0

11.0

14.2

2.66

SEMI\/I-WW-MT-9601 ©1996 Dataquest May 13,1996

I

I

Appendix B

European Revenue History and Forecast in ECU

Revenue growth in shipments to the European region differs whether the dollar or ECU is used as the currency basis. The dollar has not had any consistent long-term change with the ECU; the exchange rate in 1995 was essentially the same as in 1990, although there were armual fluctuations.

Figure B-1 shows the annual growth in each of these two currencies over both the historical 1988-through-1995 period and the forecast 1996through-2000 period. Because Dataquest does not forecast exchange rates, the forecast growth rates are the same.

Tables B-1 and B-2 provide the European revenue history and forecast in

ECU. The historical exchange rates are shown at the bottom of these tables. Tables B-3 and B-4 show the annual growth associated with the year-to-year revenue growth. The rate of dollar appreciation against the

ECU for the period from 1990 through 1995 is shown at the bottom of

Table B-3. Over the past seven years, the exchange rate has shown Uttie fluctuation, on average.

Figure B-1

Comparison of Revenue Shipment Growth in European RegionDollars versus ECU

Annual Growth (%)

50

40

30

20

Growth in Do!!ars

Growth in ECU

10

0

-10

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

962914

Source: Dataquest (May 1996)

I

SEMM-WW-MT-9601 ©1996 Dataquest

39

40

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table B-1

European Semiconductor Market, Six-Year ECU Revenue History, 1990-1995 (Revenue in

Millions of ECU)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

ECU/U.S.$ Exchange Rate

Source: Dataquest (May 1996)

1990

1,420

1,659

43

1,615

1,483

402

1,081

1,709

6,271

124

6395

1,493

319

8,207

0.788

1991

1,689

1,761

35

1,727

1,691

359

1,332

1,771

6,912

144

7,057

1,483

393

8,932

0.811

1,732

7,551

116

7,668

1,406

334

9,408

0.77

1992

2,097

2,077

29

2,048

1,645

299

1,347

11,273

154

11,427

1,518

321

13,266

0.858

1993

3,464

3,489

23

3,466

1,973

311

1,661

2347

2,831

15,153

150

15,302

1,771

483

17,556

0.84

1994

4 3 4 3

5,546

24

5,522

2,234

276

1,957

1995

5,425

7,732

15

7,718

2,510

225

2,285

3,194

18,862

185

19,047

2,363

526

21,936

0.774

CAGR (%)

1990-1995

30.7

36.1

-19.4

36.7

11.1

-10.9

16.1

1 3 3

24.6

8.4

24.4

9.6

1 0 3

21.7

Table B-2

European Semiconductor Market, Five-Year ECU Revenue Forecast, 1995-2000 (Revenue in MilUons of ECU)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Aiwlog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

ECU/U.S.$ Exchange Rate

Source: Dataquest (May 1996)

1995

5,425

7,732

15

7,718

2310

225

2,285

3,194

18,862

185

19,047

2363

526

21,936

0.774

1996

6,887

8360

11

8,349

2,933

203

2,730

3,750

21,931

202

22,132

2,727

638

25,498

0.81

1997

7,918

9,689

11

9,679

3,219

164

3,055

4,298

25,124

198

25,322

2,918

704

28,945

0.81

29,056

201

29,256

3,228

790

33,274

0.81

1998

9,206

11,273

8

11,265

3,677

141

3,536

4,900

1999

10398

13350

7

13,542

4,253

122

4,130

5,791

34,491

209

34,700

3,638

893

39,231

0.81

2000

12,504

16,134

6

16,128

5,002

109

4,893

6,950

40,590

213

40,803

4,194

1,034

46,031

0.81

CAGR (%)

1995-2000

18.2

1 5 3

-15.1

15.9

1 4 3

-13.6

1 6 3

1 6 3

16.6

2.9

16.5

12.2

14.5

16.0

SEMM-WW-MT-9601

©1996 Dataquest

May 13,1996

European Revenue History and Forecast in ECU 41

Table B-3

European Semiconductor Market, Historic Revenue Growth, 1990-1995 (Revenue

Growth in ECU)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

U.S.$ Appreciation versus ECU

Source: Dataquest (May 1996)

-20.4

-14.5

21.0

-7.1

0.2

-7.0

4.5

-0.7

-4.8

-13.22

1990

8.5

-26.6

-32.8

-26.4

-16.2

1991

18.9

6.2

-19.5

6.9

14.0

-10.6

23.2

3.6

10.2

16.7

10.4

-0.7

23.2

8.8

2.92

1992

24.2

17.9

-16.1

18.6

-2.7

-16.8

1.1

-2.2

9.2

-19.5

8.7

-5.2

-15.0

5.3

-5.06

4.2

23.3

35.6

49.3

32.1

49.0

1993

65.2

68.0

-20.8

69.2

19.9

8.0

-4.0

41.0

11.43

1994

31.2

58.9

1.5

59.3

13.2

-11.3

17.8

20.6

34.4

. -2.6

33.9

16.7

50.5

32.3

-2.10

1995

19.4

39.4

-37.5

39.8

12.4

-18.5

16.7

12.8

24.5

23.7

24.5

33.4

9.0

24.9

-7.86

CAGR (%)

1990-1995

30.7

36.1

-19.4

36.7

11.1

-10.9

16.1

13.3

24.6

8.4

24.4

9.6

10.5

21.7

-0.36

Table B-4

European Semiconductor Market, Forecast Five-Year ECU Revenue Growth, 1995-2000

(Percentage Revenue Growth in ECU)

Microcomponents

Memory Total

Bipolar Memory

MOS Memory

Logic/ASIC Total

Bipolar Logic

MOS Logic

Analog ICs

Monolithic IC Total

Hybrid ICs

Total ICs

Discrete Devices

Optical Semiconductors

Total Semiconductor

U.S.$ Appreciation versus ECU

Source: Dataquest (May 1996)

1997

15.0

15.9

0

15.9

9.7

-19.1

11.9

14.6

14.6

-1.6

14.4

7.0

10.3

13.5

0

17.4

16.3

9.0

16.2

15.4

21.3

16.2

4.65

1996

27.0

8.1

-28.4

8.2

16.8

-9.7

19.5

-18.5

16.7

12.8

24.5

23.7

24.5

33.4

9.0

24.9

-7.86

1995

19.4

39.4

-37.5

39.8

12.4

1998

16.3

16.3

-23.1

16.4

14.2

-14.3

15.8

14.0

15.6

1.2

15.5

10.6

12.2

15.0

0

1999

18.4

20.2

-10.0

20.2

15.6

-13.2

16.8

18.2

18.7

4:o

18.6

12.7

13.1

17.9

0

18.5

20.0

17.7

1.9

17.6

15.3

15.7

17.3

0

2000

14.7

19.1

-11.1

19.1

17.6

-11.3

CAGR (%)

1995-2000

18.2

15.8

-15.1

15.9

14.8

-13.6

16.5

16.8

16.6

2.9

16.5

12.2

14.5

16.0

0.91

SEI\/II\/1-WW-I\/IT-9601

©1996 Dataquest May 13,1996

^

I

Appendix C

Definitions,

Analog ICs

Analog ICs are a group of semiconductors that deal with electrical signals and electrical power. Analog components carry information as voltage, current, frequency, phase, duty cycle, or other electronic parameters.

Because they are not based on number values, analog information is not limited to a finite range of values and has no irrherent quantization noise or quantization error. The downside is that analog signal information exists in the time domain and can be corrupted as the information-carrying parameter is influenced by noise, drift, bandwidth, and component instability—all the vagaries of time.

Bipolar

These are semiconductor devices that use bipolar transistors rather than

MOS transistors. Bipolar transistors are found in both ICs and discrete products. Bipolar transistors are so named because they carry electricity with two different types of "carriers"—holes and electrons.

Digital ICs

Discrete Devices

A discrete semiconductor is defined as a single semiconductor component such as a transistor, diode, or thyristor. Although multiple devices may be present in a package, they are still considered discretes if they have no internal functional interconnection and are applied in the same manner as other discrete devices. Some discrete devices may actually be similar to

ICs in having integrated protection and sensing circuitry. Even if a device is an integrated circuit, it will be considered a discrete if it is used like one.

Hybrid IC

Digital ICs handle numbers in the binary format of ones and zeros. Digital

ICs comprise logic, microcomponent, and memory ICs. The number-handling nature of digital electronics makes the data more immune to physical changes in the electronic components.

A hybrid is an IC that mixes semiconductor technology with other electronic technologies in a single package. It is this mixing of technologies within the IC package that gives these products the "hybrid" IC name.

Other technologies include thin and thick film resistors and chip capacitors. A multiple-chip IC is not a true hybrid IC and is counted in the monolithic IC category. The mixing of technologies is most often done for analog hybrid ICs. Because of this, hybrid ICs are often added to monolithic analog IC revenue to provide the total analog IC market.

I

43

SEMM-WW-MT-9601 ©1996Dataquest

44

Semiconductor Equipment, iVIanufacturing, and Materials Woridwide

IC

An integrated circuit is a chip in which multiple transistors and diodes are intercoruiected to perform an electronic function. The function-specific nature of an IC differentiates it from the nonspecific array of discrete transistors.

Logic

This is an electronic function where bits (one and zeros) are processed.

This bit processing is defined by hardwiring, mask programming, or field programming. Microcomponents and memory ICs are logic ICs, but they are logic ICs that are either dedicated to a function (such as microperipherals and memory ICs) or are software programmable (such as microprocessors and microcontrollers). Logic ICs also include customer-specific logic

ICs.

Microcomponent

A microcomponent is a digital IC that can be programmable such as a microprocessor (MPU), microcontroller (MCU), digital signal processor

(DSP), or an application-specific logic device that provides a supporting function to an MPU, MCU, or DSR

Monolithic IC

A monolithic IC is an IC formed on a single chip of semiconducting material. This designation has been applied more broadly to mean any device, even a multiple-chip packaged device, that does not contain other, nonsemiconductor, components. This differentiates monolithic ICs from hybrid ICs that may also be multiple-chip, but represent a "hybrid" in the sense of mixing other technologies within the IC package, such as film resistors or chip capacitors.

MOS

MOS is an acronym for metal oxide semiconductor, a type of transistor used in ICs and discrete devices. Although the actual device may use different materials than metal or oxide, this acron5ntn is used to define the whole family of similar processes that provide an insulated gate fieldeffect transistor (FET). MOSFETs, like all field-effect transistors, differ from bipolar devices in having an insulated gate and only a single carrier of electrical current (either electrons or holes). MOSFETs are found in both

N and P channel varieties. A special IC process combines both the N and P channel device in a complementary configuration, an arrangement known as CMOS.

Memory IC

Memory ICs are ICs that can store and retrieve logic bits. Two major memory types are read-only memories (ROM), preloaded with data, or random-access memories (RAM), where data can be both stored and accessed. RAM subcategories include DRAM and SRAM. Memory ICs that do not lose their data when power is removed are called nonvolatile

SEIVIIVI-WW-I\/1T-9601

©1996 Dataqu est iVlay13,1996

Definitions 45

I

Optical Semiconductors

These devices are the semiconductor subset of optoelectronic products.

This family includes light-sensing products such as photosensors and

CCDs as well as light-emitting devices such as LEDs and lasers.

Optocouplers and interrupters use both functions.

Semiconductors

memories. DRAM and SRAM do not retain data when power is removed from the device. ROM, EPROM, EEPROM, and flash memory ICs are nonvolatile memory devices.

These electronic components are manufactured by introducing impurities into a semiconductor material to create special current conducting devices such as diodes, bipolar transistors, and MOS transistors. Semiconducting material is so named because its conducting capability falls between the range of insulators and metallic conductors.

I

I

SEIVIM-WW-I\/IT-9601 ©1996Dataquest iVlaylS, 1996

I

I

Appendix D

Historical Excliange Rates

Table D-1 shows 10 years of exchange rates of the yen and ECU versus the

U.S. dollar. The appreciation of the dollar against these local currencies is given in the last two columns.

Table D-1

Exchange Rates

1985

1986

1987

1988

1989

1990

1991

1992

Year

1980

1981

1982

1983

1984

1993

1994

1995

Source: Dataquest (May 1996)

Yen per U.S.$

227

221

248

235

237

238

167

144

130

138

144

136

126.5

111.2

101.8

93.90

ECU per U.S.$

-

-

-

-

-

-

0.846

0.908

-

-

0.788

0.811

0.770

0.858

0.840

0.774

U.S.$ Growth versus Yen (%)

3.6

-2.7

12.2

-5.2

0.9

0.4

-29.8

-13.8

-9.7

6.2

4.3

-5.6

-7.0

-12.1

-8.4

-7.8

U.S,$ Growth versus ECU (%)

-

-

-

-

-

-

-2.5

-

-

7.3

-13.2

2.9

-5.0

11.4

-2.1

-7.9

I

SEMM-WW-MT-9601 ©1996 Dataquest

47

For More information...

Gary Grandbois, Vice President/Chief Analyst (408) 468-8251

Internet address [email protected]

Via fax (408) 954-1780

The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible individiials in the subject companies, but is not guaranteed as to accuracy or completeness.

It does not contain material provided to us in confidence by our clients. Reproduction or disclosure in whole or in j j o t O f * \ | I f ^ C f part to other parties shall be made upon the written and express consent of Dataquest.

* ^ * ^ I d V ^ U V ^ i J I ©1996 Dataquest—Reproduction Prohibited

AGartnerGraupConpany Dataquest is a registered trademark of A.C. Nidsen Company i

DATAQUEST WORLDWIDE OFFICES

NORTH AMERICA

Woridwide Headquarters

251 River Oaks Parkway

San Jose, California 95134-1913

United States

Phone: 1-408-468-8000

Facsimile: 1-408-954-1780

East Coast Headquarters

Nine Technology Drive

P.O. Box 5093

Westborough, Massachusetts 01581-5093

United States

Phone: 1-508-871-5555

Facsimile: 1-508-871-6262

Dataquest Global Events

3990 Westerly Place, Suite 100

Newport Beach, California 92660

United States

Phone: l-714r476-9117

Facsimile: 1-714-476-9969

Sales Offices:

Washington, DC (Federal)

New York, NY (Financial)

Dallas, TX

LATIN AMERICA

Research Affiliates and Sales Offices:

Buenos Aires, Argentina

Sao Paulo, Brazil

Santiago, Chile

Mexico City, Mexico

EUROPE

European Headquarters

Tamesis, The Glanty

Egham, Surrey TW20 9AW

United Kingdom

Phone:+44 1784 431 611

Facsimile: +441784 488 980

Dataquest France

Immeuble Defense Bergeres

345, avenue Georges Clfeienceau

TSA40002

92882 - Nanterre CTC Cedex 9

France

Phone: +33 141 35 13 00

Facsimile: +33 1 41 35 13 13

Dataquest Germany

Kronstadter Strasse 9

81677 Munchen

Germany

Phone:+49 89 93 09 090

Facsimile: +49 89 93 03 27 7

Sales Offices:

Brussels, Belgium

Kfar Saba, Israel

Milan, Italy

Randburg, South Africa

Madrid, Spain

JAPAN

Japan Headquarters

Aobadai Hills 4-7-7

Aobadcii

Meguro-ku, Tokyo 153

Japan

Phone: 81-3-3481-3670

Facsimile: 81-3-3481-3644

ASIA/PACIFIC

Asia/Pacific Headquarters

7/F China Underwriters Centre

88 Gloucester Road

Wan Chai

Hong Kong

Phone: 852-2824r6168

FacsimUe: 852-2824-6138

Dataquest Korea

Suite 2407, Trade Tower

159 Samsung-dong, Kangnam-gu

Seoul 135-729

Korea

Phone: 822-551-1331

Facsimile: 822-551-1330

Dataquest Taiwan

llF-2, No. 188, Section 5

Nan King East Road

Taipei

Taiwan, R.O.C.

Phone: 8862-756-0389

Facsimile: 8862-756-2663

Dataquest Singapore

105 Cedl Street

#06-01/02 The Octagon

Singapore 069534

Phone: 65-227-1213

Facsimile: 65-227-4607

Dataquest Thailand

12/F, Vanissa Building

29 Soi Chidlom

Ploenchit Road

Patumwan, Bangkok 10330

Thailand

Phone: 662-655-0577

Facsimile: 662-655-0576

Research Affiliates and Sales Offices:

Melbourne, Australia

Beijing, China

Qty:

DataQuest

A Gartner Group Company

©19% Dataquest

SEMM-WW-MT-9601

Ms. Maria Valenzuela

Dataquest

—Internal Dist— USA SJ

FILE COPY:

MARIA VALENZUELA

DataQuest

Year-End 1996 Forecast: Capital

Spending, Wafer Fab Equipment, and

Silicon Markets

Market Trends

Program: Semiconductor Equipment, IVIanufacturing, and l\/laterials Worldwide

Product Code: SEMM-WW-MT-9603

Publication Date: March 17,1997

Filing: Market Trends

Year-End 1996 Forecast: Capital

Spending, Wafer Fab Equipment, and

Markets

Market Trends

Program: Semiconductor Equipment, IVIanufacturing, and IVIaterials Worldwide

Product Code: SEMM-WW-MT-9603

Publication Date: March 17,1997

Filing: Market Trends

Year-End 1996 Forecast: Capital Spending, Wafer Fab Equipment, and Silicon Markets

Table of Contents

Page

1. Executive Summary Year-End 1996: Searching for Rays of

Sunshine in the Storm 1

Wafer Fab Equipment Industry Backlog Status and Forecast 1

Issues of Concern to Reconcile in 1997 and 1998 1

Too Many Fabs Starting and Spending Ratio Too High in 1997... 1

Silicon Consumption in the DRAM Sector: Overcapacity

Exists Well into 1998 2

The Equivalent of 40 Idle Fabs Has Been Created in

Asia/Pacific and Japan 2

Silicon Wafer Forecast Overview 3

Dataquest Perspective: Be Prepared to Batten Down the Hatches.. 4

2. Semiconductor Capital Spending Forecast 5

Chapter Highlights 5

Capital Spending Methodology and Tables 6

1996 Was a Very Dynamic Year 9

How Will 1997 Look? 10

Why Are We Concerned about This Downturn Lasting into 1998? 11

Capital Spending Ratio and Planned Fabs Too High in 1997 12

Silicon Consumption in the DRAM Market Declining! 12

Foundry Industry, Proxy for Logic Capacity, in

Widespread Oversupply 13

Two Possible Scenarios 14

Over/Underinvestment Model Also Supports a Late 1998

Recovery Scenario 14

The Americas Market Will Exhibit Strategic Strength

Long Term 14

Japan: DRAM Capacity Additions Stop, Investment in

Technology Under Way 16

Europe Sustains Presence as a Growth Market 17

Asia/Pacific Investments Focusing on Foundry Near Term as the DRAM Falls 17

Who's Investing Where? 18

Dataquest Perspective 20

3. Wafer Fab Equipment Forecast 23

Chapter and Market Highlights 23

Annual Investment Themes for 1996 through 2001 29

When Will Demand Expand to Meet Capacity? An Update to the Over- or Underinvestment Model 29

Fundamental Capacity Analyses and Annual Review—DRAM and Foundry 32

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table of Contents (Continued)

Page

Highlights of Key Equipment Segment Markets and Forecasts 33

Steppers and Track 33

Etch and Clean: Dry Etch and Chemical Mechanical

Polishing (CMP) 34

Deposition: CVD, PVD, and Silicon Epitaxy .35

Thermal Nondeposition Processes: Diffusion and RTP 36

Ion Implantation 36

Segments That Will Fare Best the Next Two Years 37

Dataquest Perspective 38

4. Silicon Wafer Forecast 41

Silicon Forecast Tables 41

Silicon Wafer Revenue Forecast 41

Silicon Unit Forecast Overview 50

The 200mm Wafer Ramps Up—^Suppliers Have Responded with Capacity 51

What about 300mm Wafers? 52

Epitaxial Wafer Trends: Are DRAMs in the Future? 52

Update of Historical Merchant Epitaxial Market Estimates 53

Highlights of the North American Silicon Wafer Market and

Forecast 53

Highlights of the Japanese Silicon Wafer Market and Forecast 54

Highlights of the European Silicon Wafer Market and Forecast.... 54

Highlights of the Asia/Pacific Silicon Wafer Market and

Forecast 55

Dataquest Perspective 55

5. Semiconductor Consumption Forecast 57

Semiconductor Consumption 57

6. Semiconductor Production Forecast 59

Historical Semiconductor Production 59

Captive Semiconductor Production 59

The Move toward Asia Continues after a Pause, European

Growth Rests 61

Semiconductor Production Trends: Accelerating Shift to

Asia/Pacific 62

Dataquest Perspective 62

Appendix A—Macroeconomic Outlook: Fourth Quarter 1996 65

World Outlook: Global Economy at Dawn of New Era 65

Americas: U.S. Economy ... Too Hot? Too Cold? Just Right? 66

Europe, Middle East, and Africa: Resurgent Germany

Shepherding EU toward Common Currency 67

Japan and Asia/Pacific: Japan's Recovery Slowing,

Four Tigers Struggling 68

Appendix B—Exchange Rates 73

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Year-End 1996 Forecast: Capital Spending, Wafer Fab Equipment, and Silicon Markets

List of Figures

Figures Page

3-1 Net Cumulative Over- and Underinvestment of

Semiconductor Wafer Fab Equipment 31

3-2 Net Cumulative Over- and Underinvestment of

Semiconductor Wafer Fab Equipment 31

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

IV

Semiconductor EquipiriEnt, Manufacturing, and Materials Worldwide

List of Tables

Table Page

2-1 Semiconductor Capital Spending—Top 20 Spenders,

Comparison of 1995 and Projected 1996 Worldwide

Capital Spending 7

2-2 Worldwide Capital Spending by Region—Historical

1989-1995; Includes Merchant and Captive Semiconductor

Companies 8

2-3 Worldwide Capital Spending by Region—Forecast,

1995-2001; Includes Merchant and Captive Semiconductor

Companies 8

2-4 Regional Investment Patterns of Semiconductor

Manufacturers in 1995 19

2-5 Regional Investment Patterns of Semiconductor

Manufacturers in 1996 19

3-1 Worldwide Wafer Fab Equipment Market, by

Region—Historical, 1989-1995 25

3-2 Worldwide Wafer Fab Equipment Market, by

Region—Forecast, 1995-2001 25

3-3 Wafer Fab Equipment Revenue by Equipment Segment,

Historical 1989-1995 26

3-4 Wafer Fab Equipment Revenue Forecast by Equipment

Segment, 1995-2001 27

3-5 Annual Driving Forces and Investment Themes for

Wafer Fab Equipment, 1996-2001 29

4-1 Forecast of Captive and Merchant Silicon* and Merchant

Epitaxial Wafers by Region 42

4-2 Forecast Growth Rates of Captive and Merchant Silicon* and Merchant Epitaxial Wafers by Region 43

4-3 Merchant Epitaxial Silicon Demand by Application, 1994-2001 44

4-4 Worldwide Wafer Size Distribution Forecast, 1994-2001 45

4-5 The Americas Wafer Size Distribution Forecast, 1994-2001 46

4-6 Japan Wafer Size Distribution Forecast, 1994-2001 47

4-7 European Wafer Size Distribution Forecast, 1994-2001 48

4-8 Asia/Pacific Wafer Size Distribution Forecast, 1994-2001 49

4-9 Worldwide Merchant Silicon Wafer Revenue Forecast,

1994-2001 Includes Polished, Virgin Test, and

Epitaxial Silicon 50

4-10 Relative Silicon Area Consumed by DRAM, Fourth

Quarter 1995 through 1997 51

5-1 Worldwide Semiconductor Consumption by

Region—Historical; Includes Merchant Semiconductor

Companies Only 57

5-2 Worldwide Semiconductor Consumption by Region,

Merchant Semiconductor Sales Only—Forecast 58

6-1 Worldwide Semiconductor Production by Region—Historical;

Merchant and Captive Semiconductor Company Sales 60

6-2 Worldwide Semiconductor Production by Region—Forecast;

Merchant and Captive Semiconductor Company Sales 63

SEI\/IM-WW-MT-9603

B-1 Exchange Rates per U.S. Dollar 74

Chapter 1

Executive Summary Year-End 1996: Searching for Rays of

Sunshine in the Storm ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^

Overcapacity in the DRAM market has now trickled to most areas in semiconductor manufacturing. The second half of 1996 saw slightly sharper than expected contraction in shipments of wafer fab equipment. We have uncovered a number of facts that give us concern and issues that need to be reconciled during 1997 and 1998. These issues provide the basis for our belief that there is more downside risk than upside potential to our forecast in the next two years, even in the face of a strong end-use demand for semiconductor devices.

