Manual 21099139

Manual 21099139
1991 SEMMS Newsletters and Bulletins
TAB
Forecasts/Business
Conditions
Technology Trends
Company
Information
SEMMS 1991 Newsletters
TITLE
MONTH
Wafer Fab Equipment Market
Near-Term Forecast
January
Preliminary 1990 Worldwide
Semiconductor Market Share
Estimates: The Microprocessor Begins
January
1991 Silicon Wafer Forecast
January
Regional Wafer Fabrication
Equipment Market Forecast
February
First Quarter 1991 Worldwide
Semiconductor Industry Outlook:
Embattled, But Not Bombed Out
February
The Effects Of The Gulf War
On Semiconductor Capital Spending
March
Will Industry Economics Defy "Moore's Law"
April
Final 1990 Semiconductor Market Shares
May
Equipment And Contamination
In The 1990s
Februaiy
A Glimpse At Future 64Mb DRAM Technologies
March
A Quick Look At Advanced Micro Devices
January
A Quick Look At Texas Instruments
February
A Quick Look At NEC
March
1991 Dataquest Incorporated June
1991 SEMMS Newsletters and Bulletins
TAB
Other Newsletters
Wafer Fab Equipment/
Industry Trends
Lithography
Etch/Clean
Deposition
Diffusion/Implant
TITLE
U.S. Companies Top List For
Strategic Alliances
February
Alliances: Large Company
Rivalries Spawn Start-Ups
February
1990 Wafer Fab Equipment Market: Japan
Accounts For More Than Half Of World's
Supply And Demand
April
The Branson/IPC And Gasonics Merger: A
Growth Option For Small Companies Within
The Wafer Fab Equipment Industry?
April
Equipment Company Ownership-At The
Turning Point
April
Stepper Equipment Market-1990 Market
In Review
April
Wet Processing Appears To Have
Nine lives
February
Dry Etch Equipment: 1990 Market In Review
April
Wet Processing Equipment - 1990 Market
In Review
April
Silicon Epitaxy Equipment Market1990 Market In Review
April
CVD Equipment: 1990 Market In Review
April
PVD: 1990 Market In Review
April
Diffusion Equipment Market: 1990 In Review
April
Process Control
ii
MONTH
1991 Dataquest Incorporated June
SEMMS 1991 Newsletters
1991 SEMMS Newsletters and Bulletins
#
TAB
Semiconductor
Manufacturing/
Industry Trends
MONTH
TITLE
SEMATECH's Congressional Review: Is
The Best It Can Be Good Enough?
January
-
The Role Of Automation In Semiconductor
Manufacturing
March
Merchant DRAM Suppliers: Another Shake-Out
Coming?
March
Wafer Size And Manufacturing Costs:
The Push To Larger Wafers
April
The Fab Database Series: The Shift To
Submicron Geometries
April
Five-Year Capital Spending Forecast:
It's A Good Worid In Spite Of 1991's
Uncertainty
January
U.S. Merchant Semiconductor Capital Spending
Highlights: 1990-1991
April
U.S. Merchant Semiconductor Capital Spending:
Three Long-Term Trends
April
Materials/
Industry Trends
SemiGas: The Balance Between
Trade Policy and Open Markets
January
Silicon
The Decline Of The 100mm
Wafer Market
February
Flat FZ Wafer Market
March
Worldwide Fabs
Capital
Spending
Semiconductor
R&D
Photoresist
Healthy Forecast For U.S.
Positive Resist
January
Gases
Other Materials
SEMMS 1991 Newsletters
California Fabs Will Survive
The Drought
1991 Dataquest Incorporated June
Februaiy
m
Dataqyest
acomranyof
The Dun & BtadstPcct Corporation
Research Newsletter
THE FRAGMENTATION OF KNOWLEDGE AND THE FAILURE OF
ESTABLISHED FIRMS
Dataquest is pleased to publish an invited article by
Dr. Rebecca Henderson, Assistant Professor, Sloan School
of Management, Massachusetts Institute of Technology. In
this article, Dr. Henderson analyzes the failure of established firms to recognize the impact of architectural innovation in new product development, a situation that
Dr. Henderson concludes is a serious contributing factor
to subsequent loss of market leadership. Although
Dr. Henderson's article is based on her research of the
photolithography equipment industry, Dataquest believes
that her conclusions have clear ramifications throughout
all high-technology industry sectors.
Peggy Marie Wood
INTRODUCTION
Ehramatic commercial success is too often a
precursor to equally dramatic failure. In a wide
range of industries, well-established firms with a
track record of financial and technological success
have been unable to respond to subtle shifts in their
environment that have been better e?q)loited by
faster-moving entrants. In the computer industry,
IBM has been a longtime leader in mainframes and
personal computers, not in minicomputers or workstations, and E>igital Equipment Corporation has
yet to duplicate its success in minicon^uters in the
personal computer arena. In the semiconductor
industry, a handful of once-dominant U.S. companies now face significant competition from all
comers of the globe. A three-year study of the
optical photohthography aligner industry suggests
that the reasons behind an erosion in market leadership are often pervasive, critical, and under
managerial control.
ARCHITECTURAL INNOVATION
Research conducted for a doctoral dissertation
(Henderson, 1988) suggests that successful firms
erode their standing in the market through a failure
to respond to "architectural" innovation—
innovation that changes the ways in which the
elements of a product are integrated together while
leaving the core design concepts on which the
product is based untouched. Architectural innovation is particularly difficult to react to effectively
because of the way that a company's technology
base and the customer's requirements are managed
inside the majority of companies. A history of
success with one generation of the technology leads
firms to fragment their technical knowledge and
their understanding of their customer's needs, placing undue reliance on information filters, problem
solving strategies, and commumcation channels
that reflect an increasingly obsolete tinderstanding
of the technology and industry. In the photolithography equipment industry, established firms
attempted to push architectural innovation such as
the scanner and the second generation of steppers
back into the frameworks with which they were
familiar, refusing or falling to understand the
dimensions along which they could offer customers
very real performance improvements.
This concept of innovation can be clarified if
one thinks of a product as consisting of a series of
components integrated together to form the final
product. Innovation can then take the following
four forms:
• "Incremental" innovation improves the performance of individual components but leaves the
relationships between components untouched.
Think, for example, of the steady stream of
improvements in lens size and power that
characterize each generation of stepper.
• "Radical" innovation introduces an entirely new
set of components and, hence, of relationships
between them. The use of direct-write electron
beam machines in mainline production is an
example of this type of radical innovation.
01991 Dataquest Incorporated July-Reproduction Prohibited
SEMMS Newsletters 1991
0011130
The content of this report represents our interpretation and analysis (^inprmation generally available to the public or released by responsible indivitiuals in the subject companies, but
is not guaranteed as to accuracy or completeness. It does not contain nuiterial