Wafer Fab Equipment Industry Backlog Status and Forecast

In the midyear forecast, Dataquest stated the belief that the 1996 shipment levels would be supported by the larger than normal backlogs at the equipment supplier. Our estimate was that the industry was at about a six-month backlog at the time, compared to a normal backlog of four months. It is not news to anyone that bookings for equipment have taken a dramatic fall since June and have put pressure on backlog levels. The industry has now shipped down backlog to more normal levels, and several companies are below normal.

If the fourth quarter 1996 run rate for wafer fab equipment shipments were to be held flat for all of 1997, the resulting decline would calculate to just over 15 percent for 1997. Our forecast of a slightly larger decline of 18 percent suggests we have not quite hit bottom yet, and indeed we are expecting bookings in the middle of 1997 to be disappointing after a number of months of firm orders in the first half. However, in the hope of improved bookings, we expect companies to continue to ship down backlog until a minimum three-month level is reached by the middle of 1997.

At that time, quarterly shipment rates are expected to be reduced to booking levels. We do not expect sustainable bookings growth to support shipment growth until the second quarter of 1998.

Even in the face of strengthening semiconductor demand, Dataquest has a conservative view of the recovery. Why? The answer is that the fundamentals of the overcapacity simply do not allow a large volume spending recovery to occur before mid-1998, even in the mainstream logic sector.

Issues of Concern to Reconcile In 1997 and 1998

Too Many Fabs Starting and Spending Ratio Too High in 1997

It is very normal in a downturn to get a pocket of companies that will stay and continue to invest in the infrastructure in order to position themselves for the next cycle. In 1997, there are expected to be several companies continuing to make strategic investments. Examples of this are the U.S. fabs being built by Taiwan Semiconductor Manufacturing Company, Samsung, and Hyundai; the U.K. fab being built by Siemens; the array of initial

SEI\/ll\/l-WW-MT-9603 ©1997 Dataquest

Semiconductor Equipment, Manufacturing, and Materials Worldwide joint-venture fabs, such as IBM/Toshiba, TwinStar, and Motorola/

Siemens in the United States; Winbond, Powerchip, and Macronix in Taiwan, whose activities are tied to Japanese companies; and Texas Instruments' venture in Thailand and Korea (with Anam). In all, there are still 47 fabs coming on line in 1997, the same as in 1996 and a higher number than

1995. Even though the facilities will likely be equipped with a minimum equipment set and have slower ramp rates, the capacity coming on line in

1997 exceeds incremental demand.

Capital spending as a percentage of semiconductor revenue is also high at

24 percent. Dataquest believes that 35 to 40 fabs per year and a spending ratio of 22 percent represent an equilibrium for the industry.

Silicon Consumption in tlie DRAM Sector: Overcapacity Exists Well into 1998

As Dataquest has illustrated in earlier publications, the industry migration from 4Mb to 16Mb DRAMs would caxise overcapacity even in the face of high bit demand growth. Die size relationships mean that the average

16Mb DRAM has two to three times more bits per square inch than the

4Mb generation. This means that wafer starts should actually decline for a period of two to three quarters as a result of this silicon efficiency.

Indeed, we have performed a quarterly analysis of the square inches consumed by the DRAM market and have calculated that 14 percent less silicon was required in the third quarter 1996, compared to the fourth quarter of 1995, to support DRAM bit demand. We are not expecting to return to fourth quarter 1995 silicon demand levels until second quarter 1997; yet in those six quarters, a lot of capacity will have been added. Current factory utilization rates for DRAM fabs is running around 70 percent, and we would expect utilization to continue falling in 1997, perhaps to a percentage in the low 60s. According to this fundamental analysis, capacity spending in DRAM is not expected to return until late in 1998.

The Equivalent of 40 Idle Fabs Has Been Created in Asia/Pacific and Japan

Probably our largest concern is that we believe this is not just a DRAM

thing anymore. A startling fact became evident in the silicon market during our forecast process.

When we look at the actual consumption of 150mm wafers in Asia/Pacific and Japan together in 1995, the run rate is about 3.1 million wafers per month (37.2 million wafers in 1995). By the end of 1996, the 150mm wafer consumption into Japan and Asia/Pacific had fallen to 2 million wafers per month. Our 1997 forecast calls for a modest recovery to 27.3 million

''' wafers, or about 2.3 million wafers per month. When comparing the 1995 consumption with the 1997 forecast, we see that capacity of about 800,000 wafers a month has been idled, or about the equivalent of 40 fabs! (This assumes fabs with 20,000 wafer starts per month.)

Our rough analysis indicates that about 14 of these "equivalent fabs" are in Korea, with the balance likely in Japan. Let's look at the makeup of these idle fabs, most of which are a result of the shutoff of 4Mb DRAM production.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Executive Summary Year-End 1996: Searching for Rays of Sunshine in the Storm

In Korea, these fabs have been almost entirely redirected to the foundry market, and the result was a 25 percent reduction in prices in the middle of 1996. Both LG Semicon and Samsung have become very aggressive players in this market, and it is very likely that these companies will invest to upgrade the fabs to sub-0.5 micron technology. Our analysis of the foundry market without these extra fabs shows a market in oversupply in

1997 through 1999 by 15 percent to 25 percent. Foundry capacity supply and demand can be used as a proxy for mainstream logic capacity investment. This picture shows no fundamental driving force for renewed capital spending growth.

The fabs in Japan (equivalent to 26) fall into two areas. Japanese companies initially processed 16Mb DRAMs on 150mm wafers, and we estimate that perhaps 10 fabs worth of capacity is now available to be upgraded to run 200mm wafers. Because this will effectively double that square inch capacity of this block at a relatively low cost, the DRAM-driven recovery could be pushed back a bit. The other block of 16 equivalent fabs cannot be upgraded to run 16Mb DRAMs, so these will migrate to lagging processes such as those for microcontrollers, telecommunications chips, mixedsignal ICs, and analog ICs.

Silicon Wafer Forecast Overview

We anticipated that the second half of 1996 would be weaker in silicon demand, based on the migration from 4Mb to 16Mb DRAMs mentioned in earlier forecasts. However, starting about August, the silicon market flat out collapsed, with run rates by the end of 1996 about 20 percent below those of just six months earlier. These d3Tiamics can mostly be explained by the activities in the DRAM market and inventory trends. As a result, the long-term absolute millions of square inches (MSI) of silicon shipment level was reduced about 5 percent compared with our last forecast, with the six-year compound armual growth rate (CAGR) at 11.1 percent.

Theoretically, silicon consumption into the DRAM sector should be recovering now. However, actual shipments from wafer manufacturers will lag by four to six months as inventories are worked down. We do expect recovery in silicon MSI shipment growth by the middle of 1997. Our quarterly model into the DRAM sector suggests run rates by fourth quarter

1997 should be 20 to 30 percent higher than current depressed levels. Our

MSI growth forecast overall for 1997 is just over 7 percent. With end-use semiconductor and electronic equipment demand strengthening, we would expect accelerated growth in silicon consumption starting in the middle of 1997 and returning to double-digit MSI growth rates in 1998.

The silicon market, driven by a strong long-term picture for semiconductor unit demand in general, will grow faster over the next six years than the last six years. As the industry transforms into a 200mm baseline, the outlook for silicon wafer manufacturers becomes brighter. Silicon manufacturers have answered the call for 200mm capacity with significantly increased capital outlays. Activity in 300mm wafer development has accelerated, particularly in Japan with the Selete consortium. The increased visibility of I300I has also contributed to the first 300mm pilot fab armouncement. We still expect only pilot volume activity in 300mm wafers through the year 2000.

SEI\/IM-WW-l\/IT-9603 ©1997 Dataquest IWarch 17,1997

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Sales of merchant epitaxial wafers by the wafer suppliers accounted for

477 MSI in 1995, About 62 percent of these wafers were used for CMOS logic applications, while only 2 percent were used for DRAM products.

The remainder was shipped into the power/discrete device segments. By the year 2000, Dataquest expects that fully 21 percent of merchant epitaxial silicon will be used for DRAMs, primarily driven by two factors. The

CMOS logic application remains dominant at 52 percent. The overall epitaxial wafer market will experience a 19.4 CAGR from 1995 through 2001.

Bataquest Perspective: Be Prepared to Batten Down the Hatches

Dataquest believes that the wafer fab equipment market will be in for sluggish business conditions through the middle of 1998, before a sustainable bookings recovery can start, coming initially from DRAM capacity spending. Fundamentals of capacity supply and demand balance simply do not support a recovery occurring in 1997 for either memory or mainstream logic. The next 18 months are expected to be dominated by strategic investment by IC manufacturers, which includes production location positioning and investment in new technology.

Equipment areas such as chemical mechanical polishing (CMP) and epitaxial reactors will benefit from higher penetration of the technology, and rapid thermal processing (RTP) system suppliers will benefit from the much higher interest in this process in 300mm lines. Both deep-UV lithography and high-voltage implant are expected to benefit from increased penetration of the technology as well; however, suppliers of steppers and implanters will also be impacted from a dramatic falloff in their capacityoriented businesses, namely i-line steppers and volume DRAM purchases.

Our outlook for the silicon wafer market is more optimistic. The end-unit semiconductor demand remains strong, so the silicon wafer manufacturers should lead the wafer fab equipment market by about a year. We do expect recovery in silicon MSI shipment growth by the middle of 1997, with double-digit growth returning in 1998.

The silicon market has become recognized again as being strategic in the semiconductor manufacturing infrastructure. Will this continue? We believe it will, as long as silicon suppliers continue to concentrate on value-add processes and techniques as the equipment manufacturers have done, as well as adequately and smartly plan capacity additions.

Project Manager: Clark Fuhs

Contributing Analysts: Ellie Babaie, Calvin Chang, Ron Domseif, Joe D'Elia,

Mike Glennon, C.S. Kim, Ben Lee, Jason Lin, Takashi Ogawa, Nader Pakdaman,

Beth Sargent, James Seay, George Shiffler, Yoshie Shima, Yoshihiro Shimada,

Yasumoto Shimizu, J.H. Son, and Jerry Yeh

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Chapter 2

Semiconductor Capital Spending Forecast

This chapter presents data on worldwide semiconductor capital spending by region. Capital spending in a region includes spending by all semiconductor producers with plants in that region. Components of capital spending are property, plant, and equipment expenditure for front- and backend semiconductor operations.

Chapter Highlights

This chapter will discuss the following highlights:

• On the heels of booming growth in 1995,1996 was a year that turned from growth to decline, but net was just under 14 percent growth in capital spending. We saw two distinctly different markets in the past year. The first half was marked by many fab announcements, aggressive spending plans, and good bookings levels for equipment. The second half was the opposite: bookings collapsed, expansion delayed, and capital spending budgets cut.

• Our forecast for 1997 is for a 14 percent decline, with a relatively modest growth of 6 percent in 1998. This reflects our belief that industry overcapacity will delay a sustainable capital spending growth pattern until the middle of 1998.

• We would expect supply/demand dynamics to be fully corrected by early 1999, driving a robust resumption of growth in capital spending growing to just under $70 billion in the year 2000, from the just under

$40 billion in 1998.

• Capital spending in the Americas region grew at an industry average 17 percent in 1996, as a mix of companies both cut and accelerated spending. We expect that investment in advanced technology, coupled with the earlier capacity upturn from the advanced logic segment and strategic investments, will stabilize the region's spending later in 1997 and will lead the market's recovery in 1998 to 1999 as it did in 1993. We expect the Americas region to be the second-fastest growing market as foreign multinationals and foundry companies invest in capacity in the

United States.

• Japan's 9 percent increase in capital spending in 1996 was really a 5 percent decline on a yen basis, as Japanese companies were among the first to cut capital investment and retrench. Because of the early cutback in spending, the 1997 capital spending decline is slightly less than the worldwide average, with growth returning in 1998. Lagging investment patterns in Japan are expected to continue throughout the decade.

• Although spending on capacity has essentially stopped in Japan, two other types of investment are likely to be important in Japan now through 1998. First, Japanese companies will invest in any new technology and equipment targeted at the 0.25-micron production arena. Second, Japanese companies will build shells in 1997, initially at very low run rates, as a preparation to ramp when the market turns up.

SEMM-WW-MT-9603 ©1997 Dataquest

Semiconductor Equipment. Manufacturing, and Materials Worldwide

We are expecting Europe to be a reflection of the overall worldwide market because there is a good mix of multinational investment in the region. Europe is currently at equilibrium with the world, with a stable percentage of the demand and production mix. This stability is what has attracted more companies to produce in the region.

The often erratic but sustained semiconductor capital spending growth in the Asia/Pacific region continued at a market-leading pace of 26 percent growth in 1996. However, the tide has turned in the DRAM area, and we are forecasting the Asia/Pacific region to be hit the hardest in

1997, with a 27 decline in capital spending and a flat-to-down 1998 as the foundry industry deals with oversupply. Longer term, we expect

Asia/Pacific to exhibit among the most aggressive growth in capital spending of any region, but much closer to overall market growth rates than in the recent past.

The foundry industry is now a strategic industry rather than simply a tactical one, but it is now also an industry that has become susceptible to the general industry overcapacity, with dramatically falling prices for processed wafers in the second half of 1996.

Capital Spending Metiiodoiogy and Tabies

Dataquest's forecast process has several cornerstones, including semiconductor production by region, a worldwide database of existing and planned fabs, and independent comprehensive surveys of the equipment and semiconductor companies.

The survey results are one input into our several forecasting models, which include analysis of trends in semiconductor production, raw silicon consumption, spending ratios, investment cycles, new fab and expansion activity, stepper-to-DRAM price-per-bit analysis, and semiconductor revenue per square inch.

A list of the top 20 semiconductor capital spending companies and their projected spending in 1996 is presented in Table 2-1, along with indications for 1997. Our surveys are completed in December, when many companies have not yet finalized their budgets for the upcoming year. We cannot, therefore, include very precise figures for 1997. We have provided some guidance for 1997 spending by company in order to provide some account guidance to our clients.

Capital spending details by region are provided in two tables in this chapter. Table 2-2 shows historical semiconductor capital spending by region for 1989 through 1995. Table 2-3 shows the capital spending forecast by region for 1995 through 2001. Yearly exchange rate variations can have a significant effect on the interpretation of the 1989 through 1996 data. For more information about the exchange rates used and their effects, refer to

Appendix B.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

CO

en

@

CO

CO

-vl tu

Table 2-1

Semiconductor Capital Spending—Top 20 Spenders, Comparison of 1995 and Projected 1996 W

Spending (Millions of U.S. Dollars)

1995

Rank

1

3

1996

Rank

1

2

Company

Intel

LG Semicon

1995

Capital

Spending

3,550.0

2,258.1

1996

Capital

Spending

3,400.0

2,747.5

12

5

3

4

Texas Instruments:

Samsung

1,079.3

1,946.6

2,300.0

2,247.9

9

4

5

6

Hyundai

NEC

1,492.0

2,010.1

2,123.8

1,808.7

10

16

6

14

8

7

2

13

7

8

9

10

11

12

13

14

IBM MicroelectroniGS

Siemens AG

Toshiba

Micron Technology

Hitachi

Fujitsu

Motorola

SGS-Thomsciti.

1,150.0

850.0

1,624.1

960.0

1,497.6

1,592.1

2,530.0

1,001.0

1,550.0

1,450.0

1,437.7

1,400.0

1,296.7

1,275.4

1,150.0

1,000.0

17

11

22

18

15

44

15

16

17

18

19

20

Matsushita

Mitsubishi

TSMC

Chartered S^fiftiisqiRdtiEfiE*

Philips

Winbond

Total Top 20 Companies

Total Worldwide Capital Spending

Top 20 Companies' Percentage of Total

846.6

1,118.2

583.9

786.7

959.0

117.8

27,953.0

38,410.8

72.8

962.8

927.6

901.8

872.3

841.0

700.0

30,393.2

43,707.1

69.5

Note: Specific company informaton is not available for Japanese companies^ less in 1997 than in 1996.

However, Dataquest estimates that, as a group, Japanese companies

Source: Dataquest (January 1997)

CO

CO

CU

—1

o

CO

CO

CO en

C 3

CO

@

CO

CO

• v j o

K.

DJ

Table 2-2

Worldwide Capital Spending by Region—Historical 1989-1995; Includes Merchant and Captive

Companies (Millions of U.S. Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Japan (Billions of Yen)

Percentage Growth

Europe, Middle East, and Africa

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Oataquest (January 1997]

1989

3,833

15.5

5,415

21.6

748

29.2

1,198

26.0

1,884

84.1

12331

26.5

1990

4,320

12.7

5,732

5.9

826

10.4

1,598

33.4

1,580

-16.2

13,230

7.3

1991

3,895

-9.8

5,702

-0.5

787

-4.7

1,248

-21.9

2,300

45.6

13,145

-0.6

1992

4,135

6.2

3,958

-30.6

500

-36.4

1,188

-4.8

2,318

0.8

11,599

-11.8

1993

4,943

19.5

4,413

11.5

491

-2.0

1,738

46.3

3,238

39.7

14,333

23.6

1994

7,194

45.5

6,667

51.1

679

38.3

2,504

44.0

5,720

76.6

22,085

54.1

Table 2-3

Worldwide Capital Spending by Region—Forecast, 1995-2001; Includes Merchant and Captive Se

(Millions of U.S. Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Japan (BilHons of Yen)

Percentage Growth

Europe, Middle East, and Africa

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (January 1997)

1995

12,170

69.2

9,910

48.6

931

37.1

4,137

65.2

12,194

1,13.2

38,411

73.9

1996

14,185

16.6

9362

-5.5

1,009

8.5

4,756

15.0

15,405

26.3

43,707

13.8

1997

13,910

-1.9

8,160

-12.8

917

-9.1

4,228

-11.1

11,205

-27.3

37303

-14.2

1998

15,427

10.9

9,102

11.5

1,023

11.5

4,209

-0.4

11,035

-1.5

39,773

6.1

1999

18,729

21.4

11,723

28.8

1,318

28.8

5363

32.2

14399

3 2 3

50,614

2 7 3

2000

23,442

25.2

15,541

32.6

1,747

32.6

7,195

2 9 3

23312

59.7

69,490

3 7 3

Semiconductor Capital Spending Forecast

1996 Was a Very Dynamic Year

After a 24 percent growth in semiconductor capital spending in 1993, accelerated growth of 54 percent followed in 1994. Growth peaked at 74 percent worldwide during 1995, and 1996 transitioned from growth to decline with a slower 14 percent growth, based on our most recent capital spending survey. Nearly all of this growth occurred in the first half of the year, and spending contraction began in the second half of 1996 and is expected to spill into 1997.

The industry is now relying on the continued growth in personal computer unit sales, with added growth in telecommunications and networking products to create a unit demand picture that will be a healthy backdrop. The wafer fab capacity bubble has burst in all regions and for most semiconductor products, most notably DRAMs, mixed-signal, discrete, and analog. Whereas the 1995 spending growth was almost entirely driven by capacity purchases, 1996 was a year in transition, and 1997 and

1998 will be years of investment in technology.

The first companies to cut back were the U.S. companies, as they tend to be more driven by short-term cost issues. The Japanese companies quickly followed, as the overcapacity in DRAM has caused Japanese companies to quickly tighten the purse strings in hopes of avoiding a more serious price erosion. However, their Korean and Taiwanese counterparts did not fully cooperate until recently. Although Korean companies have already basically announced a 17 to 18 percent cutback in 1997 relative to 1996, the recent move to cut DRAM production significantly means more spending cuts are likely to come.

This above-industry growth for the big three Korean companies has meant that all three are in the top five for capital spending in 1996, and likely to remain among the top five in 1997. As noted earlier, Japanese suppliers of memory cut back investment early in this cycle. Japanese companies as a group will actually spend 6 percent less in 1996 in dollar terms (8 percent growth in yen terms). As a result, only two Japanese companies appear in the list of top 10 capital spenders in 1996—NEC and Toshiba. Most of the other Japanese companies do appear in the second 10. Intel still heads the list for 1996, as microprocessor demand continues to be strong on a unit basis. Intel's capital spending actually declined in 1996, however, primarily because yield ramps on its new fabs have been better than expected, so the company therefore needs less equipment to produce the same unit volume. Motorola, the long-time No. 2 spender, has dropped to No. 13 as the demand for telecom-related chips softened in conjunction with the overcapacity experienced in this area.

A mostly new crowd of Taiwanese companies that entered the DRAM manufacturing business, spending over $1 billion collectively in 1995, increased spending feverishly in the first half of 1996 and likely spent more in the first half than in all of 1995. However, the spending planned for the second half of last year has been predominantly delayed, mostly into 1997, likely later, and in some cases, indefinitely (read late 1998 or

1999).

SEIVllVI-WW-IVIT-geoa ©1997 Dataquest Marcli17,1997

10 Semiconductor Equipment, Manufacturing, and Materials Worldwide

TSMC debuts on the top 20 list for 1996 with an estimated $900 million spent on capacity, as foundry capacity expansion has now evolved into a major trend. TSMC has now joined Chartered Semiconductor on the list, and these two are the largest dedicated foundries. This industry has transformed into a bona fide business and is no longer a specialized way to use excess capacity. There are several companies tiiat have entered the foundry business as a result of today's overcapacity. Unless these companies commit to the foundry business in the long term, their success in the market will be limited. Gone are the days when the "temporary" foundry can exist. Customers of foundry are now requiring long-term relationships and contracts for winning their capacity business.

It is very normal in this type of a downturn to get a pocket of companies that will stay and continue to invest in the infrastructure or have niches that maintain growth, thus supporting an increase in spending. These companies in 1996 include IBM (advanced 16Mb DRAM to support systems), Texas Instruments (primarily DSPs and logic), TSMC (foundry),

Siemens (advanced DRAM), Lucent Technologies (modem chipsets and the Cirrus Logic foundry), and Chartered Semiconductor (foundry). AH of these companies have continued to grow investment during the current year. Micron Technology may be the surprise of 1996 to some, particularly as it has delayed the Lehi, Utah, fab. However, it has been spending aggressively in Boise, Idaho, upgrading the facilities for 200mm production and advanced technology for the 16Mb generation. Micron's spending plans for 1997 have been dramatically curtailed.

Winbond also debuts in the top 20 list, primarily as a result of an alliance to manufacture chips for Toshiba. This is an example of a "strategic" investment pattern that we believe will be dominant, particularly in the first half of 1997.

With the cutback of the big Japanese players in the industry, and with some smaller companies continuing to be aggressive in spending plans, the concentration of capital spending by the top 20 has decreased in 1996 by a few percentage points to just under 70 percent.

How Will 1997 Look?

Six months ago, we believed that the logic sector would be in a position to drive a capacity recovery late in 1997. Developments in the last half of

1996 in Asia/Pacific company strategies and a deeper look into trends in silicon consumption have led us to change this belief. We now believe that, after a firmness in the first half of 1997 as a result of a set of "strategic" investment projects, the overcapacity of the industry will drive the second half of 1997 to be below the first half in capital spending.

Even in the face of the strengthening semiconductor demand currently under way today and the strong end-use electronic equipment and PC markets, Dataquest has a relatively conservative view of the recovery.

Why? The answer is that the fundamentals of the overcapacity we are in today simply do not allow a large volume spending recovery to occur before mid-1998, even in the mainstream logic sector, which has traditionally been the part of the market less affected by the cycle.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Capital Spending Forecast 11

We do not see a stoppage in advanced technology investment, indicating the continued belief in the customer base of a strong end-user market for semiconductors. In some respects, it is this and other "strategic" investments that are pushing out the timing of the recovery in our minds and leading to the current fimmess in orders on the equipment side.

What do we mean by "strategic" investment? After the industry forcefully shut off the valve in the middle of 1996, companies have had time to reevaluate their spending plans. Because the end-use markets for semiconductors remain strong, many companies are investing "strategically," positioning themselves for market share increases and the next major ramp in capacity. Strategic investments include any fabs that open up new production market locations and those that are tied to partnerships and joint ventures. Of particular interest are those companies in a new alliance, where there are factors that are more important than the capacity being added, such as a strategic development or supply relationship.

There are at least 20 fabs starting or upgrading in 1997 that fall specifically into these categories, including the following:

• The U.S. fabs being built by TSMC, Samsung, and Hyundai

• The U.K. fab being built by Siemens

• The array of initial joint venture fabs such as IBM/Toshiba, TwinStar, and Motorola/Siemens in the United States

• Winbond, Powerchip, and Macronix in Taiwan, whose activities are tied to Japanese companies

• Mosel Vitelic, which is tied to Siemens

• Texas Instruments' ventures in Thailand and Korea (with Anam)

These are fabs that will not be stopped but will likely have slower ramp rates than plarmed. Also, the Korean companies are upgrading their older

4Mb DRAM fabs in order to run more advanced logic for the foundry market. A recent order announcement from ASM Lithography for deep-

UV steppers from a Korean company is evidence of this activity.

Does the recent cancellation of a new joint-venture fab of Texas Instruments and Toyota signal the start of a trend that will lead to a worse 1997 than we expect? Only time will tell; however, there is more downside risk than upside potential, in our view, over the next two years. We will now review our concerns in the market to support this belief.

Why Are We Concerned about This Downturn Lasting into 1998?

Our longer-term forecast projects this contraction to be sharp and relatively deep, but lasting about an average length by historical norms—at least two years. Despite the strengthening demand in the semiconductor market, the overspending in the 18 months ending in the middle of 1996 represents a significant "lead balloon" that will drag and delay the spending recovery into later in 1998.

SEIWIVi-WW-MT-9603 ©1997 Dataquest l\^arch 17,1997

12 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Overcapacity in the DRAM market has now trickled to most areas in semiconductor manufacturing. We have also uncovered a number of facts that give us concern and issues that need to be reconciled during 1997 and

1998. These issues provide the basis for our belief that there is more downside risk than upside potential to our forecast in the next two years, even in the face of a strong end-use demand for semiconductor devices. The key point to remember is that the industry needs a capacity driving force to resume sustainable growth prospects.

We have the following three concerns which we will expand on in this section:

• The number of fabs and spending ratio currently factored into in our forecast for 1997

• The consumption of silicon into the DRAM sector declining in the face of increasing bit demand

• Existing excess capacity, which we have now about quantified, being upgraded and redirected

Capital Spending Ratio and Planned Fabs Too High in 1997

With all the "strategic" investment happening in the world, 1997 is simply turning out to be looking "too good," if there is such a thing in this environment. In all, there are still just under 50 fabs that have been announced to come on line in 1997. Although this is significantly lower than the number revealed by a spot check of plans six months ago, this is the same number that started in 1996 and a higher number than the 40 or so in 1995.

Even though the facilities will likely be equipped with a minimum equipment set and have slower ramp rates, the capacity coming on line in 1997 exceeds incremental demand. We estimate that silicon area demand, excluding test and monitor wafers, will grow about 4 percent in 1997, while we are expecting about 8 to 10 percent to be added to the capacity base during the year. For reference, about 13 to 14 percent more square inches of silicon capacity was added in total in 1996, compared to an incremental demand of 4 percent.

Capital spending as a percentage of semiconductor revenue, which is a significant figure of merit, is also high at 24 percent in 1997, after peaking at above 31 percent in 1996. We believe that to support long-term growth in the chip industry, about 35 to 38 fabs per year and a spending ratio of 22 percent represents an equilibrium level. This means that 1998 and 1999 are likely to be years under these equilibrium levels—our current forecast supports this. We would also expect that fab delays and outright cancellations will continue through 1997.

Silicon Consumption in the DRAM Market Declining!

As we have mentioned in past publications, the industry migration from

4Mb to 16Mb DRAMs would cause overcapacity even in the face a high bit demand growth. Die size relationships mean that the average 16Mb

DRAM has two to three times more bits per square inch than the 4Mb generation. This means that wafer starts should actually decline for a period of two to three quarters as a result of this silicon efficiency. Indeed, we

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Capital Spending Forecast 13 have performed a quarterly analysis of the square inches consumed by the

DRAM market and have calculated that 14 percent less silicon is required in the third quarter of 1996 compared to the fourth quarter of 1995 to support DRAM bit demand (see Table 4-10 in Chapter 4).

We are not expected to return to fourth quarter 1995 silicon demand levels until second quarter 1997, yet in those six quarters, a lot of capacity will have been added. Current factory utilization rates for DRAM fabs are running at about 70 percent, and we would expect utilization to continue falling in 1997, perhaps to the low 60s. According to this fundamental analysis, capacity spending in DRAM is not expected to return until late in

1998.

Foundry Industry, Proxy for Logic Capacity, in Widespread

Oversupply

Probably our largest concern is that we believe this is not just a DRAM

thing anymore. A startling fact became evident about the silicon market during our forecast process.

When we look at the actual consumption of 150mm wafers in Asia/Pacific and Japan together in 1995, the run rate is about 3.1 million wafers per month (37.2 million wafers in 1995). By the end of 1996, the 150mm wafer consumption into Japan and Asia/Pacific had fallen to 2 million wafers per month. Our 1997 forecast calls for a modest recovery to 27.3 million wafers, or about 2.3 million wafers per month. When comparing the 1995 consumption with the 1997 forecast, we see that about 800,000 wafers a month capacity has been idled, or about the equivalent of 40 fabs! (This assumes fabs with 20,000 wafers start per month.)

Our rough analysis indicates that about 14 of these "equivalent fabs" are in

Korea, with the balance likely in Japan. Let's look at the makeup of these idle fabs, most of which are a result of the shutoff of 4Mb DRAM production.

In Korea, these fabs have been almost entirely redirected to the fotmdry market, and the result has been a 25 percent reduction in prices during the middle of 1996. Both LG Semicon and Samsung have become very aggressive players in this market, and it is very likely that these companies will invest to upgrade the fabs to sub-0.5 micron technology. Our analysis of the foundry market without these extra fabs shows a market in oversupply in 1997 through 1999 by 15 to 25 percent. With the Korean capacity, the oversupply is much more severe, so much so in fact that we expect that many foundry suppliers will be forced to cut spending by late in 1997 in response.

Foundry capacity supply and demand can be used as a proxy for mainstream logic capacity investment. This picture shows no fundamental driving force for renewed capital spending growth.

The fabs in Japan (equivalent to 26) fall into two areas. Japanese companies initially processed 16Mb DRAMs on 150mm wafers, and we estimate that perhaps 10 fabs worth of capacity is now available to be upgraded to run 200mm wafers. Because this will effectively double that square inch

SEiViM-WW-MT-geoa ©1997 Oataquest IVIarch 17,1997

14 Semiconductor Equipment, Manufacturing, and Materials Worldwide capacity of this block at a relatively low cost, the DRAM-driven recovery could be pushed back a bit. The other block of 16 equivalent fabs cannot be upgraded to run 16Mb DRAMs, so these will migrate to lagging processes such as those for microcontrollers, telecommunications chips, mixed signal ICs, and analog ICs.

Two Possible Scenarios

These three issues together represent a possibly different scenario from our current forecast. One can think of the next two years as being one of two possible types of markets. Semiconductor companies could continue to invest strategically, keeping spending levels for 1997 high enough to not fully correct the excesses. This puts a drag on growth prospects for

1998. If the chip industry were to fully correct for the excesses of 1995 to

1996, the 1997 market could be significantly worse—perhaps down 25 to

30 percent, and recovery could be significant sequential growth in the second half of 1998. The first scenario represents our forecast scenario. The second, more dramatic, scenario represents the downside risk, given the facts of the day.

We believe that capital spending may be influenced in 1998 and 1999 positively with the facility construction and purchase of equipment toward the world's first 300mm wafer fabs. We have built this infrastructure investment in our model.

i

Over/Underinvestment Model Also Supports a Late 1998 Recovery Scenario

A few years ago, we introduced a model that quantifies the over/underin- I vestment picture for wafer fab equipment and semiconductor capacity.

Although the early 1990s created and sustained a net underinvestment, this picture was corrected to create about a 36 percent overinvestment by the end of 1995 (see Chapter 3). Clearly, this was in the danger zone, and we are seeing the results of this overinvestment today. By the end of 1998, should our forecasts for investment and semiconductor demand be on target, we would expect the indusfry to return to being 15 percent net underinvested, just barely within the range to properly set the stage for a robust recovery in 1999.

The Americas Market Will Exhibit Strategic Strength Long Term

Capital spending in the Americas region grew at an industry-average 17 percent level in 1996, as a mix of companies both cut and accelerated spending. We expect that investment in advanced technology, coupled with the earlier capacity upturn from the advanced logic segment and strategic investments, will stabilize the region's spending later in 1997 and will lead the market's recovery in 1998 to 1999 as it did in 1993. We expect the Americas region to be the second fastest-growing market as foreign multinationals and foundry companies invest in capacity in the United

States. Dataquest is forecasting the Americas region growing at a 14.5 percent CAGR for 1995 through 2001, driven by the relative low cost of capital and the need for foreign multinational and foimdry manufacturers to be closer to their customers, building fabs in the United States.

i

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Capital Spending Forecast 15

The relatively strong growth in capital spending had been driven by the ongoing growth in PCs, telecommunications, and networking. This key driver has not disappeared, as these products have seen increasing use as tools to increase the productivity of tiie workplace. Electronic products with increased semiconductor content have created enormous demand for microprocessors, microcontrollers, SRAM, programmable logic and memory, digital signal processors (DSPs), standard logic, and peripheral controllers. The U.S. companies dominate many of these market segments.

These segments combined are expected to maintain fairly stable growth rates over the next few years, with PC growth slowing (however, still maintaining a 17 percent CAGR) and networking and telecommunications expanding. The near-term market for PCs has remained robust worldwide, despite the slower growth and penetration into the U.S. home market.

N e w consumer digital products and services, such as personal communicators, interactive television, DVD systems, digital cameras, and video-ondemand, provide the potential for enormous growth in semiconductor sales longer term, especially for highly integrated complex logic and signal-processing chips that will be the core engines of future systems.

Although the strategic strength of core logic products enables a healthy and flourishing semiconductor production environment, it is also an environment that is less volatile in capital spending. In the boom years of 1994 and 1995, the Americas region grew at somewhat lower than the market rates. This trait will also enable the Americas market to grow in capital spending faster than market rates (or remain more stable) in the slower years, such as 1997 and 1998. We believe companies will strategically invest in technologically advanced capacity to preserve competitive advantage. Our forecast is for an essentially flat capital spending market in 1997 and a low double-digit growth in 1998.

Capital investment trends in the Americas region for 1996 have a definite split personality. While Intel has finished expansion of Fab 11.2 in N e w

Mexico and ramping Fab 12 in Arizona, it has reduced spending because of increased yields. Micron Technology stopped investments in Lehi,

Utah, but continued aggressive expansion and conversion to 200mm at all three fabs in Boise, Idaho. Memory-sensitive plant expansions such as

Fujitsu's Greshem, Oregon, plant and Integrated Device Technology's

Oregon fab have been delayed or have slower ramps, yet IBM has been very aggressive in its Burlington expansion of 16Mb DRAM production.

Logic investment has seen a slowdown as well, with LSI Logic's pushout of its Oregon fab, the delay of Motorola's North Carolina MPU fab, and the slow ramp-up of Advanced Micro Devices' Fab 25 in Austin, Texas.

Yet Cirrus Logic and Lucent Technologies have increased spending dramatically to ramp fabs in the eastern United States, with Atmel expanding aggressively in Colorado and Rockwell's emerging success creating opportunity for equipment companies in California. SGS-Thomson has remained aggressive in spending in the United States, placing finishing touches on its new Arizona facility, as well as starting up Fab 4 in Carrollton, Texas. Texas Instruments remains aggressive, spending in capacity expansion for DSP chip capacity. Although neither Samsung nor TSMC placed equipment into their new U.S. fabs in 1996, capital spending on the shells is progressing.

SEiV!l\/l-WW-l\/IT-9603 ©1997 Dataquest March 17,1997

16 Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

In 1997, we expect Intel and AMD to lead the charge for spending in the

United States. Also, equipment orders are expected to be placed for the first fabs in the United States for Hyundai, Samsung, and TSMC. We would also expect logic-oriented companies with a primary capacity base in the United States to come back into the investment picture, finishing up projects started but left incomplete. We would be cautious about new projects being announced, however, as these companies will likely take advantage of the favorable foundry pricing environment before putting in place rather risky capital investment plans.

Japan: DRAIM Capacity Additions Stop, Investment in Technology Under Way

Japan's 9 percent increase in capital spending in 1996 was really a 5 percent decline on a yen basis, as Japanese companies were among the first to cut capital investment and retrench. Because of the early cutback in spending, the 1997 capital spending decline is slightly less than the worldwide average, with growth returning in 1998. Lagging investment patterns in

Japan is expected to continue throughout the decade.

Some of the Japanese electronics giants that experienced good profit growth in 1995, driven by semiconductor operations, have seen those profits evaporate with the precipitous fall of DRAM prices. While spending on capacity has essentially stopped, two other types of investments are likely to be important in Japan now through 1998.

First, Japanese companies will invest in any new technology and equipment targeted at the 0.25-micron production arena. This technology will not likely be in volume production until 1999, but the Japanese companies are expected to take advantage of this slowdown to understand and progress down the learning curve on these new process technologies.

Second, the Japanese found that the shells built in 1990 and 1991 became an asset during the ramp in 1993 and 1994. By building a fab shell, equipped with a skeleton equipment set, they were positioned to more quickly ramp up production when the market turned up. We see that same pattern repeating, so we would expect several new fabs to be started in 1997, however at very low run rates. Once these fabs are in place, then the Japanese can continue to review the market every six months, making course corrections in April and October, as they have been doing through the last cycle ramp.

Although new facilities by Japanese companies will come on line outside

Japan throughout the rest of this decade, DRAM investments inside Japan are really the only driving force today, although diversification has come to the forefront again in Japan. Japanese companies will continue to invest, but will grow to depend on production outside of Japan faster than within

Japan. We are therefore forecasting a below-average CAGR of 9.9 percent in Japan for the years 1995 through 2001.

One bright spot is that a PC boom could emerge in Japan over the next year or two, spawned by the networking infrastructure that is currently being built. This would breathe new life in the Japanese semiconductor market, and our forecast would be brightened a bit. This PC boom would be seriously in jeopardy if the Japanese economy retracted into recession.

We do not think that even a PC boom, however, would create a forecast

SEMI\/l-WW-l\/IT-9603 ©1997 DataquBSt IVIarch 17,1997

Semiconductor Capital Spending Forecast 17 different from several percentage points below the world average. The fundamentals of a Japanese production capacity are still too heavily concentrated in DRAMs, with no clear future direction emerging as yet, which keeps us from being more optimistic for capital activities in Japan.

Europe Sustains Presence as a Growth Market

After a higher-than-expected growth bubble of 46 percent in 1993, European spending "moderated" to a slower than market growth in 1994 as multinationals (Intel) substantially completed the majority of their expansions in 1993. The growth of 44 percent in 1994 is nonetheless extremely healthy, primarily being fueled by the Europeans themselves, with the ever-present SGS-Thomson, Philips, and Ericsson equipping their expansions. Europe continued to attract the capital in 1995, growing 65 percent.

Large multinationals are still present, with Motorola upgrading the Scotland fab bought from Digital Equipment, the new IBM/Philips venture in

Germany, Analog Devices expanding in Ireland, Texas Instruments expanding again in Italy, and the IBM/Siemens fab in France continuing to ramp 16Mb DRAMs.

Also, the key expansion is Siemens' new fab in Dresden, the key driver pulling Siemens into the top 10 in capital spending worldwide in 1996. As in the United States, Europe experienced slowdowns as well during 1996, with capital spending growing 15 percent. While SGS-Thomson and Siemens remained strong. Philips and the Japanese companies have significantly pulled back investment in capacity. However, partly because we do not believe Siemens can sustain their current spending, we are calling for an 11 percent decline in spending in Europe for 1997, and a flat-to-down

1998 as we expect multinationals will ramp domestic memory fabs before

Europe. All three big Korean companies have now announced plans for fabs in Europe to come on line during 1999, with only Samsung undecided about the exact location.

We are expecting Europe to be a reflection of the overall worldwide market because there is a good mix of multinational investment currently in the region. Europe is currently at equilibrium with the world, with a stable percentage of the demand and production mix. This stability is what has attracted more companies to produce in the region. Europe is viewed as a strategic location for production longer term to take better advantage of

European and 16/64Mb DRAM growth in the future, driven by PC production, without the import tariffs. We therefore expect Europe to be a near-average investment region in the long term, with a six-year CAGR of 12.3 percent.

Asia/Pacific Investments Focusing on Foundry Near Term as the DRAM Falls

The often erratic but sustained semiconductor capital spending growth in the Asia/Pacific region continued at a market-leading pace of 26 percent growth in 1996. However, the tide has turned in the DRAM area, and we are forecasting the Asia/Pacific region to be hit the hardest in 1997, with a

27 percent decline in capital spending and a flat-to-down 1998 as the foundry industry deals with oversupply. Longer term, we expect Asia/

Pacific to exhibit among the most aggressive growth in capital spending of any region, but much closer to overall market growth rates than in the recent past, with a 17.2 percent CAGR for capital spending.

SEI\/ilVI-WW-IWT-9603 ©1997 Dataquest l\/larcli 17,1997

18 Semiconciuctor Equipment, Manufacturing, and Materials Worldwide

Spending in 1995 and early 1996 came primarily from two areas, DRAMs and foundry capacity. The Korean conglomerates are continuing their relentless DRAM capacity expansion plans, although more moderately in

1996. We do expect these companies to finally succumb to the inevitable reality of overcapacity, with significant cutbacks for 1997. The real story of interest in 1995 and 1996 were the new Taiwan players. But the tide has turned quickly, likely accelerated by the fact that the Taiwanese stock market is very close to the U.S. stock market in its reaction to bad news. Many companies in DRAM have been cutting back feverishly to save near-term profitability. Yet many companies in Taiwan are tied strategically to other companies outside the region, lending some investment stability there in the near term.

The foundry industry is now a strategic industry rather than simply a tactical one, but it is now also an industry that has become susceptible to the general industry overcapacity situation, with dramatically falling prices for processed wafers in the second half of 1996.

While Taiwan chip companies TSMC, Macronix, and UMC along with

Chartered semiconductor in Singapore and SubMicron Technology in

Thailand continued with their planned projects in expansion of foundry capacity, the Korean companies entered the foundry market very aggressively in the second half of 1996, causing us to be concerned about plans for 1997 and 1998 for the dedicated foundries.

The reason for the continued interest in spending capital and new entrants in this area comes from the fact that the core business is dependent upon the logic and PC unit demand rather than DRAM. Further, Dataquest estimates that only about 32 percent of the contracted manufacturing of semiconductors originates from fabless companies in 1995, but the trend for this share is increasing. The remainder is from integrated device manufacturers (IDMs) who wish to place manufacturing lower value-added products away from their own facilities in order to maximize resources and cost, or to reduce investment risks using foundries as an extension of their own capacity, or to use the more advanced technology of foundries

(in some cases) as a growth strategy.

Who's Investing Where?

In our capital spending survey recently completed, Dataquest gathered picture of how money is being spent. Table 2-4 summarizes how companies based in different regions are spending their money abroad for 1995, and Table 2-5 summarizes the same for 1996. About 79 percent of money spent went into domestic economies worldwide in 1995, and that ratio increased slightly to 83 percent in 1996 as companies tend to cut back externally first.

Asia/Pacific companies have historically placed all of their investments domestically, but 1994 saw the first year of diversification, which continued in 1995. Asia/Pacific companies spent about 3.6 percent of their money abroad in 1995, and that is increasing to about 4.2 percent in 1996.

We would expect this ratio to increase significantly over the next two to three years. Europeans have been the most aggressive capital exporters, historically, placing only 59 percent of their investment inside of Europe during 1995. This figure has grown slightly to 64 percent in 1996 and should expand in 1997 as European companies reign in spending.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

cu o

CO

CO

CO

CO

O)

@

CO

CO

•vl

K.

cu

Table 2-4

Regional Investment Patterns of Semiconductor Manufacturers in 1995 (Millions of U.S. Dollar

American Companies

Japanese Companies

European Companies

Asia/Pacific Companies

All Companies

Percentage Growtli from 1994

Source: Dataquest (January 1997)

Worldwide

13,863.3

12,042.3

3,301.5

9,203.8

38,410.8

73.9

Americas

10,135.2

1,328.7

420.2

285.5

12,169.5

69.2

Japan

655.1

9,247.7

7.2

0

9,910.0

48.6

Europe, Middle East, and Africa

1,423.3

738.7

1,936.0

38.9

4,136.9

65.2

Asia/

1

Table 2-5

Regional Investment Patterns of Semiconductor Manufacturers in 1996 (Millions of U.S. Dollar

American Companies

Japanese Comparues

European Companies

Asia/Pacific Companies

All Companies

Percentage Growth from 1995

Source: Dataquest (January 1997)

Worldwide

15,657.8

11,268.8

3,802.0

12,978.5

43,707.1

13.8

Americas

12,289.5

1,045.2

359.7

490.0

14,184.5

16.6

Japan

445.3

8,916.5

0

0

9,361.8

-5.5

Europe, Middle Eas^ and Africa

1,503.4

751.0

2,446.2

54.9

4,755.5

15.0

Asia/

1

1

20 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Japanese companies are very close to the worldwide average with about

77 percent domestic investment in 1995, rising to 79 percent in 1996. North

American companies are also high domestic spenders, with about 73 percent staying at home for 1995 increasing to 78 percent in 1996. We would expect both of these groups to fluctuate between 70 and 80 percent long term, depending on market conditions.

The North American and Japanese regions are net investors, while European and Asia/Pacific regions are net beneficiaries of that investment.

This parallels those regions being net exporters and net importers of semiconductors, respectively.

While all regions are spending in Asia/Pacific, and all multinational regions investing in Europe, only North American companies have the strategic vision to invest in Japan. Japanese companies are also investing on a worldwide basis. We believe this is one of the key elements necessary in a strategic plan for a semiconductor company to be competitive on a global basis.

Dataquest Perspective

The capital spending boom experienced in 1993 through the first half of

1996 is over, and the industry is now in what we would characterize as a two-year "pause," with investment in capacity initially declining and stabilizing as demand catches back up to supply.

Six months ago, we believed that the logic sector would be in a position to drive a capacity recovery late in 1997. Developments in the last half of

1996 in Asian company strategies, and a closer look into trends in silicon consumption, have led us to (iiange this belief. We now believe that, after a firmness in the first half of 1997 as a result of a set of "strategic" investment projects, the overcapacity of the industry will drive the second half of 1997 to be below the first half in capital spending. Even in the face of strengthening semiconductor demand currently under way today, and strong end-use electronic equipment and PC markets, Dataquest has a relatively conservative view of the recovery. The fundamentals of the overcapacity situation we are in today simply do not allow a large volume spending recovery to occur before mid-1998, even in the mainstream logic sector, which has traditionally been the part of the market less affected by the cycle.

We do not see a stoppage in advanced technology investments, indicating the continued belief in 9ie customer base of a strong end-user market for semiconductors. In some respects, it is this and other "strategic" investments that are pushing out the timing of the recovery in our mind, and leading to the current firmness in orders on the equipment side.

Overcapacity in the DRAM market has now trickled to most areas in semiconductor manufacturing. We have also uncovered a number of issues that need to be reconciled during 1997 and 1998. These issues provide the basis for our belief that there is more downside risk than upside potential to our forecast in the next two years. The key point to remember is that the industry needs a capacity driving force to resume sustainable growth prospects.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Capital Spending Forecast 21

Our concerns are the following:

• The number of fabs and spending ratio currently factored into in our forecast for 1997 is still too high.

• The consumption of silicon into the DRAM sector has declined in the face of increasing bit demand.

• Existing excess capacity, which we have now about quantified, is being upgraded cind redirected into the foundry market.

The foundry industry is now a strategic industry rather than simply a tactical one, but it is now also an industry that has become susceptible to the general industry overcapacity situation, with dramatically falling prices for processed wafers in the second half of 1996. Foundry capacity supply and demand can be used as a proxy for mainstream logic capacity investment. This picture shows no fundamental driving force for renewed capital spending growth near term.

By the end of 1998, should our forecasts for investment and semiconductor demand be on target, we would expect the industry to return to be net underinvested within the range to properly set the stage for a robust recovery in 1999. We believe fiiat capital spending may be influenced in late 1998 and 1999 positively with the facility construction and purchase of equipment toward the world's first 300mm wafer fabs. We have built this infrastructure investment in our model.

SEI\/IM-WW-IVIT-9603 ©1997 Dataquest l\/larch 17,1997

I Chapter 3

Wafer Fab Equipment Forecast

This chapter presents data on worldwide spending by region for wafer fabrication equipment. Wafer fab equipment spending in a region includes spending by all semiconductor producers with plants in that region. Included are all classifications of equipment for front-end semiconductor operations.

Chapter and Market Highlights

This chapter will address the following points:

• Overcapacity has gripped the industry and 1996 was a year that turned from growth to decline, but net was just under 12 percent growth in wafer fab equipment. We saw two distinctly different markets in the past year. The first half was marked by many fab armouncements, aggressive spending plans, and good booking levels for equipment. The second half was the opposite: bookings collapsed, expansion delayed, and capital spending budgets were cut.

• The only reason 1996 remains a double-digit growth year is due to the strong backlogs coming out of 1995. Our estimate was that the industry was at about a six-month backlog at the time, compared to a normal backlog of four months. The bookings decline has put pressure on backlog levels, and the industry has shipped down backlog now to more normal levels, with several companies below normal.

• The second half of 1996 contained slightly more severe declines than we anticipated six months ago. If the fourth quarter 1996 run rate for wafer fab equipment shipments was held flat for all of 1997, the resulting decline would calculate to just over 15 percent for 1997. Our forecast for

1997 is for an 18 percent decline, which would suggest we believe we have not quite hit bottom yet, and indeed we are expecting bookings in the middle of 1997 to be disappointing.

• However, we anticipate that equipment companies, in the hope of improved bookings later in the year, will continue to ship down backlog in order to maintain a minimum financial performance, maintaining the run rates of fourth quarter 1996. We expect to reach a minimum three-month level backlog by the middle of the year 1997. At that time, with backlogs razor thin, quarterly shipment rates are expected to be reduced to booking levels, leading to a sales dip in the third quarter of this year.

• We do not expect sustainable bookings growth to support shipment growth until the second quarter of 1998, and shipment growth is expected to remain constrained until very late in the year, keeping

1998 as a single-digit growth year.

• We would expect supply/demand dynamics to be fully corrected by early 1999, driving a robust resumption of growth with the wafer fab equipment market growing to over $33 billion in the year 2000, from the iust under $19 billion in 1998. The worldwide wafer fab equipment market is forecast to grow at 13.7 CAGR between 1995 and 2001, essentially equivalent to semiconductor market growth.

SEMM-WW-MT-9603 ©1997 Dataquest 23

24 Semiconductor Equipment, Manufacturing, and Materials Worldwide

• We believe the market in 1997 and early 1998 will concentrate on advanced technology equipment. A purchasing behavior that will be prevalent in the near term will be new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. These segments will be those related primarily to deep-UV lithography, silicon epitaxy, maskmaking equipment, 300mm technologies, and 0.25-micron multilevel metallization process schemes.

• Our model, which measures the net cumulative under- or overinvestment, indicates that by the end of 1995, the semiconductor manufacturing world was be overinvested in wafer fab equipment to the tune of

$6.8 billion, or 36 percent of the market. This is above the peaks exhibited in 1984 and 1989, so it is no surprise that excess capacity has emerged during 1996 in the DRAM market where capacity has been added recently.

• We have factored in an infrastructure investment in equipment for late

1997 through 1999, which will affect the forecast size of the markets positively. This additional investment will be for initial equipment to fill a couple of 300mm fabs to run silicon by 1999. The bulk of this "300mm equipment bubble" occurs in 1998. However, our outlook for a significant 300mm equipment market will wait until 2001 through 2002.

This chapter presents historical and forecast data on the worldwide wafer fabrication equipment market, by region and by key equipment segment.

In this year-end forecast outlook on wafer fab equipment, we have chosen to focus our forecast of equipment categories on specific segments and issues, namely:

• The annual investment theme for 1995 through 2000

• Steppers and automatic photoresist processing equipment (track)

• Dry etch and chemical mechanical polishing (CMP)

• Silicon epitaxy, CVD, and PVD

• Diffusion and RTF

• Ion implantation (medium current, high current, and high voltage)

• Segments that will fare best the next two years

These segments of the equipment market not only represent the majority of all wafer fab equipment expenditure in the world today, but also embody the key technological capability for advanced device production.

Highlights of some of the factors affecting individual equipment segment forecasts also are presented.

Equipment spending in a region refers to spending by all companies— both domestic and foreign—within the region. Also, we note tiiat yearly exchange rate variations can have a significant effect on 1989-through-

1996 data appearing in the tables in this chapter. Appendix B details the exchange rates used in this document.

i i i

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast

SEI\/IM-WW-MT-9603

25

Tables in this chapter provide details on the following:

• Table 3-1: Historical market data, by geographic region, for the years

1989 through 1995

• Table 3-2: Forecast market data, by geographic region, for the years 1995 through 2001

• Table 3-3: Historical data for key equipment segments for the years 1989 through 1995

• Table 3-4: Forecast data for key equipment segment for the years 1995 through 2001

Table 3-1

Worldwide Wafer Fab Equipment Market, by Region-—Historical, 1989-1995

(Millions of U.S. Dollars)

CAGR (%)

1989-1995

21.5 Americas

Percentage Change

Japan

Percentage Change

Europe, Middle East, and Africa

1989

1,657

7.9

2,813

23.9

721

8.7

1990

1,589

-4.1

2,996

6.5

764

Percentage Change

Asia/Pacific 820

6.0

522

-36.4

Percentage Change

Total Wafer Fab Equipment

Percentage Change

58.1

6,011

20.6

5,871

-2.3

Note: Columns may not add to totals shown because of rounding.

Source: Dataquest (January 1997)

1991

1,519

-4.4

3,012

0.5

641

-16.1

832

59.5

6,003

2.3

1992

1,576

3.8

2,098

-30.3

641

0

783

-5.8

5,098

-15.1

1993

2,118

34.4

2,450

16.8

988

54.1

1,312

67.5

6,868

34.7

1994

3,190

50.6

3,661

49.4

1,370

38.6

2,567

95.7

10,787

57.1

1995

5,332

67.1

6,157

68.2

2,313

68.8

5,208

102.9

19,010

76.2

13.9

21.4

36.1

21.2

Table 3-2

Worldwide Wafer Fab Equipment Market, by Region—Forecast, 1995-2001

(Millions of U.S. Dollars)

Americas

Percentage Change

Japan

Percentage Change

Europe, Middle East, and Africa

Percentage Change

Asia/Pacific

1995

5,332

67.1

6,157

68.2

2313

68.8

5,208

1996

6,154

15.4

6,263

1.7

2,623

13.4

6,205

Percentage Change

Total Wafer Fab Equipment

102.9

19,010

19.1

21,245

Percentage Change 76.2 11.8

Note: Columns may not add to totals shown because of rounding.

Source: Dataquest (January 1997)

-30.1

17,439

-17.9

1997

5,728

-6.9

5,129

-18.1

2,243

-14.5

4,339

1998

6,439

12.4

5,614

9.5

2,223

-0.9

4,110

-5.3

18,386

5.4

1999

8,062

25.2

7,253

29.2

2,946

32.5

5,716

39.1

23,976

30.4

2000

10,451

29.6

9,739

34.3

3,862

31.1

9,425

64.9

33,477

39.6

2001

12,197

16.7

10,803

10.9

4,691

21.5

13,298

41.1

40,988

22.4

CAGR (%)

1995-2001

14.8

9.8

12.5

16.9

13.7

©1997 Dataquest March 17,1997

(O

Oi

o

GJ

@

CO

CO

• v j a cu

Dl

—1

o

CO

CO

•vl

i

Table 3-3

Wafer Fab Equipment Revenue by Equipment Segment, Historical 1989-1995 (Millions of U.S. D

Equipment Segment

Worldwide Fab Equipment

Change(%)

Steppers

Track

Maskrtiaking Lithography

Other Lithography^

Total Lithography/Track

Automated Wet Stations

Other Clean Process

Dry Strip

Dry Etch

Chemical Mechanical Polishing

Total Etch and Clean

Tube CVD

Nontube Reactor CVD

Sputtering

SiHcon Epitaxy

Other Deposition^

Total Deposition

Diffusion

RTF

Total Thermal Nondepo^ition

Medium-Current Implant

High-Current Implant

High-Voltage Implant

Total Ion Implantation

Optical Metrology

CD-SEM

Thin Film Measurement

Patterned Wafer Inspection

250

7

370

59

88

NS

105

68

153

1,296

325

33

358

114

118

690

NS

1,208

259

457

359

1990

5,871

-2.3

1,052

317

47

195

1,612

268

132

1989

6,011

20.5

1,181

322

69

192

1,764

243

134

121

670

NS

1,168

220

388

320

75

170

1,173

332

25

357

131

301

25

457

74

81

NS

116

1993

6,868

34.7

1,007

500

52

120

1,679

285

198

138

1,083

44

1,748

283

585

584

83

115

1,650

342

45

388

108

233

18

359

43

83

72

144

1,300

246

36

283

83

164

16

20

1,213

213

437

446

84

119

263

40

78

58

109

1992

5,098

-15.1

646

353

53

106

1,158

286

103

123

682

1,281

268

474

425

89

147

1,403

326

46

372

108

228

18

353

59

93

43

90

1991

6,003

2.2

979

364

48

158

1,549

291

143

119

717

11

213

1,592

64

2,548

442

885

1,012

114

101

1994

10,787

57.1

1,833

695

79

125

2,732

468

213

2,553

490

76

567

242

391

27

659

67

154

100

281

CU

—1

n

CO

CO

CO

o

GO

@

(O

CO

•vl

C3

CU

Table 3-3 (Continued)

Wafer Fab Equipment Revenue by Equipment Segment, Historical 1989-1995 (Millions of U.S. D

Equipment Segment

Auto Unpattemed Detection

Other Process Control^

Total Process Control

Factory Automation

Other Equipment

Total Factory Automation/Other

Equipment

Total Wafer Fab Equipment

1989

37

369

676

195

222

417

1990

45

313

609

216

202

418

6,011 5,871 6,003

NS=Not surveyed

NA^Not applicable

''includes contact/proximity, projection aligners, dired-wiilBB-beam, and X-ray lithography

^Includes evaporation, MOCVD. and MBE

^Includes auto review/classification, manual detection/revisw, and other pracess control equipment

Note: Columns may not add to totals shown because of rounding.

Source: Dataquest (January 1997)

1991

41

307

632

227

185

412

1992

30

228

542

194

146

340

5,098

1993

32

270

644

250

151

401

6,868

1994

56

446

1,103

412

213

625

10,787

Table 3-4

Wafer Fab Equipment Revenue Forecast by Equipment Segment, 1995-2001 (Millions of U.S. Do

Equipment Segment

Worldwide Fab Equipment

Change(%)

Steppers

Track

Maskmaking Lithography

Other Lithography

Total Lithography/Track

Automated Wet Stations

Other Clean Process

Dry Strip

Dry Etch

1995

19,010

76.2

3,332

1,413

82

118

4,944

928

389

343

2,842

1996

21,245

11.8

3,768

1,397

179

105

5,448

1,062

447

369

3,317

1997

17,439

-17.9

2,556

1,087

184

89

3,916

912

389

327

2,797

1998

18,386

5.4

2,913

1,279

190

88

4,469

945

391

339

2,847

1999

23,976

30.4

4,157

1,954

247

106

6,464

1,161

453

418

3,506

200

33,47

39

6,03

2,53

33

13

9,03

1,55

60

59

4,69

CO

at

o

CJ

@

L

CO

CO

-vl

a

cu

•-+

( U cu

- 1

Ci

CO

CO

Table 3-4 (Continued)

Wafer Fab Equipment Revenue Forecast by Equipment Segment, 1995-2001 (Millions of U.S. Do

Equipment Segment

Chemical Mechanical Polishing

Total Etch and Clean

Tube CVD

Nontube Reactor CVD

Sputtering

Silicon Epitaxy

Other Deposition^

1995

179

4,680

780

1,803

1,567

207

1996

221

5,416

765

2,067

1,664

274

1997

241

4,666

646

1,818

1,437

262

124 115 83

Total Deposition

Diffusion

RTP

Total Thermal Nondeposition

4,480

773

154

927

4,886

740

188

928

4,246

611

200

811

Medium-Current Implant

384 403 310

High-Current Implant

550 640 496

High-Voltage Implant

119 223 163

Total Ion Implantation

1,053 1,265 968

Optical Metrology

98 106 81

CD-SEM

Thin Film Measurement

Patterned Wafer Inspection

Auto Unpatterned Detection

313

196

535

111

345

209

596

127

717

267

184

485

101

604

Other Process Control^

601

Total Process Control

Factory Automation

1,854

686

2,099

830

1,722

788

Other Equipment

Total Factory Automation/Other Equipment

386

1,072

373

1,203

322

1,110

Total Wafer Fab Equipment

19,010 21,245

^Includes contacrt/proximity, projection aligners, direct-write e-beam, and X-ray lithography

^Includes evaporation, MOCVD, and MBE

^Includes auto review/classification, manual detection/review, and other process (»ntroi equipment

Note: Columns may not add to totals shown because of rounding.

Source: Dataquest (January 1997)

17,439

5,518

805

315

1,120

363

674

234

1,271

140

1999

437

5,975

835

2,366

1,860

368

90

2,328

910

391

1,301

23,976

446

257

658

134

692

1998

301

4,823

675

1,905

1,480

283

78

4,422

624

230

854

297

478

175

951

95

316

201

504

106

. 589

1,810

739

318

1,057

18,386

10

7,70

1,13

46

1,59

55

1,05

36

200

63

8,08

1,15

3,36

2,59

48

19

96

3,23

1,28

58

1,86

33,47

1,96

18

60

33

94

Wafer Fab Equipment Forecast 29

Annual Investment Themes for 1996 through 2001

Behind our equipment and segment forecasts are assumptions about how semiconductor producers will perform and invest. These are summarized in Table 3-5 for the years 1996 through 2001. The following areas are considered: the availability of profits for reinvestment, memory versus logic growth, technology shifts, and brick and mortar versus equipment purchases.

When Will Demand Expand to Meet Capacity? An Update to the Over- or

Underinvestment Model

In our forecasts over the last few years, we have shown a model that provided a measure of the net cumulative over- or underinvestment in wafer fab equipment to support capacity needs in the industry. As equipment purchases precede actual capacity on line by a number of months or quarters, this model could be viewed as a gross "leading indicator" to capacity shortages and excesses. The results of this model are closer to a 1.5-to-3year indicator of turning points in the equipment industry. The methodology of the net cumulative investment (NCI) model is linked to our longerrange forecast model.

To review, our methodology starts with a few key assumptions and baselines:

• Long-term growth rates for semiconductors and wafer fab equipment are correlated. In other words, semiconductor revenue and profits are needed before money can be spent on equipment, and vice versa.

• Also, net cumulative investment equals zero over time—meaning that in a noncyclical environment where annual growth rates are constant, investment and capacity are at equilibrium at all times. Of course, our industry cycles through over- and underinvestment.

Table 3-5

Aiuiual Driving Forces and Investment Themes for Wafer Fab Equipment, 1996-2001

1998

Solid

1999

Strong

2000

Solid

2001

Solid Logic Semiconductor Unit

Growth*

Investment in Logic Capacity*

Memory Semiconductor Unit

(Not Bit) Growth*

1996

Moderate

Moderate

Moderate

Investment in Memory

Capacity*

Front-End Equipment versus

Facilities Loading of Capital

Moderate

Balanced

Primary Technologies

Invested

0.35-0.5 micron

'Scale: Strong > Solid > Moderate > Weak > Dead

Source: Dataquest (January 1997)

1997

Moderate to solid

Moderate

Weak

Dead

Facilities

0.3-0.5 micron

Solid

Moderate

Solid

Solid

Weak

Facilities

0.3-0.4 micron

Moderate Strong

Equipment

0.25-0.4 micron

Strong

Strong

Equipment

0.25-0.4 micron

Solid

Moderate to weak

Solid to moderate

Balanced

0.2-0.35 micron

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

30 Semiconductor Equipment. Manufacturing, and Materials Worldwide

• The output is a tangible number, and is in dollars of over- or underinvestment at year-end. However, the more useful output of the model divides this gross dollar number by the wafer fab equipment market size. The result is a percentage of market figure that is repeatable in level from cycle to cycle.

• To take into consideration the long-term growth of the semiconductor and equipment industries changing over time—the model has a factor allowing the fundamentals of the industry to change over time.

A net positive or negative investment is calculated relative to the longterm growth baseline annually, and then added to the prior year. The calculation resulted in a dollar value net cumulative over- or underinvestment, and has correlated well with historical patterns.

We consider the region of plus 20 percent or higher, or minus 15 percent or lower net cumulative investment is necessary to place the wafer fab equipment market within the time frame of a "turning point" in the market.

Figures 3-1 and 3-2 show the most recent results of the model, little changed from our midyear forecast update. In absolute dollar terms, by the end of 1995 the industry was $6.8 billion overinvested, or 36 percent of the wafer fab equipment market, exceeding levels witnessed during the

1984 and 1989 peaks. These levels were driven to these extended levels for two basic reasons. First, the PC unit demand was growing at low-20s annual growth rate in the early 1990s. About one-third of the semiconductor industry, and over one-half of the capital spending on new capacity is to support this demand.

Second, the DRAM market had not yet converted to run the more siliconefficient 16Mb DRAM by the end of 1995, placing this past investment cycle about seven years behind the last cycle. As the equipment being installed since late 1994 is fully "convertible" to run 16Mb DRAMs, we can think of these fabs building "pent-up supply" in bits. The size of this overspending means that even with good fundamental demand in semiconductors, the industry will not be spared at least the average length of a downturn, namely, two years.

We have factored into the model an investment in a couple of 300mm fabs starting in late 1997 through 1999, with the bulk in 1998. This is considered an equipment "bubble demand" as the equipment will be shipped into a nonproductive fab (meaning no semiconductor revenue will be initially generated).

With our forecast for a momentum style growth in 1996, and two pause years in 1997 and 1998, the model indicates a reacceleration of equipment spending starting in 1999. This model also shows that if our forecast of an

18 percent decline for wafer fab equipment coupled with a 13 percent growth in semiconductors is met for 1997, the equipment market will still be net slightly overinvested by the end of 1997, and clearly not close to the levels necessary to indicate the market would be near a turning point. We would not enter the turning point region until the later half of 1998.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast

Figure 3-1

Net Cumulative Over- and Underinvestment of Semiconductor Wafer Fab Equipment

(Millions of Dollars)

Millions of Dollars

7,000-rt-

6,000

5,000

4,000

P

M

3,000

2,000

. #

1,000

0

-1,000-1

•'""•o-CT

^.r^.

m.

-2,000

-3,000

m

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 199511996 1997 1998 1999 2000 2001

Forecast

871E77

Source: Dataquest (January 1997)

Figure 3-2

Net Cumulative Over- and Underinvestment of Semiconductor Wafer Fab Equipment

(Percentage of Wafer Fab Equipment Market)

Millions of Dollars

40-(I

30

20

10

0

-10

I

I

m, w

'7//.

i

9>Z\

'm

I

h

-20

m

-30-It

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 199511996 1997 1998 1999 2000 2001

Forecast

871S78

Source: Dataquest (January 1997)

31

March17,1997

SEMM-WW-MT-9603 ©1997 Dataquest

32 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Fundamental Capacity Analyses and Annual Review—DRAM and Foundry

The NCI model described above is only a tool to indicate possible future turning points and should not be relied upon to actually forecast capacity supply versus demand absolutely. A more fundamental, basic approach is required—looking at square inches of silicon capacity. How many wafers can be processed, and of what type of process, is much more enlightening in measuring capacity than for a particular device type. Capacity is very fluid: A stepper does not care whether the picture it takes is for a DRAM,

SRAM, or logic device. But there are limits in transferring capacity. For example, logic processes have specialized process techniques that are not found in DRAMs, and vice versa. SRAMs can use a DRAM equipment mix or a logic-oriented process scheme, the latter tending to have faster access times.

There are two major markets we can isolate in order to understand basic capacity supply and demand—the DRAM market and the foundry market. The latter is particularly interesting for two reasons. First, foundry capacity has tended to be more heavily logic and ASIC process oriented, giving us a second perspective on capacity versus supply. Second, the major customer base for the foundry is the fabless company, whose products tend to be those that are placed within the PC logic and graphics chipsets. As PCs now account for about one-third of all semiconductor use, about two-thirds of all DRAM consumption, and is the main engine for the current semiconductor boom, looking at PC-related capacity issues is important for understanding potential future equipment market turning points. Put simply, the foundry market can be used as a proxy for the mainstream logic supply and demand.

The details of these analyses are provided in other reports, namely,

Dataquest's Quarterly DRAM Supply/Demand Report and ongoing research on the foundry market. A summary of the basic results and impacts will be given here.

In any supply/demand trend, there is a cycle between oversupply and undersupply. Investment in capacity tends to be in reaction to these situations, and there is inevitably overshoot in both directions. Analysis is based on square inches of silicon, and not on revenue, bit demand, nor unit demand. If demand for silicon area exceeds supply, the market is technically in undersupply.

Here are the basic conclusions and review of the market impacts by year:

• The DRAM market transitioned to oversupply very late in 1995. A high level of capacity investment continued until the middle of 1996, and

,/ with the conversion to a more silicon efficient 16Mb DRAM, this condition is expected to last into the middle of 1998. This window may be later if significant Japanese capacity is converted from 150mm wafers to

200mm wafers.

• The foundry market has also transitioned to oversupply. Investments in the dedicated foundry have continued to accelerate, while the overcapacity in U.S. logic manufacturers has meant a slightly lower demand.

In the last half of 1996, the Korean manufacturers aggressively entered the market, driving prices down 25 percent for processed wafers.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast 33

f

I

• 1996—A year of transition and spotty investment in capacity. DRAM prices crashed early in the year. The market begins to transition to the

16Mb during the year, and prices will continue to decline gradually into

1997. Foundry prices collapsed late in the year.

• 1997—A year of contraction. Weak DRAM capacity investment, but continued strong investment in technology and strategic investments.

But the oversupply in the foundry capacity means equipment purchases will likely lose steam quickly through the year. We expect

DRAM investment to decline some 20 to 30 percent, while logic investment remains flat.

• 1998—A frustrating year. Even though DRAM capacity is likely to see its peak oversupply in 1997, significant reinvestment will not likely occur until late in 1998 or 1999 as companies delay investment and

Japanese companies finish 200mm conversions. As the foundry market deals with oversupply and new entrants, logic investment will remain cool. Yet we expect that semiconductor demand will again begin to accelerate, signaling an underlying strength in the market and anticipation of the eventual upturn. U.S. logic investment will lead, but will remain subdued because of the redirection of production into the foundry market.

• 1999—The first growth year in the next boom cycle. DRAM investment is likely to pick up again, while foundry investment will lag. U.S. logic capacity will likely be strong and lead foundry investment out of the

"pause." By 2000, the next equipment boom will be well under way and likely to last into 2001.

A more detailed discussion of the quantity of the silicon square inch oversupply is in Chapter 2 of this document.

Highlights of Key Equipment Segment Maricets and Forecasts

steppers and Tracic

From 1989, the peak year of stepper shipments at more than 950 units, the market tumbled to less than 400 tools in 1992 before recovering. During this DRAM-sensitive ramp, the industry experienced its first year exceeding 1,000 steppers shipped; indeed, 1,228 steppers shipped in 1995. The peak year was 1996, in which we currently estimate just under 1,400 steppers shipped into the market. Shifts in the product mix toward higherpriced deep-UV systems, i-line systems and wide field lenses have also driven up average selling prices (ASPs). This trend was offset somewhat in 1996 by the weak yen, to drive a revenue increase of 13 percent on a dollar basis in 1996, essentially at market growth. Steppers, being a capacitysensitive segment, is forecast to lag the market performance in 1997 and

1998, and then lead market growth starting in 1999.

Stepper revenue is forecast to grow at 13.4 percent CAGR, slightly below the market average for 1995 through 2001. Our forecast for stepper unit growth over the five-year forecast horizon remains modest, about 4.3 percent CAGR between 1995 and 2001.

I

SEMM-WW-l\/IT-9603 ©1997 Dataquast March 17,1997

34 Semiconductor Equipment, Manufacturing, and Materials Worldwide

With the adoption of phase shift mask technology, off-axis illumination techniques, as well as conventional i-line tools with variable numerical aperture, i-line is clearly a viable technology down to the 0.3-micron regime and will continue to dominate the overall stepper technology mix through the 2000 time frame. Excimer/deep-UV steppers will begin to represent a more significant portion of the product mix from 1997 onward for use in below 0.3-micron devices, and ICs with large chip areas such as advanced microprocessors, which require large field size capability.

Dataquest believes that field size pressures accompanied by shrinking geometry will drive the industry toward step-and-scan technologies for the majority of excimer/deep-LTV shipments, begirming in 1997.

Track equipment is forecast to grow at a 12.4 percent CAGR between 1995 and 2001, slightly below the industry growth of 13.7 percent. While we believe that the rapid shift in the product mix toward higher-priced systems has been recently completed, we do expect another product shift to occur in the track market associated with the ramp of deep-UV steppers, which require more sophisticated environmental control systems, and will translate to slightly higher ASPs. Unit demand will parallel the stepper market, however.

Etch and Clean: Dry Etch and Chemical Mechanical Polishing (CMP)

Dataquest began covering the chemical mechanical polishing (CMP) market in 1993. At this time Dataquest includes the post-CMP clean system, usually sold in conjunction with a CMP tool, as part of the cleaning segment, and not in the CMP segment.

The year 1995 was a critical one for adoption of the technology into production, with CMP systems growing 180 percent to $179 million. Even though the application appears to be limited to devices with at least three levels of metal, which tends to exclude the DRAM market, the acceptance of the technology into logic and particularly the foundry market has been the key turning point. Based on the demand of the customer, however, and because of the introduction of more robust equipment, we now believe most foundries offer CMP at 0.5-micron. CMP is one of the few markets Dataquest believes will grow during 1997, albeit at single-digit levels, during the market contraction in 1997.

These systems are used to remove material from the surface of the wafer resulting in a flat surface over the entire wafer. This global planarization, primarily of dielectric layers, is required to achieve high yields in devices where three or more levels of metal are used. Today's advanced logic and

ASIC devices are fueling this market growth. Dataquest believes that this technology and market will become a major part of semiconductor manufacturing in the long run. CMP is our fastest-growing segment with a

31 percent CAGR for the years 1995 through 2001.

Dry etch systems continue to exhibit long-term revenue growth, with a

CAGR of 12.5 percent forecast for the years 1995 through 2001. Unit shipments are expected to grow as greater multilevel intercormect process capacity is brought online, increasing the need for dielectric and metal capacity. Relatively strong ASP growth will lend additional momentum to dry etch revenue growth as new high density plasma systems for 0.35micron applications enter the martet, and multichamber cluster tools

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast 35 continue to increase their presence. The success of CMP will hold etch below market growth, however, particularly in metal etch as stringer removal becomes a nonissue. The metal etch market will be the slowestgrowing subsegment of the etch market through the year 2000, and may actually flatten altogether when dual damascene processes enter production around the turn of the decade.

Deposition: CVD, PVD, and Silicon Epitaxy

CVD equipment sales are predicted to grow at a 14 percent annualized rate fron\ 1995 through 2001, only slightly above overall equipment growth. The steady growth in multilevel intercormect capacity will continue to generate demand for dielectric and metal CVD systems. ASP growth will also contribute to revenue growth, as more highly integrated systems with improved productivity and particle control appear in the marketplace. Advanced dielectric deposition systems utilizing HDP sources have been introduced and will gain momentum for intermetal dielectric (IMD) applications for sub-0.5 micron processes. Metal CVD will continue to exhibit strong growth, driven by blanket tungsten for contact and via plugs. For these reasons, the forecast for nontube CVD systems outperforms tube furnaces.

Sputter deposition systems are forecast to grow at an annualized rate of

12.6 percent for the years 1995 through 2001. As in the case of dry etch and

CVD equipment, continued expansion of multilevel interconnect process capacity is the primary driver behind sputter system growth. Rapid growth in average system ASP has helped to drive total revenue growth prior to 1995, primarily from the quick and expanding dominance of

Applied Materials in the market. With Applied now accounting for more than 50 percent of the market, the bulk of tiie ASP increases are behind us.

Changes in system architecture, pioneered by the Applied Materials

Endura system, will continue to yield improvements in film properties, equipment productivity, and defect density. This is a market segment that will be somewhat buffered from a slowdown in DRAM investment, as the fundamental growth in the number of metal layers in ASICs and logic devices drive a more stable outlook.

The need for single-wafer epitaxial systems has been the primary driver on epitaxial deposition systems, as the need for 200mm epitaxial wafer capacity for CMOS logic applications was needed. However, this capacity is more expensive than wafer suppliers would like, so we expect the concept of "minibatches" to emerge as a viable production strategy, as it has in

CVD. Moore Technologies is known to have such a product on the market, but has yet to be widely accepted in the market. A strong automotive, power, and discrete device market has increased demand for the specialty batch units, and growth in this segment actually exceeded CMOS logic in

1995 and continued to be strong in early 1996. There are clear indications that epitaxial layers will be required for some 64Mb DRAMs (see Chapter

4). This, along with an increased product mix of logic semiconductors at

200mm wafer capacity, and sustained demand for discrete devices and power ICs will be the primary drivers for epitaxial deposition equipment growth. In fact, epitaxial reactors are expected to be a star performer with an average annual growth of 19.5 percent, and will be somewhat insulated from the near-term slowdown. We believe the subsegment of 200mm systems will actually grow through the 1997 contraction.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

36 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Thermal Nondeposition Processes: Diffusion and RTF

Diffusion systems are expected to demonstrate a CAGR of only 9.7 percent for 1995 through 2001. Newer vertical systems will be configurable as multitube clusters with integrated dry clean capability, to compete with single wafer cluster tools. Tube systems will also incorporate small batch capabilities to offer greater flexibility for custom and semicustom circuit manufacturers. Diffusion tubes are growing significantly slower than the market for two reasons. First, RTF processes are slowly penetrating steps that had been done by the tube furnace. While we believe this penetration will remain slow until a change to 300mm wafers, RTF will keep diffusion tube market growth below average. Second, and more serious in the near term, is the prospect of recently developed high voltage implant schemes actually starting to replace diffusion steps in logic processes.

RTF is expected to grow at an annualized rate of 22.9 percent for 1995 through 2001. This market grew much faster than we anticipated, nearly doubling in 1994 and slightly more than doubling in 1995. The growth in

1994 was primarily fueled by the growing acceptance of self-aligned silicide processes in logic process flows. The growth in 1995 comes from new offerings in the market from Applied Materials, CVC, and Mattson, and from the expansion of the application into traditional tube diffusion steps.

The real growth for this segment will come from transitioning of the thermal "nondepositing" processes away from diffusion tubes and into singlewafer RTF systems for 300mm wafers. We have factored in a large complement of systems into initial 300mrn facilities starting in late 1997, largely contributing to the higher-than-the-market growth. RTF systems are primarily used today for salicide anneals, and some implant drive-ins, and are primarily driven by logic and ASIC capacity.

Dataquest believes that batch tube systems will continue to resist penetration by RTF in areas such as well drive, BFSG reflow, and thermal oxidation because of the demonstrated cost of ownership benefits in these areas, at least through 200mm wafers. For 300mm wafers, there will also be a strong desire on the part of the semiconductor manufacturer to continue to use the batch tube systems, as these systems offer much better cost efficiencies.

Ion implantation

Overall ion implantation system revenue is forecast to grow at a CAGR of

15.5 percent for the years 1995 through 2001. This market segment will continue to be one of the most volatile, because of the highly device specific nature of the implant segments and of the dependence on new fab capacity for unit growth. The fastest-growing segment is expected to be high energy implantation, which is evoking intense interest because of its potential for process simplification and manufacturing cost reduction. The first year of true production ramp is expected to occur in 1998 as 0.4micron technologies become mainstream, although early adopters such as

Samsung have placed high-voltage implant into 16Mb DRAMs. A recent process developed by Genus is attracting attention and gaining acceptance in logic processing, offering a technique to laterally isolate transistors.

Near term, high voltage systems will decline in sales because of the current dependence on the DRAM market.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast 37

>

>

New implant systems will continue to offer improvement in uniformity, particle control, charging, and wafer throughput. The number of implant steps requiring medium current ion sources is expected to grow faster than high dose implant steps, again driven by the higher worldwide semiconductor logic component, with the shallow junctions preferentially driving the trend toward medium ciirrent implants.

However, our forecast does not reflect this trend, with medium current implant sales lagging the market with only 9.7 percent CAGR. Why?

Recently Eaton, Applied Materials, and Genus have introduced expanded capability to existing systems or new systems that are targeted to compete in more than one segment. For example. Applied Materials' 9500 systems now basically have a range to cover both high current and medium current capabilities, and Gentis' new 1520 model expands the range effectively across high voltage and medium current. Traditional batch medium current systems are effectively being "squeezed" out of the market. This is occurring because equipment utilization for implanters tend to be among the lowest in the fab. Semiconductor manufacturers, in an effort to increase utilization and reduce cost, are tending to buy equipment with broader ranges.

There are medium current applications that will still require dedicated medium current systems, namely high-tilt (greater than 42 degrees) implants and V-p adjustment implants. However, these will be better executed using implanters with single wafer end stations rather than the traditional batch systems.

Dataquest is investigating the "redefinition" of ion implant segments in order to better capture this market dynamic, as well as segment the components of the high-voltage implant segment into its well structure application (which performs implants at the 700- to 800-keV range) and the lateral isolation application (requiring implant energy greater than 1.5

MeV).

Segments That Will Fare Best the Next Two Years

During the coming slowdown, there will be two kinds of purchasing behaviors that equipment companies can take advantage of to buffer them from sales declines. The first behavior is tied to what types of capacity will be required early in the recovery cycle. While logic capacity will be a larger part of the mix, only advanced technology in the logic area is expected to grow during 1997, as well as equipment that supplies the maskmaking industry. Included in this category are the segments of nontube CVD, silicon epitaxy, maskmaking lithography, process control systems in the materials business, and RTF.

The second purchasing behavior that will be prevalent over the next 18 months will be new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. By no means will this listing be complete, but we will highlight a few of these areas here.

The movement to deep-UV lithography will pick up steam through this slowdown. The unit forecast for deep-UV steppers is for more than 360 systems to ship in 1998, up from the 105 or so shipped during 1996.

SEMI\/l-WW-l\/IT-9603 ©1997 Dataquest March 17,1997

38 Semiconductor Equipment, Manufacturing, and Materials Worldwide

There are several equipment segments that will benefit from this movement—those inspection and process control systems which must operate in the deep-UV wavelength in order to detect defects and materials on the surface of the wafer, namely, mask inspection and thin-film measurement equipment. Thin-film measurement is key to intra- and interlevel metal interconnect, and storage capacitance applications. Thin-film measurement systems are used in-line to monitor the in etch, lithography

(photoresist), deposition, and diffusion steps. Although this market was driven by logic applications more than memory production, in the past several years DRAM manufacturers have begun integrating thin-film measurement stations into their process lines.

Chemical mechanical polishing systems are becoming more robust and able to meet tighter specifications. Japanese and Asian companies are just begirming to investigate these systems. While dielectric CMP is relatively common, metal CMP is still emerging. Above three levels of metal, it becomes more beneficial to implement a metal CMP step to the process.

Gap fill dielectric CVD has been a highly competitive development arena.

Novellus Systems appears to have product momentum with its Speed product line. We believe this product area offers a way for companies to effectively buffer the current slowdown.

Silicon epitaxial reactors will benefit from the conversion and ramp of the industry in 200mm wafers, as wafer suppliers have to retool their epitaxial capability. Also the possible increased use of epitaxial layers for advanced

DRAMs means increased need for epitaxial reactors.

Wafer cleaning systems can also offer an area where a company's technology can shine. The recent announcement by Steag AG in Germany of its vapor dryer was built into a successful strategy when the company sold these systems as part of an integrated automated wet clean station. Wafers are placed in the dryer system after a wet cleaning process to dry, but do not spin. Instead, they are pulled from a bath of IP A/water in a controlled fashion into an IPA vapor atmosphere. As the wafer is pulled out, the liquid sheets off the surface in such a way to not leave any water spots, which often hold killer residue defects. The concept of automating cleaning processes will be a key product issues over the next few years. Spray processors also address the key cost issue of chemical use per square inch of silicon. These systems may be more common in 300mm fabs.

Dataquest Perspective

Overcapacity has gripped the industry and 1996 was a year that turned from growth to decline, but net was just tinder 12 percent growth in wafer fab equipment. We saw two distinctly different markets in the past year.

The first half was marked by many fab announcements, aggressive spending plans, and good bookings levels for equipment. The second half was the opposite: bookings collapsed, expansions were delayed, and capital spending budgets cut.

The only reason 1996 remains a double-digit growth year is that there were strong backlogs coming out of 1995. The bookings decline has put pressure on backlog levels, and the industry has shipped down backlog

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Wafer Fab Equipment Forecast 39 now to more normal levels, with several companies below normal. However we anticipate that equipment companies, in the hope of improved bookings later in the year, will continue to ship down backlog in order to maintain a minimum financial performance, maintaining the run rates of fourth quarter 1996. We expect to reach a minimum three-month level backlog by the middle of the year 1997. At that time, with backlogs razor thin, quarterly shipment rates are expected to be reduced to booking levels, leading to a sales dip in third quarter 1997.

We do not expect sustainable bookings growth to support shipment growth until the second quarter of 1998, and shipment growth is expected to remain constrained until very late in the year, keeping 1998 a singledigit growth year. We would expect supply/demand dynamics to be fully corrected by early 1999, driving a robust resumption of growth with the wafer fab equipment market growing to over $33 billion in the year 2000, from the just under $19 billion in 1998.

We believe the market in 1997 and early 1998 will concentrate on advanced technology equipment. A purchasing behavior that will be prevalent in the near term will be new processes and equipment directed at solving issues for 0.35-micron and 0.25-micron manufacturing. These segments will be those related primarily to deep-UV lithography, silicon epitaxy, maskmaking equipment, 300mm technologies, and 0.25-micron multilevel metallization process schemes.

We have factored in an infrastructure investment in equipment for late

1997 through 1999, which will affect the forecast size of the markets positively. This additional investment will be for initial eqviipment to fill a couple 300mm fabs to run silicon by 1999. However, our outlook for a significant 300mm equipment market will wait until well after 2000.

SEMM-WW-l\/IT-9603 ©1997 Dataquest March 17,1997

>

I

Chapter 4

Silicon Wafer Forecast

Dataquest's current forecast for regional silicon wafer demand reflects significant silicon wafer growth during the first half of 1996 similar to 1995, with a significant slowdown in the second half of 1996 carrying over into

1997. This slowdown is expected to be short-lived as strong semiconductor unit growth coupled with the essential completion of the migration from the 4Mb DRAM to the 16Mb DRAM. Our latest forecast, along with highlights of some of the key factors affecting the regional markets, is presented here.

Silicon Forecast Tallies

Tables in this chapter include Dataquest's most recent forecasts of regional unit silicon wafer consumption. Three tables (4-1 through 4-3) detail unit consumption by region. Individual forecasts of major product segments such as prime, epitaxial, and test and monitor wafers are included. For the first time in this publication, the merchant epitaxial wafer market is forecast by application (Table 4-3). Also five tables (4-4 through 4-8) of worldwide and regional forecasts for wafer size distribution are presented.

Silicon Wafer Revenue Forecast

Table 4-9 contains the revenue forecast for silicon wafers worldwide. Our methodology takes the wafer size distribution forecast in Table 4-4 and multiplies this matrix by our view of wafer price trends and test wafer consumption trends (by wafer size) in the forecast horizon. We do not publish our specific wafer price forecast, but do make comments on trend outlooks from time to time. For the first time, we have broken out the polished and epitaxial wafer components of the revenue forecast, and is consistent with how Dataquest has been collecting revenue information since

1985.

In our analysis, we have concluded that the revenue forecast would resemble the semiconductor industry more closely than the capital spending markets. The concept of semiconductor revenue per square inch is more closely tied to silicon consumption than raw wafer capacity of the industry, although this metric is cyclical. The six-year CAGR of 16.2 percent is slightly higher than the semiconductor forecast of 13.9 percent. This is somewhat contrary to the conventional model that semiconductor manufacturers will attempt to control the costs associated with manufacturing, which includes using silicon more efficiently in the future, which would tend to keep the growth rate of silicon revenue slightly below that of the semiconductor revenue. Yet, our forecast does not reflect this, primarily because the industry is migrating seriously over the next several years to

200mm wafers, which have a silicon wafer revenue per square inch significantly higher than the smaller wafer sizes. Further, the migration toward epitaxial wafer use, however slight, will also tend to favor higher silicon market growth relative to semiconductors. Therefore, the mix shift will tend to increase the growth rates in revenue, bringing the forecast at a slightly higher level to the semiconductor growth rates.

SEI\/IM-WW-IVIT-9603 ©1997 Dataquest 41

m

CO

Oi

o

CO

@ t o

CO

•-J

o

91 r—I-

cu

J=]

Table 4-1

Forecast of Captive and Merchant Silicon* and Merchant Epitaxial Wafers by Region (Millions

Worldwide Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

Americas Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

Japanese Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

European Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

Asia/Pacific Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

NA = Nat applicable

'Includes prims, test, and monitor wafers

Source: Datai^usst (January 1997)

1994

2,907.6

1,869.2

129.7

1,277.9

878.1

111.8

42.0

246.0

353.8

222.4

359.7

147.0

531.7

829.0

440.0

172.3

87.0

57.3

6.0

68.1

446.9

328.7

18.3

12.0

87.9

226.5

49.0

182.6

1,483.4

1,036.6

137.1

22.0

287.7

1995

3,523.8

2,271.4

477.4

74.0

701.0

943.1

485.0

457.7

281.4

86.1

3.0

87.2

639.6

468.4

27.7

0

143.5

1997

4,048.2

2,348.3

635.3

74.0

990.6

1,158.2

573.3

297.4

49.0

238.5

1,588.4

958.8

182.1

22.0

425.5

562.4

317.1

123.3

3.0

119.0

739.2

499.1

32.5

0

207.6

1996

3,777.5

2,307.4

559.6

74.0

836.5

1,063.5

534.1

266.5

49.0

213.9

1,502.2

982.9

159.0

22.0

338.3

519.3

303.6

109.4

3.0

103.3

692.5

486.8

24.7

0

181.0

1998

222.7

22.0

469.3

641.1

349.3

145.3

3.0

143.5

871.7

591.7

42.7

0

237.3

4,600.8

2,631.5

754.4

74.0

1,140.9

1,323.5

640.0

343.7

49.0

290.8

1,764.5

1,050.5

551

841

449

197

3

192

1,254

868

845

449

49

347

2,221

1,265

382

22

81

200

6,009

3,428

1,110

74

1,395

1,691

1999

5,252.6

3,025.7

268.0

22.0

517.1

729.2

401.0

159.6

3.0

165.6

865.1

74.0

1,287.8

1,515.9

754.0

381.9

49.0

331.0

1,961.8

1,154.7

1,045.7

716.0

55.6

0

274.1

304

CO

CO

-vl

@

CO

CO

D l

I—(-

cu

tn

CO en

Dl

O

Table 4-2

Forecast Growth Rates of Captive and Merchant Silicon* and Merchant Epitaxial Wafers by Reg

Worldwide Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

Americas Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Morutor

Japanese Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial SUicon

Captive Silicon

Merchant Virgin Test and Monitor

European Total Silicon -^ Epi

Merchant Prime Polished Sihcon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

Asia/Pacific Total Silicon + Epi

Merchant Prime Polished Silicon

Merchant Prime Epitaxial Silicon

Captive Silicon

Merchant Virgin Test and Monitor

NA - Not applicable

'Includes prime, test, and monitor wafers

Source: PataquBst (January 1997)

1996

7.2

1.6

17.2

0

19.3

12.8

10.1

17.7

0

17.1

1.3

-5.2

16.0

0

17.6

13.5

7.9

27.1

0

18.5

8.3

3.9

-10.8

NA

26.1

31.5

-43.7

40.8

16.1

18.1

22.6

-47.6

17.0

29.4

26.5

50.3

-50.0

28.0

43.1

42.5

51.4

-100.0

63.3

1995

21.2

21.5

32.7

-49.7

31.8

13.8

10.2

15.6

9.5

28.1

17.6

21.6

12.7

11.9

16.0

13.5

1994

18.9

16.5

26.0

17.6

23.2

13.8

22.2

14.2

37.4

20.0

41.3

46.3

47.5

41.9

33.3

45.0

0

20.6

17.9

18.6

31.4

NA

14.3

14.3

11.6

15.6

0

21.9

11.1

9.6

22.3

0

10.3

14.0

10.2

17.8

1998

13.7

12.1

18.7

0

15.2

1997

7.2

1.8

13.5

0

18.4

8.9

7.3

11.6

0

11.5

5.7

-2.5

14.5

0

25.8

8.3

4.4

12.7

0

15.2

6.7

2.5

31.6

NA

14.7

CO

CO

•vl

44

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Co S

^1

U a^

'*

o\

I-H

f\ S

U

in

ni

1—1

( N

I N

I—1

oq vd

0 0 i n i n

I - H

o

CN

I - H

CN

CS

I-H

VO t< i n

VO cn

CN

0 0

• *

I - H

vq

I-H t - H

2

VD t<

I - H

VO

»*

I - H

o cn

t - H

I-H

CM

VD

OS

0 0

CM t s

OS

T-H

• ^ iri

I-H

T-H

^

T-H

I

g

i

QO cs

0 0 cn

f-H

o

f^

^

r H

^

CD rs.

VD

O

"*

cs

0 0

•<1<

^ m

•>i<

^

t

S

I-H

Ov

as

0 0

•^

•^

0 0

t - H

I-H

OS

s

I-H

N .

VO

OS

I N

CS o o

l - H

o

s

0 0

^

b-. vd

CM

t - H

I-H

I-H

VD

0 0

iri t - H oq

OS rJH en

Oi

u

HH

0)

> H

(0

3 cr

CA o

1

f - H d

I—1

I - * '

tv t N tfj t^ in

Ot3 t N

O f O t>.

l - H

ro t N

as

O N

• *

• ^

I - H

as

I - H

cn in

^

t-» cn

VD in in cvi

0 0 cn in

VD

O

I-H

I-H

VD

0 0

OS

CM

I-H

O

!<

OS

I-H

t N cn o

I-H

VD

QO

VO

0 0

• ^

CM cn

I-H

0 0 oq vd

•^

IS.

cri

I-H

oq

d

CM

s s

I-H ffi

0 0

0 0

0 0

( N

iri

I-H i - <

• < *

I-H

OS

I-H

0 0

CO o\

S

CM

ON

CN i n

I-H

^

o

0 0

VO

CM t>^

Ov i n o

t - H

oq

0(3

VO

VO

OS

Ov

d

Ov

m

^

CN

d

t N

VO

ffi

CM

OS cn cn

OS

I-H

Is!

3

0^

0^ rH

O

•»->

(0

u

• F N

"a

D H

<

•d

03

01

m

Tt^ u

•1M

^ H

•1H

C/3

1 - ^

n

• F N

X

(Q

••a

u

r^

E5

a

S

c o

_u

c/5

"(3

*H

ni j : ; tn

( 0

E

"P,

PU

+ j o o a

(IQ

" - V

U

•a o en

O

T 3

1

^

t j

s s

• *

• ^ j '

CJ\

m

in

• ^

T-H

[ < t N

CN

•<*

P

h ^

r^

^

en

VO

VD

OS ifi i n cn

i - <

cn

I X

CM

CM

CM oq

VD

0 0

P t - H t - H t - H

Ov

•^

CM cn in

• ^ t - H

t N

d

0 0

p

I-H m t - H

^

I-H

IS.

CN

^

S;

S m

in

s

a^

• ^

^ m

in

0 0

Y-H o

in

^

^

!<

OS c^

I N

c3

a

VD

iri

• ^

VD r-H

CN|

I-H

CM

QO

I-H

cn

Ov tx

I N

t - H

Ov

^>.

I-H

I-H

cn cn

CN

I-H

oq

I-H

t V

' t

t - H

xj^

I-H d

I-H

m cvi cn

s s

in a\ in in

• ^

J j '

[ V

CD tv

•*

« m d

CN in

VO

VC

CN

^

• « ^

I—1

CM

VO

d

ti< in

I-H

I-H

P

OV in

I-H

ov

^

[ ^ cn

I-H

0 0

0 0

CM

• < *

OS

T - l

isl

VD bN

VD cn

VD

iri

Is.

• ^

CM cn

d

cn cn

Ov

t - H cri

ts

cri

CN

I N ts vq

I-H

o

0 0

I-H t - H

vd

VO

d

s s

• ^

N; t->

• *

O

M5

CT\ t N

OJ

1-H

t X

1-H

VO o\ i n

VD

CN

CN

•^

N; t^

I-H

cn

Qi

0 0

Vf3

t - H

t V cn

I-H

oq

T-H m

• ^

i n

0 0

1

I-H

VO

0 0

O

VD

-*

in t < cn in

CM t s t s

CM t s

d

CN ts vd

CN

d

s s

in o\ in en

m

t - H

t N t-^ t N

I—1

I-H

in in cn

I-H

t N

l-H

VD

VO t N in

iri

oq

I-H t - H

I-H

I - H

CM i n

VD

OS

m

1

cn

Iv

m

in

0 0

?i

cn

0 0

l - H m t - H

t^

cri

1 c o

u

-TH

•a

rO j : : in

[ 0

E

t".

[ Q

'o

cu

+J cu u tn

s

w

ca

,

Q

1

•c

CJ cn ra u

1

1 tn

O

u

•a o u

-w m

t H

u tn

i3

< u;

Q

c

8

EJ5

•H

(Q

-^

'E.

til

-s

(D

E w i

^^

"o

O-

(3

u

'Eb

o

QJ m tu

C m d .

(0 crt

O u

HI

- ^ j

u in

B

<

Di

D

§

u

M

tH

"ro

•H

la

4-*

"B. w

E

"o

Bin

r

(fl

H-"

O

H

C

ra tu

n. o

=1

W

u

u

• a

2

cn

O

u

t;

s

<

Pi

D

to

A

- I H

'p.

a

H

u

•0 ra

& •

ffi

^^

u

•a

5

tn

O

^ _

U

ja

E

'^

( H

f a

"o

CJ

tJ

D

E!

D

r«. a> c»

I— e-

n

tts

| S

o at

E 'S

B °

Z jb

II s.

Z CO

SEMM-WW-MT-9603 ©1997Dataquest March 1 7 . 1 9 9 7

I

I

Z661.'ZmOJB^ lS8nbE;[email protected]

E096-i^-MM-l/^W3S

CO

5

i

S

1

^

?

3

>

<

1

TO

0 )

1

CO

TO

1-1

s

Bl

3

n

TO

• I

H o r - f

D l

• - t i

TO

•-t

V I

1

C*l

O

I'

3 hO o

I'

3

H-»

U l

f

3 k - i ro

U l

5" n

3-

U i

N>

>-'

• - >

W

O S

1—* o

N O

U l

O N

• p .

0 0

O S

V ]

|NJ

V I w

0 0 o h — '

S O o

S J

S J u>

4^

N O

N O u

V I

V I

> - i o o

3

3 fei

N«4

3

i~i

rr

1 — ' ro

I->

V I

V I o

V I w

U l

O J

U l

O S

C

3. ro

1 1 a n

TO

U l rV a cn ff e n o

^

^

tu f? tn r i -

01

1-1 r + cn

2

o

H o

[ U

1

1

( U

' ' y :

C i ) o

f

3 f

O t o

C 3

C 3

3

3 w

• - ' ht^. o so o

"-K

t

a>

• - K

TO

• - 1 cn

• - 1

0 0 so ro so o

0 0

(-» o o o o

(-' o so

U I

O S

•f^

0 0

OS

V I

1—» ro t o t o

V I

CO

0 0

I—1

U l

H-* t o

1*

3

3

3

4^

V I o

•—1 so o ro ro

4^

V I

K-» o

f

3 h-» t o

H-1

V I h-1

4:^ bo h-»

4^ u

TO

C/I a"

I O

( 0

tl>

i"

rtj

• 1 3

TO l - l n

Cfi

e

B>

- l

TO

S' n

3" cn

V I o

V ]

CO h-»

4^

5

tu

3 ro

9

a.s I

n

1

£? a "

ffi* t

C * ^ ^ - k ^

I - H

3

1

ss-

^ ^

B)

HK n

CD

N n a.

3

o

1 ^

n

S I cn cr

F:

I f

• ^ • o

V O

> l ^

I

M a p

i-i

^

IS n t-(

U l w

^

I - '

U l

U l o

1 — '

4^

J-"

U l

•><

NO w so w w

V I so

U l

U l o so rsj

1 — ' ro w

Ul to

I - " o o o o h->

N O

U l

^

o ro

(-»

• - » h - l

H - ' p

• - » cn

13 l - H

B> tn

n

CD tn

O S u

•—1

U l o

M u ll:^

U l

OS

U l w

U l

• - 1 u>

N>

I n

•f^ ro o u

V I ro w

"vi

VI

1 — ' o o o o

U J p l-»

4^

K-» o

V I U l o bo o

U l bo

• ( 5 .

(->

U l

I - " o o

O N u u

4^

U l w u> w ro so

V I u

O S o ro ps o

1—»

>->

O N p k)

O S

^

O S u

a

U)

U) o ro

0 0

NO

U )

H-» o t o p\

H - i

O N

I - "

V ]

O N

4^ o

U l u

U l

U )

O S

U l

0 0

U l w

K )

KJ

K>

0 0

O N ro bo o v j t o

4^ g

0 0 o o o o t o

4^ p

H-N

CO

4 ^

U l t—1

U l

O S

0 0

N O o

O S o

CO

V I

O N o h — ' o o o o

4 i .

4 ^

O N

K-»

CO

(-»

V I

>—*

C O

O N

V I

OS o

U l o t - l

>l^ ro

U l

>o

U l

O J

>-» o o o

I - "

H - '

4^

N O

V I

CO o

U l t — »

H-»

V I

C3S

OS o

4^ o

s

* s

s

w

s

V O

»

a

^^* fii

B n ro t-»

u

CL

3

13

1 ^

3 cr

CO

cr

O N N C

JS

U l

U l p so

O S w

U l

O N

U l

OS u w hp>. ro

N O

U l ro

U l C3 t - i

4^

4^

O N

O o o o

U l

(-»

U l ro so so t — * o

O N

O N o o

ip

o i a^

It^ o ro o

O N

O

1 — '

O S

^

V J rf^ bo

O N

OS

OS w ro bo ro

0 0 ro ro

• - 1

O i - i

O u>

O N

O N h - l o o o t o bo

U l

4^ so ro va

U l

N O

4^

U l io o io o

M i

s.

o

f3

cr

^seosjod -is^BM uooiiis

9l7

46

Semiconductor Equipment, IVIanufacturing, and iVIateriais Worldwide o

•w^

0

•c

s

1 o en o

V D

Is.'

C S

0 0

o

m m

en

0 0

• < t

N O

I N

o

a

o

I—1

N O

^.

t - H

00

O N

o

1 o

m

o i n

0 0

C N

O N

o

m m

f O t s

'*

t v

I - H

o o o

I—1 r H

O N

N O

T - H

N O

I - i

I - H

o

0 0

o

N O

I - H

I - H

P

0 0

• ^

C N

C M

•<*

00

I - H

^ o

N O

I - H

N O

O N

I - ;

V O

o ts o

0 0

I - H

I - H

C N

0 0

• *

O

CN

•^

N O

I - H

N O

C M

o

0 0

ts! m

O t - H

N O

G

t3

C

ns

u

Ol

H-"

Ol

OS

s

s

o

•^ o

• ^

O N

p o

I—1

N D

C O

en

• *

i n

• *

C N i - i

o o o

r H

N O

T H

i n

T H "

i n

• ^

I - H

o

O N

o tv

T-l

I - H

p

0 0

N O

0 0 t - H

I - H

• ^

I H t s

I - H

o

• ^

rn i n

I - H

o

N O

ss s o i n o t>. o

T-l

o

(N

T - l

o>

•<* en i n

i - <

• > *

•<1<

o o o o

r - l

•<^

C S

en

•"^

I - H t - H

o

O N

o

N O

I - H

I—1

^

0 0

O N

N O

I - H

en

I - H

I - H

i n p o

I - H

O N

•>*

>o

0 0

i n ja en

QJ

J5 u

d

fc s o

V O

o

C N

C N

I—1

^ D

T - H

o

0 0

en

• ^

^

C M

o o o

00 m

I - H

I - H

O N

0 0

o p

r H

N O

I - H t - H

O N

0 0

I - H

N O

I-H

C N

0 0

C M

o o

0 0

^

0 0

N O

iri

•-H

(S

;d

cr

CD

V

00 na

•s

PUl

1

s

s

o

t - H

o

00 o o

•<:l'

r H

p

T - i

i n

N O

T - H

• ^

O N

en en

O N

C N

o tv

IS!

I—1

tS

O N

I - H

I—1

o

• < *

O N f—(

CM o o o o

1—1

o o o

I—1 p

1 - H

0 0

C N

I - H

5

O N

00 en

I - H

o

C M t - i

C M

C M

I - H

CM

O N

en i n

I - H

"*

N O

o en o en

i - <

I X en

I - H

in

O N

S O

en

I - H

(N

• ^

o

• ^

^

O N

C N

^

W

i n i n

O N

C N iri

s

s

I—1

o

r H

I - <

i n o

C S

C D

I N

( N

00

I N

• < *

CN en

I—1

o o o o

l - H

O N

CN

00

N O

i n

I - H

en o cn

I - i

o

• *

I - H

Is.

O N

o en

I - H

(N

CM o

•<* o t

t - H

I - H

i n rh

0 ^

0 ^

CB

OS u

Oi

VH

o

^ c

&

M

1

'*

I - l

(6

tv o tv

t s

»—1

o i

T - (

C S

p

O N l - H

0 0

en

t s

cs tv

N O

0 0

'*

N O

i n o\ o

l - H

. ^ • N

en d

CO

t

I - H

cn

t S

o l<

b s

I - H

C M

I - H

CM p

O N

I - H

% en

t s

C M

ts

N O

0 0

• ^

N O

i n

O N

o

I - H o

•J3

3

a*

N

"1H

CD

Wi

V

>-,

43

CO

V

-6

C

>-i

0 )

4 - »

0 1

a

(3

^ ' w '

<» t :

m

tt

u I*-*

"C ^ m a* J-

^ •< 5 ja Si >

J3

E 2 S cr t n

lU

tn

.2

P

1 bC

IQ

4 - * c

OJ

c

C M c

m

frH

0 )

121

u

OJ

u

t n

E

6

o o

1—<

E

6

i n

(N

i - i

B

E

m

r ^

E

6

o r4

E

6

o tn

r o

o

H

M - ,

o

t o o

• ^ H

• ^ H

'c?

1

C

[ O t;

CO

-4-»

CD

M - i

CO

> N

i

c a

o o

t H

g

-B

•J3

3

CO

1

5

H - *

[ O

0 1

-6

c

C N

•B

P

cn

O l

-s

e

E

o

V H

E

B

B

B

o

I - H

I - H

E

B

o

C N

E

E

o en

( H

0 )

4 - »

0 1

1

•^H

Q i

CO

I H

0 1

M - I

CO

i-l

0 1

• 4 H

CO

" c O

4 H

o

H

0 1

bO

a

( H

0)

>

<

SEMM-WW-MT-9603 ©1997Dataquest

March 17,1997

I

i i

Silicon Wafer Forecast

>

I

>>

Xi

G

o

•43

Xi

* J

en s

1 s i s s

o o

( N

o

i n

• ^ o

"*

I - H

O)

I - H

CO

m

i n

•^

OS

CO

o o o

0 0

I - H

^^ cs"'

Ov

0 0

CO

o

I—1

i n

^ iri

I - H

OS

' t

CO

0 0

I - i

'^l^ i n

cvi o o o

I - H

I - H cs cs cs" cs

CO

I - H o

'* o

VC i n

ON

VD

I - H

O S

CO

CO

VD

I - H

'* sq

I - H p o o

I - H cs so cs l - H

I - H o o o

tv

d

O S

0(3

0 0 tv

l - H cs

a6

cs

VO

CS

cs so

0 0

d l - H

O S l~N

^

CN

SO

OS

d

CO

O S

P

0 0

l - H

CO

0 0

CM

I - H o\ l - H l - H

i n

d l - H

SO tv

OS

p

so

I - ;

I - H p

OS

• ^

t<

l - H

CO

• ^

rs|

0 0 so

l - H

OS

CN

d

OS

0C3

SO cs

p

so

S

•a

(Q

VH

01

• M

u

a

(B

>^

s s fc s

o

i n

o

^

VO

Q O

GO

I - H

0 0

CO

CO

OS

OS

CO

VO

o o o o o

[ ^ o m

so

I - H l - H

I-H cs

t>:

^ o

I N

so

i n

CO

0 0

iri

CO

CO

o o o o

I - H

0 0

0 0 tv

iri o cs i - <

CO

O S t

!< i—t

0 0

l - H

CM i n

r j H l - H o l - H d

CO

"^

so

I—1

OS

iri o

s o

I - H

^

O S

O t<

i - i

tx

d cs

ts.

l - H l - H

• * p d

»* d

so

OS tv

iri

en

0)

u

l-H

u

OS

cr en

at bO

• M c u

IH

Ol s s s s s s

o o o

DO

O

( N

0 0 r-J

I - H

i n

OS

• ^

I - H

O

CN

cs

i n

S D

1 *

• ^

VO

(N

IV.

OS

'd^

m cs

CN

o

CO

CO

I - H

<* o t - H

• ^ o\

( N

0 0

gj

• ^ p

O S

o o o o o

I - H p o o

I - H o o o

CS

o

in

I - H

CO

I - H

CO

0 0

I - ;

so

I - H

I - ^ "

gp

CN

I - H

^s

Csj

I - H o o o

t>>

I - H

I - H o l - H

CO

cs

S O i - i

l - H

i n

CS)

O S

d l - H

^

K

l - H

SO

d

C S |

0 0

OS

l - H

i n i n rsi

o\

so

OS so

cs l - H

• *

(N

CO

CM

• < ^

CN

o o o

t s

l - H

S O

m iri

so t s

CO

iri

0 0

0 0 i n tv i n

iri

so

cs iri

«v

**

0^

0 ^

(Q

U t i

o

G

o

J3

• F H i i

en

D

01

N

CD

t r

w

a

1

i-i

01 e to

(3

43

Ol t ^

as

"* r H

CO

t N o

t<

t - N

r-l

ci

I - H

CN

O

OS

I-H

M

CO

rsi

CN

rv

so

0 0 so i n

OS

tn

(H

t o

»+^ o to

C o

• ^ H

Uc

to

en

OJ

M-i

to

C

cn

vo 5 -2

^ ^ ^

V

• ^

( Q

1 C

u u

01

Ol

H S ^

D

lU

ro

Ul

V

u c tM tn

V

u

(Z

"" tn

e

J

o

I - H

6

B in

C 4

I - H

g jl i n

I - H

s

J

o

<N

6

a o

e

13

• 4 " * o

H

!

ro

4-»

O

CO

OJ

£

•B

5

Q

IH

u

'B

CO

"2

D

J=

u cs

«*

I - H

CO

O v o

t<

r>. l - H oi l-H cs o

OS

T-H

Q O

CO

( N

^^

so

o6

so i n

C3S

^—^

CO

0 1 en

Ol

u g

^

*~*

CO

o l - H

6

1

CS

I - H

6 d i n

l - H

B d

o cs

I

01

"-M

IH

B d

o

CO to

01 to

5

+-»

O

H

<u

bO to

i-i

01

>

<

u

B

t i

\

I

SEMM-WW-MT-9603 ©ig97Dataquest iVIarch 17,1997

47

48 Semiconductor Equipment, Manufacturing, and Materials Worldwide

?^

J3 i:; o

• I H

-u

•c ri-i

w

S

•d

(0

«H o

s

1

o

O

m

vd

CO

\6

o

Cvj r H

VD

CM

r^ o o o t - l

CO

o

ON

co

I S ! i

o

1 - H o

• ^

t N

o

t v

1 — <

t s

CS

O t s

m

• ^ i - < o o o

1—1

5)

0 0

• > * iri

1—1

2

S

o

1—1 o p

0 0

VO l <

0 0

o

CO

Cs

CM

i n

0 0

o o o o

I—1 o\

E:!

t s

CO

1—1

ss s

o

CN

o

m as

0 0

CT\ t v

cvi

CO

0 0 t s

• *

"^ o p o o

1—1

I—t

NO

O

• > ^

1 — < o o

t s

• ^ o

CO

ON t s

"*

1-5

v - H

0 0

1 - ; d

1—1

t s

CM

i - < i n

NO

o

I—1

d

1—1 i n

1—1

CO

CO

0 0

ON

ON

1—1

1 - H d

NO

N£5

CM

•<1<

CO

\ D

o

I—1 d

0 0

• *

ON

c^

CM

0 0

ON t s

i n o

d

O

• ^

CM

CM

CM

NO

o

CM

d

ON

"*

CO

CO

^%

IS;

CO

Nx5

1—1 o

d

CO

CS

CM

^

P

NO

(Q

m cr

CD u bO

(0

"S

01

u

PUl

>.

43 in

01 u

H H fc

s

o

( N

O

0 0

1 - H

1—1

CO

I - H

1—(

Cv|

IS!

CO

in

ON

CO

o o o o

1—(

CM

NO i n

CO

0 0

1

o

•^ o

'^ i n

1—1

CS

' t

T-(

[ s

a\

CO

CO

o

CO

o o o o

I—1

ON

1 — < m m

CO

1 - 1

s

s

1—1 o

VD

O

o o

CM

ts.

0 0

I—1

CO

1 - H

• ^

CO

OS

I—1 o o c5 o

1 - 1

0 0

^

•^ o\

CN

o

CM d i n i n

CO

CO

ND t s

NO

•«1<

o

CM

i - <

CM

CM

0 0

i n o

CO d

NO

NO

O N

CO

m

t s

CM

CO

o i n i - <

CNl

S

i n

1—1

d

•^ d i n t s i n

•<;h

ON

NO

0 0

I - i

o

CO

1 - <

CN

CO

CS

i n

s

s

I—1

C 3 t>>

o

i n

o cs

I S .

I - H

CM

o

CO

• ^ o

• ^

1—1 o o

d

o

I—1

t

m

CO

CS

CNI

CNl

1 - 1

d rj< d

p

NO

o

•^

NO

iri p

1—1 o o

t s

1—1

•^

1—1 i n

' - ? e

t-^

0^ en

(0 u

&

S

A

3

^ o

M H

G

o

"1H

• ( - •

u

^-»

01

j a

•n

* j

n l

(S

• F H

43

to

D

HI

N

- ^ H

CD

u

M H

l - H

s

(Q y - v

I

^1

2 c« ii g^ S

•« ^ ^

8

3 c en m bo

(Q

.2

P

1

4-»

s

^^

OJ

u

C

• " ^

CM tn lU

-flS 3 ^

CO

^

1—1

H W >

SEMM-WW-MT-9603

t>%

o

[ V

cn

Ol

-C u

C

t v

1—1

CS

1—1

CS

P

CJ\ r - l

0 0

CO

IS!

CM t s

SD

0 0

^

vo i n

ON

o

1—1

Hi

M-. o

s

o

•»H

•»H

ID

t:

CO

^-»

CD

M - t

nS

^

1—t

CO t s

O

is:

t s

1 - 1

CM

1—1

©1997Dataquest

CM

p

ON

1 - 1

B

6 o o

'"'

B

B

i n

CM

1 — '

T - ^

a

B

B

B

o

CM

B

B

3

o

H o

CO

E -

»—<

g

•J3

1

x ; o

1 cn

a

B

•4->

H H

•s

n

01

CM

1

-s

CO

P

t o

<u

B

B

o

1—1 1 - 1

B

B

s

m

1 - 1

0 0

CO

ts:

CM t s

NO

0 0

• < *

NO

i n

ON

o

1—1

B

B

o

CM

B

B

o

CO l a

O

H

I

[O

'tis"

OJ

-5

^

IH

^-•

01

B

ca

• f H

cd

01 r>.

O)

O)

T—

£«

CO

Ol bO to

( H

OJ

>

<

cr

CD

March 17.1997

I

i i

Z661.'Z1.M0JEW

isanbBiBQzeei.®

E096-il/\|-MM-l/\l^3S

CO o c: n o a s.

<D

S-

n

01 t » fD

»

=1

(U

•5

5

01 fD

" i

D

CO

(O

• > !

fD fD

-t

1

H o

r4-

D l i

ST" t?

3 w o

1

3

B

N J o o

5* rtj en

O i to

VO ro o w o

(->

ON

0 0

J-»

. 1 ^ to

4^ j ^

1 — '

It^ o

ON

O l

ON o h o

O l

VO o

k—1

( j i o

3

3

I—*

ro

3

p j

Q

3

f

u

5 " n rr

»—» o

ND

U l

ON

tl^

0 0

ON

^ ro

V I

CO

0 0

I — '

NO o ro

• - 1

^

^ o

^

CO

K-»

ht^

(-«•

o

3

Ul

O

M l

OJ

M l

fD

Ul

c.

to

1 1

3 n

5

Cfl fO cn gcr. o

3

ct

^

^ o> fo

• 1

n

3

3

cn

D J

• - 1

5r

H o r +

C D

01 r rtm

5*

CO

3

3

*>. o

(-»

4 ^

^ cn

• - » h - "

j s . o

IF>. rvj o

3

^ cn

3 ro

O l

3

3

3

3 w to

3" n

3"

0

iy

^ 3

*^ n

3

i-i"

01

fD

a

3 fD

i-H

n w S-g-

HK B) 2 1 m

• — *

o

NO

O l

ON

• ^

0 0

ON

V I to

V I u

0 0

> — 1

NO o ro

» — 1

ro

K - '

V I

V I o

V I

4 i .

ON

CO

;^ v j

1—>

o o o o

CO

h-^

4 ^

X I

•-) fD sT n

3"

Ul cr v;

5

Dl ff

•-J

>

>-«

H Q

^^ p

« n

^ ^

Hh n

cn

M"

N n w

cr

S.

o

13

• T l o

H( n

n

»

VA v£

I

M

I—'

V I

NO

4 ^

NO

V I

h-i

V J bo

(-> t o o to to

0

4 ^

M v e vo

»t^

<-i n

I t

n

^

• - '

ON

>-' o o o o

CO o b i

:& o

4 i o

NO

V ]

h - 1

ON

0 fo

i - i

v s fO

cn

>n

00

ON o

t—*

ro

• ( ^

o

ON bo

0 0 to

*•

ON w

U l

• - »

o o u

0 0

CO

ON

NO

U

l - i

o o o o

45.

V I

V I

CO to

CO

>-» to

V I

OS ro

(-' b p

t-i

VO

0 \ k l

VO ro to bo o

VD o

O N

ON

>(^ o w ro o bo o t o

O N

4 ^

0 0 t o

ON j ^ o

t - " t - "

ON

V J o

CO

NO

CO

• - »

o

^ o t o

O N

OS

OS

CO o o o o

CO

I - '

-f^ bo

^ fe.

CO

NO

CO

• - >

o

^ o

ON

V I o3

NO

>-> o o

Q o

>-» v j

NO

00

V ]

ro

I - '

8 o o

O l

NO o ro

4 i b i

(-> o w

O l

CO o bo p

• - '

s

O l

|:i o oo

O l

4:^.

U l p

ON p

H - 4

ro o o

( - J

o

ON

(-> o o o o

CO

ON

NO o

1 — '

NO b i

V I

f - J

CO

ON o b i

0

H I va vs

^

H I

VD

H I

VO

VO

VO

>-i n

I—t

3

s-

n

cr

a

I

I

ON

s

CO en o p

K - i

• - » f - "

0 0 k>

0 0

U l

>t^ ro

CO

CO o

V I o

CT\

VD

»^

CO

0 0 t o p

l-i

0 0 ro ro i^

0 0

t-i

>f>. o

CO o p

ON o to o o

(—>

"ro

O l

O l o o o

h-i

o

V I o

O l ON

ON

CO

CO ro o

*> 0 j ^

i - i

^;&

0 0

l - »

o o o

1 — '

4 ^ va

O l t o

4 ^

O l ro t o b i o

U)

0

1 0

0

0

0 t«) a

0

H t

CO r<-

g=

&. o

53

cr

^SeOSJOd JS^BM U03||!S

6t'

50

Semiconductor Equipment, Manufacturing, and Materials Worldwide

Table 4-9

Worldwide Merchant Silicon Wafer Revenue Forecast, 1994-2001 Includes Polished,

Virgin Test, and Epitaxial Silicon (Millions of U.S. Dollars)

CAGR (%)

1995-2001

14.0 Polished Silicon*

Percentage Growth

Epitaxial Wafers

Percentage Growth

Worldwide

Percentage Growth

1994

3,252

26.2

1,331

35.0

4,583

28.6

'Includes prime and virgin test/monitor wafers

Source: Dataquest (January 1997)

1995

4,483

37.9

1,815

36.4

6,298

37.4

1996

5,217

16.4

2,188

20.5

7,405

17.6

1997

5,421

3.9

2,397

9.6

7,818

5.6

1998

6,259

15.5

2,836

18.3

9,095

16.3

1999

7,439

18.8

3,352

18.2

10,790

18.6

2000

8,665

16.5

4,438

32.4

13,103

21.4

2001

9,860

13.8

5,680

28.0

15,540

18.6

20.9

16.2

Our revenue forecast has factored an 8 to 10 percent price decline for wafers starting in 1997. However, product migration to 200mm wafers and a higher epitaxial wafer mix will mostly offset this price decline on a per-square-inch basis. These same mix issues will continue to move toward higher silicon manufacturer revenue per square inch, resulting in a

CAGR for revenue five points higher than the area CAGR.

Silicon Unit Forecast Overview

We anticipated that the second half of 1996 would be weaker in silicon demand, based on the migration from 4Mb to 16Mb DRAMs mentioned in earlier forecasts. However, starting about August, the silicon market flat out collapsed, with run rates by the end of 1996 about 20 percent below those of just six months earlier. These dynamics can mostly be explained by the activities in the DRAM market and inventory trends. As a result, the long-term absolute MSI shipment level was reduced about 5 percent compared to our last forecast, with the six-year CAGR at 11.1 percent.

As we have noted in prior publications, the industry migration from 4Mb to 16Mb DRAMs would cause a dramatic reduction in the amount of silicon required per bit. Die size relationships mean that the average 16Mb

DRAM has two to three times more bits per square inch than the 4Mb generation. Even in the face of high bit demand growth, the industry does typically see a short period of time in which square inch requirement actually decreases.

This means that wafer starts should actually decline for a period of two to three quarters as a result of this silicon efficiency. Indeed, this can be shown to be true, as we have performed a quarterly analysis of the relative square inch consumption by the DRAM market. These normalized figures are shown in Table 4-10.

As Table 4-10 illustrates, 14 percent less silicon is required in third quarter

1996 compared with fourth quarter 1995 to support a higher DRAM bit demand. Also, not taking inventories into consideration, wafer starts should have declined into the DRAM sector during first quarter 1996. Naturally, this did not happen.

SEMM-WW-MT-9603 ©1997 Dataquest Marcli 17,1997

Silicon Wafer Forecast 51_

>

>

>

Table 4-10

Relative Silicon Area Consumed by DRAM, Fourth Quarter 1995 through 1997

Q4/95 Ql/96 Q2/96 Q3/96 Q4/96 Ql/97 Q2/97 Q3/97 Q4/97

Percentage of Q4/95 Wafer Area 100.0 89.1 86.1 86.0 97.5 99.2 106.0 120.9 132.8

Consumption Rate

Ratio of 16Mb versus 4Mb (Units) 0.30 0.50 0.63 0.87 1.15 1.55 2.09 2.76 3.37

Source: Dataquest (January 1997)

Optimism about demand for DRAM and fear of a silicon shortage meant

DRAM manufacturers continued to produce in the first half, expanding

DRAM chip and wafer bank inventories as well as keeping inventories of raw silicon wafers high. This corrected during third quarter 1996. Silicon area growth for 1996 will end up at about 7 percent growth, in terms of MSI.

Theoretically, silicon consumption into the DRAM sector should be recovering now. However, actual shipments from wafer manufacturers will lag by four to six months as inventories are worked down. We do expect recovery in silicon MSI shipment growth by the middle of 1997. Our quarterly model into the DRAM sector suggests run rates by fourth quarter

1997 should be 20 to 30 percent higher than current depressed levels. Our

MSI growth forecast overall for 1997 is just over 7 percent. With end-use semiconductor and electronic equipment demand strengthening, we would expect accelerated growth in silicon consumption starting in the middle of 1997 and returning to double-digit MSI growth rates in 1998.

The 200mm Wafer Ramps Up—Suppiiers Have Responded with Capacity

Dataquest has been studying the subject of 200mm wafers and their ramp rate closely over the last three years, particularly in light of the massive announcements in fab capacity and requirement through the rest of the decade. We are expecting more than 200 fabs will be online by the year

2000 processing 200mm wafers. Future demand for 200mm product wafers has been affected positively by the semiconductor industry's excess capacity, and have not been reduced significantly since our last forecast in mid-1996. Why? Semiconductor companies now have some time to convert 150mm fabs to the more cost-effective 200mm wafers. This has added incremental product wafer demand to the industry at 200mm wafers. A large factor in 200mm demand, the consumption of test wafers, has remained essentially unchanged since our last update. The industry average is about one test wafer per seven product wafers for sizes below

200mm. By the year 2000, we now expect this ratio to be one test wafer for per over three product wafers at 200mm.

Wafer suppliers have answered the need with several new wafer plants

' and billions of dollars of committed investment during 1996. We continue to believe that the ramp of 200mm wafers will be supply constrained from time to time through the decade, although the faster-than-anticipated conversion to the 16Mb DRAM will mean adequate supply to meet demand in 1997 and 1998. While we are stopping short of calling a shortage of

200mm wafers to the point of restricting ramp-up plans, we would expect buyers of wafers will experience firm to rising prices again by 1999, and rjiay be placed on allocation from time to time.

SEI\/ll\/l-WW-l\/IT-9603 ©1997 Dataquest IVIarcii 17,1997

52 Semiconductor Equipment, Manufacturing, and Materials Worldwide

What about 300mm Wafers?

There has been a significant amount of activity in 300mm wafer development recently. A new consortium of Japanese companies, Selete (Semiconductor Leading Edge Technologies Inc.), has been formed to spearhead the

300mm technology development. We understand equipment began shipping into the clean room setup at a facility made available by Hitachi during the second half of 1996. SEMATECH has also coordinated an effort that has companies participating from other regions outside of Japan, called the International 300mm Initiative (I300I). I300I recently completed a survey that concluded that there would be as many as 11 pilot facilities on line running 300mm wafers by the year 2000, the majority starting during 1999. Although we believe that many companies have committed internally to build such pilot plants, it was only very late in 1996 where the first public announcement took place. Surprisingly the plant is to be located in Europe, a Texas Instruments facility in Italy, to be specific. We believe that during the first half of 1997, many more such announcements will be made, with most of the activity centered in Japan.

Silicon wafer manufacturers have noticeably ramped commitment, as at least five different companies have announced or have committed to dedicated pilot lines for 300mm wafer production. Sample volumes have been available for a couple of years already. The earliest of these starts production in late 1997, with the latest coming on during 1999.

We believe the first commercially productive plant will be started in the

2001 through 2003 time frame (after the feasibility noted above), with serious volume ramp up in the years 2003 through 2005. This would be consistent with shrink 0.25-micron technology being primarily produced on

300mm wafers. This means that 200mm wafers represent at least a two technology generation wafer size, and fabs being built today may have longer lives than history would indicate.

Our current forecast reflects the increased activity at the R&D and pilot facility level, and is likely somewhat conservative. Prices are still being quoted in the $1,000 to $1,300 per wafer area for mechanical test samples with prime quality wafers still significantly higher. This market is expected to be supply constrained for qviite some time.

Further details and issues regarding the move toward 300mm wafers will be included in a Dataquest Perspective newsletter in the next couple of months.

Epitaxial Wafer Trends: Are ORAMs in tlie Future?

Sales of merchant epitaxial wafers by the wafer suppliers accounted for

477 MSI in 1995. About 62 percent of these wafers were used for CMOS logic applications while only 2 percent were used for DRAM products.

The remainder were shipped into the power/discrete device segments. By the year 2000, Dataquest expects that fully 21 percent of merchant epitaxial silicon will be used for DRAMs, primarily driven by two factors. The

CMOS logic application remains dominant at 52 percent. The overall epitaxial wafer market will experience a 19.4 CAGR from 1995 through 2001.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Silicon Wafer Forecast 53

^

^

i

First, trench capacitor designs benefit from the use of epitaxial silicon. It is believed that both IBM and Siemens are currently using epitaxial silicon while Toshiba has been using hi-wafers (wafers that include a hydrogen bake surface treatment to reduce surface contamination) to improve the performance of their trench capacitor design. We believe that, for lategeneration shrink versions of 16Mb and for 64Mb DRAMs, companies that employ the trench design (which include the companies noted plus Texas

Instruments and affiliated fabs, perhaps including Hitachi) will need to use epitaxial silicon.

Second, even in the stacked capacitor design, some companies (primarily in Japan) are finding that epitaxial silicon can be used to enhance yield by eliminating the very localized silicon "pit" defect which has recently become important at the 0.25-micron level technology in larger wafers

(200mm and 300mm). Fujitsu is believed to be seriously evaluating and nearly committed to the use of epitaxial silicon, and several other Japanese companies, including Matsushita and Oki, also have epitaxial silicon under consideration for DRAMs.

We will watch this area of device construction technology very closely over the next several years, as there is a distinct possibility that this need for epitaxial silicon will be a single-generation need. At the 0.18-micron generation, the use of other capacitor dielectrics and electrode materials may enable companies to return to planar or simple stack structures, perhaps eliminating the driving need for the epitaxial layer.

Our semiconductor manufacturing analyst in Japan is currently preparing a Perspective newsletter that will describe the issues and offer a low-end and high-end forecast for epitaxial wafer use in the DRAM segment.

Update of Historical Merchant Epitaxial Marlcet Estimates

From time to time, Dataquest refines forecast and data collection methodologies in order to provide greater value to our clients. Whenever we implement these changes, we run the risk of the new model adding new information, which requires we restate historical figures. With the implementation of new data collection methods with regard to the power/discrete segment, and an updated forecast model for CMOS logic,

Dataquest has restated the merchant epitaxial wafer MSI figures back to

1991. These changes will be reflected fully in the Silicon Market Share publication expected for a Jime 1997 release.

Highlights of the North American Siiicon Wafer Market and Forecast

Silicon consumption in North America is forecast to grow 13 percent in

1996 to 1,064 MSI, and we expect a more mild 9 percent growth in 1997.

Microprocessor and other logic chip demand have been and will continue to be key drivers behind increased silicon demand in North America, and epitaxial wafer demand will be focused on CMOS logic.

Merchant epi wafer consumption increased 17 percent in 1996 to 267 MSI, driven in large part by microprocessor manufacturers, such as Intel and

AMD, which build their microprocessors on epi wafers. Weakness in the automotive and discrete segments of the chip market significantly slowed

SEIVIIVI-WW-IVIT-geOS ©1997 Dataquest IVIarch 17,1997

54 Semiconductor Equipment, IVIanufacturing, and Materials Worldwide the market compared to 1995 growth levels. By 2000, epitaxial silicon will account for 27 percent of the square inches consumed in North America— the highest concentration in any region.

Dataquest's longer-term forecast for North American silicon consumption remains firm, primarily because we believe that the United States will attract a larger share of foreign multinational fabs, such as those currently being built by Hyvmdai, TSMC, Samsung, and others expected to be announced over the next couple of years. We are projecting that total silicon MSI will grow at an 11.9 percent CAGR for the time period 1995 through 2001.

Highlights of the Japanese Silicon Wafer Market and Forecast

Our Japan silicon consumption forecast calls for growth of just over 1 percent to 1,503 MSI in 1996 as the bottom fell out of the DRAM-sensitive region. We expect moderate growth of 6 percent in 1997, recovering to double-digit expansion again in 1998 as the industry completes its migration to 16Mb DRAMs.

Unlike North America, with its sizable CMOS epi wafer market, Japan's merchant epitaxial wafer market is more focused on discrete and bipolar applications today. However, future growth prospects may be more closely linked to the application in the DRAM device. Because we believe that a number of suppliers will migrate toward epitaxial silicon for shrink

64Mb DRAMs, Japan is among the fastest-growing epitaxial markets with a CAGR of over 23 percent. Near-term economic weakness, however, may put a lid on this growth, but we are still calling for an above average 14 percent growth for epitaxial demand in 1997, after a 16 percent growth in

1996, and a 23 percent increase in 1995.

Dataquest remains moderately conservative with its longer-term growth scenario for silicon wafer demand in general in Japan, as the country continues to work through the evolution of its semiconductor production infrastructure (seen as being too heavily dependent on the DRAM).

Investment patterns during 1995 suggested that Japanese semiconductor manufacturers were willing to come to the table and invest, preserving their stake in the memory business against the Koreans; yet Japanese companies were among the first to pull back during 1996. Also, through the use of joint ventures and other alliances, we expect that Japanese company production will gradually depend less on domestic production. We are estimating that silicon demand will grow 8.5 percent CAGR for the years

1995 through 2001, the slowest growth of all regions.

Highlights of the European Silicon Wafer Market and Forecast

Demand for silicon wafers in Europe, as well as wafer fabrication equipment, remains heavily dependent on the fab activities of foreign semiconductor firms. With increased presence from European and Korean companies, the outlook for silicon consumption has brightened.

The European market grew 14 percent in 1995 to 519 MSI. European silicon demand will moderate significantly in 1997, to the 8 percent growth level. Siemens Dresden and other DRAM production, and multinationals

SEMM-WW-l\/IT-9603 ©1997 Dataquest Marclil?, 1997

Silicon Wafer Forecast 55

I

I

Intel, Motorola, Hyundai, LG Semicon, and Texas Instruments will continue to ramp to answer the demand in Europe, helping silicon consumption grow at healthy double-digit levels from 1998 through 2000. The longterm picture for Europe, with a 12 percent CAGR through the year 2001, places this region as the second fastest growing, narrowly outpacing

North America.

Epitaxial wafer demand in the region increased dramatically in 1996 to 109

MSI, growing 50 percent. This has been a direct result of Intel in Ireland,

Philips, IBM in France, and Siemens ramping DRAM production in Germany. In the near term, epitaxial wafer demand increases will come from predominantly European producers, primarily from the power/discrete.

Later in the decade, the DRAM capacity ramps and the migration to epitaxial silicon in DRAMs from multinationals with fabs in Europe will drive the epitaxial market to a CAGR over nine percentage points above polished bare wafer consumption through the decade.

Highlights of the Asia/Pacific Siiicon Wafer IMarket and Forecast

Silicon consumption grew at an 8 percent pace in 1996, as the DRAM cutbacks affected the region in the second half. This growth is somewhat misleading as about 70 percent of the growth came from test and monitor wafer consumption because of the larger mix of 200mm wafer consumption. DRAM overcapacity will continue to put a significant drag on demand in early 1997, with growth expected to be under 7 percent (with over half of the growth again in the test and monitor wafer). Past 1997, the trend will resume that Asia/Pacific wiU represent the fastest-growing region as DRAM driven consumption will return by 1998 and new fab projects will be revitalized beginning in 1999. The six-year CAGR forecast of 14.6 percent is 3.5 points over the worldwide market rate.

Epitaxial wafer consumption will grow at the fastest rate in the Asia/

Pacific region as well with a CAGR of 29 percent. However, this comes from, and will remain, a low base of consumption. While the foundries will require a significant portion of their capacity to consume epitaxial wafers and remain the primary growth driver for epitaxial wafers in the region, the DRAM fabs in the region do not use epi wafers. With the exception of a few joint venture Tabs, the epi consumption in Asia/Pacific

DRAM capacity will remain quite low compared to other regions. Epitaxial wafer consumption will remain below 9 percent of total Asia/Pacific

MSI by the year 2000 (compared to 18 percent worldwide).

Dataquest Perspective

We anticipated that the second half of 1996 would be weaker in silicon demand, based on the migration from 4Mb to 16Mb DRAMs mentioned in earlier forecasts. However, starting about August, the silicon market flat out collapsed, with run rates by the end of 1996 about 20 percent below those of just six months earlier. These dynamics can mostly be explained by the activities in the DRAM market and inventory trends. As a result, the long-term absolute MSI shipment level was reduced about 5 percent compared to our last forecast, with the six-year CAGR at 11.1 percent.

SEI\/ll\/l-WW-l\/IT-9603 ©1997 Dataquest IVIarch 17,1997

56 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Theoretically, silicon consumption into the DRAM sector should be recovering now. However, actual shipments from wafer manufacturers will lag by four to six months as inventories are worked down. We do expect recovery in silicon MSI shipments by the middle of 1997. Our quarterly model into the DRAM sector suggests run rates by fourth quarter 1997 should be 20 to 30 percent higher than current depressed levels. Our MSI growth forecast overall for 1997 is just over 7 percent. With end-use semiconductor and electronic equipment demand strengthening, we would expect accelerated growth in silicon consumption starting in the middle of

1997 and returning to double-digit MSI growth rates in 1998.

The silicon market, driven by a strong long-term picture for semiconductor unit demand in general, will grow faster over the next six years than the last six years. As the industry transforms into a 200mm baseline, the outlook for silicon wafer manufacturers becomes brighter. Silicon manufacturers have answered the call for 200mm capacity, with significantly increased capital outlays. Activity in 300mm wafer development has accelerated, particularly in Japan with the Selete consortium. Increased visibility of I300I has also contributed to the first 300mm pilot fab armouncement. We still expect only pilot volume activity in 300mm wafers through the year 2000.

Sales of merchant epitaxial wafers by the wafer suppliers accounted for

477 MSI in 1995. About 62 percent of these wafers were used for CMOS logic applications while only 2 percent were used for DRAM products.

The remainder was shipped into the power/discrete device segments. By the year 2000, Dataquest expects that fully 21 percent of merchant epitaxial silicon will be used for DRAMs, primarily driven by two factors. The

CMOS logic application remains dominant at 52 percent. The overall epitaxial wafer market will experience a 19.4 CAGR from 1995 through 2001.

The silicon market has become recognized again as being strategic in the semiconductor manufacturing infrastructure. Will this continue? We believe it will, as long as silicon suppliers continue to concentrate on value-add processes and techniques as the equipment manufacturers have done, as well as adequately and smartly plan capacity additions.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Chapter 5

Semiconductor Consumption Forecast

This chapter presents data on the worldwide semiconductor market by region. The regional semiconductor market, or regional semiconductor consumption, deals with where chips are consiomed; this contrasts with regional semiconductor production, which deals with where chips are manufactured. The data presented here is for the merchant market and does not include the value of chips made by captive semiconductor manufacturers for internal use.

This is an excerpt from the Semiconductor Five-Year Forecast, published by Dataquest in October 1996 (SEMM-WW-MT-9603). Further details regarding this forecast can be found in that publication.

Yearly exchange rate variations can have a significant effect on the 1989 through 1996 data in the following tables. For more information about the exchange rates used and their effects, refer to Appendix B.

Semiconductor Consumption

Table 5-1 shows revenue and growth from semiconductor shipments for the years 1989 through 1995 broken down by region. Table 5-2 shows revenue and growth from semiconductor shipments for the years 1995 through 2001 broken down by region of consumption.

Table 5-1

Worldwide Semiconductor Consiimption by Region—Historical; Includes Merchant

Semiconductor Companies Only (Millions of Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Europe, Middle East, and

Africa

Percentage Growth

Asia/Pacific

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (January 1997)

1989

17,070

7.7

21,491

3.5

9,498

1990

16,540

-3.1

20,257

-5.7

10,415

1991

16,990

2.7

22,496

11.1

11,014

1992

20,430

20.2

20,579

-8.5

12,218

1993

27,926

36.7

24,645

19.8

15,461

1994

35,939

28.7

31,010

25.8

20,819

1995

48,349

34.5

42,164

36.0

28,341

CAGR (%)

1989-1995

18.9

11.9

20.0

11.9

6,280

9.2

54,339

6.8

9.7

7,333

16.8

54,545

0.4

5.8

9,194

25.4

59,694

9.4

10.9

12,034

30.9

65,261

9.3

26.5

17,486

45.3

85,518

31.0

34.7

22,812

30.5

110,580

29.3

36.1

32,417

42.1

151,271

36.8

31.5

18.6

SEMM-WW-MT-9603 ©1997 Dataquest 57

58

Semiconductor Equipment, Manufacturing, and IVIaterials Worldwide

Table 5-2

Worldwide Semiconductor Consumption by Region, Merchant Semiconductor Sales

Only—Forecast (Millions of U.S. Dollars)

Americas

Percentage Growth

Japan

Percentage Growth

Europe, Middle East, and

Africa

Percentage Growth

Asia/Padfic

Percentage Growth

Worldwide

Percentage Growth

Source: Dataquest (January 1997)

1995

48,349

34.5

42,164

36.0

28,341

36.1

32,417

42.1

151,271

36.8

1996

44,177

-8.6

36,574

-13.3

27,136

1997

50,440

14.2

40,476

10.7

29,642

1998

62,330

23.6

48,368

19.5

34,951

1999

79,222

27.1

58,842

21.7

42,225

2000

97,714

23.3

70,483

19.8

53,099

2001

112,469

15.1

78,095

10.8

60,214

-4.3

29,090

-10.3

136,977

-9.4

9.2

34,047

17.0

154,605

12.9

17.9

43,273

27.1

188,922

22.2

20.8

54,627

26.2

234,916

24.3

25.8

68,923

26.2

290,219

23.5

13.4

80,157

16.3

330,936

14.0

CAGR (%)

1995-2001

15.1

10.8

13.4

16,3

13.9

SEMM-WW-MT-9603 ©1997 Dataquest

March 17,1997

^

t i

Chapter 6

Semiconductor Production Forecast

This chapter presents data on the worldwide semiconductor production by region. Semiconductor production is defined by the place where the wafers are fabricated, and regional semiconductor production includes all production in the region, including merchant and captive producers, and all foreign producers. For instance, American semiconductor production includes Digital Equipment and Delco fabs as well as Japanese company and European company fabs in America.

Yearly exchange rate variations can have a significant effect on the data up to 1996 in the following tables. For more information about the exchange rates used and their effects, refer to Appendix B.

The semiconductor industry has a global manufacturing infrastructure.

Production of semiconductors is constantly shifting among regions, as new capital money is flowing toward areas of relative lower capital cost, and higher growth areas of consiimption. Dataquest reviews some of the trends, and potential impacts, for the future.

Historical Semiconductor Production

Table 6-1 shows the historical semiconductor production for the years

1989 through 1995 broken down by region. Dataquest follows a methodology that employs the use of our fab database, estimating the memory, microcomponent, logic, and analog/power/discrete production components separately, and estimating net production among regions for foundry activity. This approach provides insight in observing and forecasting production trends for front-end manufacturing.

Because of the reclassification of the MOS portion of IBM Microelectronics' business as merchant, the captive production figures changed dramatically in 1993. However IBM's bipolar production, which is consumed internally, is still classified as captive by Dataquest.

Captive Semiconductor Production

Semiconductor production from captive manufacturers is estimated to be

$1.81 billion in 1995, down from just under $2 billion in 1994. IBM has restructured and entered the merchant semiconductor market as of 1993.

Dataquest has reclassified IBM's MOS semiconductor production to merchant, but the bipolar products (exclusively used internally) are still reported as captive. This part of IBM's business will be converted to MOS over the next three to five years, and had resulted in a lower figure for captive production in 1995 and future years.

Many captive producers may consider the move to merchant to better take advantage of the worldwide growth of semiconductors, leveraging their sunk costs in plant and equipment for higher return in a larger end-user base. Still others may elect to take advantage of the now evolving and maturing foundry business, electing to contract out their manufacturing rather than invest in expensive new facilities for their relatively small

SEMM-WW-IVIT-geoa ©1997 Dataquest 59

60 Semiconductor Equipment, Manufacturing, and Materials Worldwide production base. We have not, however, included any such movement to merchant or fabless in our captive production forecast. The four largest captive producers accounted for nearly 70 percent of the $1.81 billion in

1995. These producers are IBM (bipolar only), Delco Electronics, Digital

Equipment, and Northern Telecom.

Clearly, the foundry industry is real and represents a key change in the semiconductor industry manufacturing infrastructure. We continue to expect U.S. companies in particular, both fabless and IDM companies, to continue to expand their use of foundries in Asia/Pacific and Japan for semiconductor manufacturing.

Table 6-1

Worldwide Semiconductor Production by Region—Historical; Merchant and Captive

Semiconductor Company Sales (Millions of U.S. Dollars)

Total Americas

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Japan

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Europe, Middle East, and Africa

Merchant

Captive

Percentage Growth

Percentage Worldwide

Total Asia/Pacific

Merchant

Captive

Percentage Growth

Percentage Worldwide

Worldwide

Percentage Growth

Merchant

Percentage Growth

Captive

Percentage Growth

NA = Not applicable

Source: Dataquest (January 1997)

1989

22,232

18,464

3,768

8.3

37.6

28,527

28,119

408

6.7

48.2

6,451

1990

24,202

20,453

3,749

8.9

40.8

26,384

25,977

407

-7.5

44.5

6,350

1991

26,039

22,275

3,764

7.6

40.4

28,338

27,925

413

7.4

44.0

6,979

1992

29,207

24,998

4,209

12.2

41.5

28,273

27,914

359

-0.2

40.1

8,589

1993

32,643

30,942

1,701

11.8

37.3

35,515

35,515

0

25.6

40.6

11,741

1994

39,617

37,887

1,730

21.4

35.2

45,289

45,289

0

27.5

40.3

15,463

3.3

59,184

7.6

54,339

6.8

4,845

17.4

5,782

669

10.2

10.9

1,974

1,974

N A

5.7

4.0

59328

0.2

54,545

0.4

4,783

-1.3

5,723

627

-1.6

10.7

2,392

2,392

NA

21.2

4.8

64,453

8.6

59,693

9.4

4,760

-0.5

6,396

583

9.9

10.8

3,097

3,097

NA

29.5 41.8

6.2

70,460

9.3

65,260

9.3

5,200

9.2

7,957

632

23.1

12.2

4,391

4,391

NA

11,421

320

36.7

13.4

7,636

7,636

NA

73.9

8.7

87,535

24.2

85,514

31.0

2,021

-61.1

15,243

220

31.7

13.7

12,095

12,095

NA

58.4

10.8

112,464

28.5

110,514

29.2

1,950

-3.5

20,551

160

33.9

13.5

21,732

21,732

NA

79.7

14.2

153,084

36.1

151,271

36.9

1,813

-7.0

1995

49,535

47,882

1,653

25.0

32.4

61,106

61,106

0

34.9

39.9

20,711

CAGR (%)

1989-1995

14.3

17.2

-12.8

13.5

13.8

-100.0

21.5

23.5

-21.2

49.2

49.2

17.2

18.6

-15.1

SEMM-WW-MT-9603 ©1997 Dataquest

March 17,1997

Semiconductor Production Forecast 61

Tlie Move toward Asia Continues after a Pause, European Growtii Rests

The production trends of the last three years may contain two or three surprises to some. Of no surprise is the strong growth in Asia/Pacific production, breaking over 14 percent of worldwide production in 1995. The strength in Asian DRAM producers and the emergence of the foundry market have been, and will continue to be, in the long term the key drivers in that growth. Expected regional capital spending and electronic equipment production will certainly keep this long-term production trend in place, however the dramatic price collapse in DRAM as well as foundry wafer pricing pressures have meant a temporary pause in the trend today.

European production has expanded from just under 11 percent of the semiconductors produced in 1991 to 13.5 percent in 1995. This is additionally remarkable in that the last four years have been good overall growth years, resulting in near tripling of the region's production in four years.

Why the move to Europe? With the region's economies recovering and with the PC boom in the early 1990s, Europe has attracted PC production, particularly in the United Kingdom. Semiconductor production has moved along with the PC, with Intel and DRAM producers worldwide taking part. Also the acceleration of telecommunications related semiconductor production benefits the European companies. Data quest believes that while multinationals will continue to invest heavily in Europe, the trend is currently in a holding pattern in part because of the concentration of current strategic spending growth in America, and the recent pause in growth of electronic equipment production.

There has been a significant relative decline in the percentage of the world's production being done in the United States. Over the last three years. North American production decreased from about 42 to 32 percent of the world's production overall. This is despite the fact that U.S. companies have increased their share of the worldwide semiconductor market in the same time period. Several factors are at work here. First, North American multinational companies have been investing heavily overseas. North

America has been a net exporter of capital for several years now, as foreign companies have yet to balance the scales with investments inside the

United States. This trend should stabilize over the next several years, as

Japanese, Taiwanese, and Korean companies have started to accelerate their investment in the United States.

Second, while U.S. companies are recognized as technology leaders, they have recently begun calling upon foreign producers to manufacture their products in the fotmdry market. While fabless companies have been the key driver of the market to this point, starting in 1994 we saw a major shift in the "integrated device manufacturer," or IDM (a merchant supplier of semiconductors that has a fab), to increase use of foundries. This imbalance in the concentration of foundry capacity is starting to make U.S. semiconductor companies a little nervous, and have actually impressed upon key foundry suppliers to begin building production in the United

States. In fact, TSMC announced plans to build a major fab in the United

States, and we believe that other foundry companies are likely to follow.

Clearly, users of foundry have stated a preference for close access to the fab. Any foundry provider who has capacity in the United States is likely to have a more stable customer base than those that do not.

SEMI\/l-WW-l\/IT-9603 ©1997 Dataquest IViarch 17,1997

62 Semiconductor Equipment, Manufacturing, and Materials Worldwide

And third, while Japanese and European companies have invested somewhat outside their own country, these companies have remained "patriots of the domestic economy" and have kept the vast majority of investment within the region, with perhaps the exceptions of NEC and SGS-Thomson.

This, along with the strong DRAM market over the last three years, has stabilized the Japanese production proportion over the last two years, keeping the same approximate 40 percent share of the production market.

However, as we now fully realize, the DRAM market is cyclical and Japanese foundries will feel pressure from Asian producers, so we expect a resumption of the gradual decay in the base of production in Japan through the rest of the decade.

Semiconductor Production Trends: Acceierating Siiift to Asia/Pacific

Table 6-2 shows forecast semiconductor production by region for the period from 1995 through 2001. The major trend is the growth of the Asia/

Pacific region at mostly the expense of Japan. By 2001, Dataquest believes that Asia/Pacific-ROW will expand to about 18 percent on a revenue basis.

North America will remain steady on a percentage basis, as lower cost of capital and clear leadership in technology and innovative design motivate companies to invest in the United States. Also, Asian companies have begun a regional production diversification program, with several major fabs earmarked for the United States. Europe production share is expected to expand slightly with a product mix shifting to contain a higher memory component, driven by the need for proximity to PC production, as more

DRAM capacity is added by large internationals very late in the decade.

Japan's share of production is likely to continue to erode, as the cost of capital remains high and as Japanese companies increasingly invest in capacity overseas.

Dataquest Perspective

Where the PC and telecommunications production goes, so goes semiconductors. This is true from the perspective of the business forecast, as well as the production line. Europe and Asia/Pacific, with very large capital spending upticks over the last several years, and expected to continue that trend, will continue to gain share in world production over the next several years.

The shifts and currents in semiconductor production trends mean that equipment and material suppliers will absolutely need a global presence in every sense of the word to remain competitive in the market. Product supply can no longer depend on local trends, as all major semiconductor companies have made it clear they are investing on a worldwide basis.

However, local service and support is a requirement to maintain customer satisfaction.

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Semiconductor Production Forecast

63

Table 6-2

Worldwide Semiconductor Production by Region—Forecast; Merchant and Captive

Semiconductor Company Sales (Millioiu of U.S. Dollars)

Total Americas

Merchant

Captive

Percentage G r o w t h

Percentage

Worldwide

Total Japan

Merchant

Captive

Percentage Growth

Percentage

Worldwide

Total Europe, Middle

East, and Africa

Merchant

Captive

Percentage Growth

Percentage

Worldwide

1995

49,535

47,882

1,653

25.0

32.4

61,106

61,106

0

34.9

39.9

20,711

20,551

160

33.9

13.5

Total Asia/Pacific

Merchant

Captive

Percentage Growth

Percentage Worldwide

Worldwide

21,732

21,732

NA

79.7

14.2

Percentage Growth

Merchant

Percentage Growth

Captive

Percentage Growth

NA = Not applicable

Source: Dataquest (January 1997)

153,084

36.1

151,271

36.9

1,813

-7.0

1996

47,701

46,161

1,540

-3.7

34.4

52,599

52,599

NA

-13.9

37.9

19,152

19,040

112

-7.5

13.8

19,177

19,177

NA

-11.8

13.8

1997

54,549

53,030

1,519

14.4

34.9

58,286

58,286

NA

10.8

37.3

21,570

21,490

80

12.6

13.8

21,799

21,799

NA

13.7

14.0

138,629

-9.4

136,977

-9.4

1,652

-8.9

156,204

12.7

154,605

12.9

1,599

-3.2

1998

67,072

65,556

1,516

23.0

35.2

68,957

68,957

NA

18.3

36.2

26,308

26,260

48

22.0

13.8

28,149

28,149

NA

29.1

14.8

1999

81,432

79,871

1,561

21.4

34.4

84,335

84,335

N A

22.3

35.7

32,918

32,888

30

25.1

13.9

37,821

37,821

NA

34.4

16.0

2000

97,674

96,062

1,612

19.9

33.5

103,608

103,608

NA

22.9

35.5

40,651

40,631

20

23.5

13.9

49,918

49,918

NA

32.0

17.1

2001

109,880

108,216

1,664

12.5

33.0

116,158

116,158

NA

12.1

34.9

46,674

46,662

12

14.8

14.0

59,899

59,899

NA

20.0

18.0

190,486

21.9

188,922

22.2

1,564

-2.2

236,507

24.2

234,916

24.3

1,591

1.7

291,851

23.4

290,219

23.5

1,632

2.6

332,612

14.0

330,936

14.0

1,676

2.7

CAGR (%)

1995-2001

14.2

14.6

0.1

11.3

11.3

14.5

14.6

-35.1

18.4

18.4

NA

13.8

13.9

-1.3

SEi\/ll\/l-WW-MT-9603

©1997 Dataquest; IVIarchI?, 1997

64 Semiconductor Equipment, Manufacturing, and Materials Worldwide

Taiwan is clearly the new major production growth area. We would expect Malaysia and Thailand to be the next major growth countries in three to five years. Evidence of this includes recent joint venture fab announcements by Texas Instruments and others. Silicon plants are now being strategically placed, such as SEH's Malaysian plant and announced joint venture in Taiwan, Komatsu's joint venture with Formosa Plastics in

Taiwan, and MEMC's joint ventures in both Korea (Posco-Hiils), Taiwan

(Taisil), and Malaysia (MEMC-Kulim).

Further, the concept of contract manufacturing in semiconductors is clearly here to stay. Equipment and material suppliers could find themselves selling their technical products to an international team from several companies, including the manufacturer and the designer. However, the emergence of the dedicated foundry company, taking ownership of the process and manufacturing flow, will tend to centralize this activity.

Dataquest has started the Semiconductor Contract Manufacturing research program and will continue to explore the key trends in contract manufacturing and foundries, including technology trends and supply/ demand balance through the decade.

i

I

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

i

Appendix A

Macroeconomic Outlook: Fourth Quarter 1996 ^^^^—^

World Outlook: Global Economy at Dawn of New Era

The worid economy appears to be witnessing the dawn of a new era as it enters the final quarter of 1996 and drives headlong into 1997. The makings of this new era have been in the works for some time now. The collapse of communism, the spread of digital technology, the ideology of smaller, more efficient, and fiscally responsible government, the global integration of economic activity, and the aging of populations in the world's developed economies—all these have contributed to forces that are changing the trajectory of the world economy and its individual constituents. There is still considerable uncertainty about the path the world's economies will ultimately find themselves following. Moreover, there is still substantial room for debate about what can and should be done to enhance the positive impact of economic change and minimize its negative side effects. Nonetheless, it increasingly appears the world's economies are entering a new age characterized by limited inflation and significantly greater reliance on private initiative. To this point, the forces responsible for these traits have also produced low GDP growth, rising income inequality, and relatively high unemployment in the world's leading economies. One would hope that these maladies prove to be no more than passing side effects accompanying the world economy's transition to a new era. If not, chronic persistence of these maladies would almost surely unleash additional forces that could dramatically alter the future course of the global economy.

The most recent forecast for the world economy from the WEFA Group indicates global real GDP growth will conclude 1996 on a positive note.

Although global real GDP growth is now expected to be somewhat slower in 1996 than previously anticipated, growth is still forecast to be above the rate experienced in 1995. WEFA anticipates continued acceleration of world real GDP growth through 1998, at which time growth is expected to level off and remain near 4 percent for several following years. Aggregate real GDP growth for the world's developed economies is likewise expected to conclude 1996 on an upbeat note. Real GDP growth for the world's developed economies as a group in 1996 is anticipated to just better the rate posted in 1995. WEFA now expects real GDP growth in 1997 to match 1996 growth. Beyond 1997, growth is forecast to accelerate somewhat and remain around 2.5 percent through 2001. As for inflation, WEFA continues to expect lower price growth among the world's developed economies in 1996 as compared to 1995 when measured by an aggregate

Consumer Price Index (CPI) for all developed economies. WEFA forecasts a very gradual acceleration of inflation through 1999, at which point CPI growth will run about 2.5 percent. Still, this figure is below the average posted over 1990 to 1995. However, this rosy picture for developed economies as a group masks marked variation in the experiences anticipated for individual developed economies.

SEI\/IM-WW-l\/IT-9603 ©1997 Dataquest 65

66 Semiconductor Equipment, Manufacturing, and Materials Woridwide

The world's developing or emerging economies are expected to continue experiencing real GDP growth rates about two to three times better than their developed counterparts. Emerging Asia/Pacific economies, spearheaded by China, remain forecast to lead the list of fastest-growing economies. Growth rates among Asia/Pacific economies, however, are slowing. And as we shall see later, several of the region's heretofore star performers actually find themselves struggling. Developing Latin American economies are still anticipated to experience real GDP growth rates that lag their Asia/Pacific rivals. Nonetheless, boosted by Mexico's continued recovery and continued economic progress throughout South America, Latin America should experience growth rates about twice those experienced by developed economies. Prospects remain bright for many emerging eastern European economies. Several eastern European economies are already posting rates of real GDP growth that rival those experienced by emerging Asia/Pacific economies. Overall, WEFA forecasts this region to experience about 4.5 annual real GDP growth to the end of the decade. Although the potential for growth remains strong among the economies of the former Soviet Union, WEFA continues to expect a contraction of aggregate real GDP for these economies in 1996. Growth is anticipated to commence in 1997 and eventually move to rates that rival eastern Europe and Latin America by decade's end.

Americas: U.S. Economy... Too Hot? Too Cold? Just Rigiit?

The current circumstances of the U.S. economy remind us somewhat of the children's story about Goldilocks and the Three Bears. In the case of the U.S. economy, economists are Goldilocks and the (dreaded) three bears are falling real GDP, rising unemployment, and (most ferocious of all!) runaway inflation. Month after month throughout 1996, economists have been sampling the porridge of U.S. economic data available to them.

Each time they sample the data, they hope the bears will remain away for yet another month. Sometimes the porridge of statistics they sample has been too hot, meaning runaway inflation could be back anytime. Sometimes the porridge has been too cold, meaning falling real GDP and rising unemployment may be just outside the front door. And sometimes the porridge has been just right, meaning all three bears are off hibernating

(thank you very much!) The upshot of this analogy about the U.S. economy is that economists generally have very mixed views about the current state of U.S. economy activity. They also have varying degrees of fear about just how soon one of those dreaded bears is going to show up and cause trouble. There seems to be general agreement that the economy is very close to being just right. Inflation and unemployment are unbelievably low. And real GDP growth is proceeding, albeit in quarterly fits and starts and at slow average armual rates. Just how long the U.S. economy can maintain the delicate balance between low inflation, low vinemployment, and real GDP growth remains to be seen. Economists know there are bears out in the woods somewhere. The only question is: When will they return and how ravenous will they be?

Elsewhere in the Americas, Canada's near-term prospects remain poor.

Although recent economic data indicates a sustained recovery may be just about to begin, Canada continues to struggle. Canadian fiscal policy remains tight even in the face of falling budget deficits. And recent appreciation of the Canadian dollar, despite accommodative money policy by

SEMM-WW-MT-9603 ©1997 Dataquest March17,1997

Macroeconomic Outlook: Fourth Quarter 1996 67

I

I

I

the Bank of Canada, could delay Canada's recovery. Longer term,

Canada's prospects appear much better. WEFA expects Canada to experience robust real GDP growth accompanied by low consumer price inflation within two years.

Mexico continues to show signs of recovery from the crisis that nearly crushed it two years ago. Real GDP is growing, inflation declining, and the peso now reasonably stable. International investors remain somewhat gun shy. But Mexico's success in reducing inflation together with the steps it has taken toward fighting corruption should help attract vital foreign investment. As for the rest of Latin America, the outlook also appears bright. Brazil and Argentina look to be recovering from recent economic setbacks. There is a growing spirit of cooperation among the region's economies that should encourage economic specialization, foster economic integration, and so buoy economic growth. As in Mexico, the pace of economic development critically depends on the progress of government-initiated economic and social reforms. It also very much depends on prudent management of exchange rates. Latin America's emerging economies would do well to always keep in mind the lessons of Mexico's experience.

WEFA's real GDP growth forecasts for the Americas region are reported in

Table A-1.

Europe, iViiddle East, and Africa: Resurgent Germany Sliepherding EU toward

Common Currency

Recent economic data from western Europe suggests circumstances are improving for many of the region's economies. Prospects for Germany in particular now appear much better. Economic statistics for Germany indicate that real GDP jumped sharply in spring and that industrial production has now sustained increases over several consecutive months. To this point, Germany's recovery appears rooted in the economy's export sector.

Germany's domestic sector remains stagnant and must improve if economic recovery is to continue.

The United Kingdom continues to exhibit robust economic health. Economic data for the United Kingdom indicates that real GDP is continuing to grow and that consumer spending growth, buoyed in part by tax cuts, is especially strong. There is mounting concern that the United Kingdom economy may be on the verge of overheating. This has prompted debate on the advisability of further tax cuts. Britain's Conservative government may find the temptation to cut taxes too hard to resist with general elections scheduled for next spring.

Economic data indicates France's economy remains very unsettled.

Although the economy's export sector has recently shown some improvement, the domestic sector remains stymied. Business confidence is low and households are struggling with the impact of double-digit unemployment. Given France's strong commitment to the Masstricht process, it appears unlikely that government policymakers will intervene on behalf of the economy anytime soon.

Finally, economic statistics for Italy indicate that economy remains mired in the doldrums. Real GDP and private consumption growth remains very

SEI\/IM-WW-MT-9603 ©1997 Dataquest March 17,1997

68 Semiconductor Equipment, Manufacturing, and Materials Worldwide weak. Inflation is sharply down but unemployment is high and looks likely to remain so. The possibility of government policy relief seems very remote at this point. Italy is under intense pressure both at home and abroad to meet Masstridit requirements for entry into the European Monetary Union. Italy's efforts to meet these requirements are likely to hold economic recovery back for sometime.

The long-term outlook for western Europe largely depends on the outcome of the Masstricht Treaty process. Championed by Germany and

France, the treaty calls for the establishment of a European Monetary

Union in 1999 and the institution of a common currency for Union members in 2002. Entry into the proposed Monetary Union will be granted based on economic and fiscal criteria evaluated in 1997. In order to meet these criteria, many treaty signatories are finding it necessary to impose tight fiscal and monetary policies on their economies. While these policies are proving successful in pushing economies toward meeting Masstricht criteria, they are taking their toll on employment and economic growth.

There are strong currency-related incentives for treaty signatories not to be left out of the treaty's proposed Monetary Union. At the same time, countries must balance these concerns against the welfare of their populations. Only time will tell how the Masstricht process plays out. One thing is for sure, however—strong commitment to the process on the part of all treaty parties will hold economic growth back in most of western Europe.

Elsewhere in the region, prospects for eastern Europe remain good. As in

Latin America, economic success for former Warsaw Pact economies has largely been linked to the progress of necessary economic and social reforms. Although carrying out reforms has not been easy, leading economies of the region like the Czech Republic, Slovakia, Poland, and Slovenia are now beginning to reap the rewards of their efforts. These economies are attracting foreign capital and have found markets for their products.

Others in the region have not been so successful and still find themselves struggling. Likewise, many economies of the former Soviet Union continue to struggle. Many thought, or at least hoped, that Yeltsin's victory in

Russia's July presidential elections would herald the dawning of a new era for the Russian economy. Sadly, Yeltsin's failing health combined with infighting among his subordinates seem to have delayed a new economic era for Russia. Still, economic progress is being made. Privatization continues and inflation has been reduced. Nonetheless, basic structural reforms in the Russian economy are badly needed. Foremost among necessary reforms is a tax system that can regularly and reliably collect all the taxes owed the government. Also needed are commercial laws that establish contracts and provide recourse to damaged contract parties.

WEFA's real GDP growth forecasts for the Europe, Middle East, and

Africa region are reported in Table A-1.

Japan and Asia/Pacific: Japan's Recovery Slowing, Four Tigers Struggling

Japan's economy continues to recover, but progress appears to be slowing.

Not surprisingly, real GDP growth decelerated to a more sustainable level in spring following very rapid growth over the winter. To this point,

Japan's recovery appears primarily fueled by robust business investment in plant and equipment as well as by strong housing demand. Contrary to

SEMM-WW-MT-9603 ©1997 Dataquest March 17,1997

Macroeconomic Outlook: Fourth Quarter 1996 69 expectations, public spending remains an important component of real

GDP growth. Unfortunately, consumer spending remains sluggish and exports apparently have yet to respond to the yen's sharp retreat from historically high levels in 1995. There is justifiable concern about the nearterm course of Japan's economy. Funds available for government priming of the economy through public investment are dwindling and likely to be exhausted soon. Moreover, Ja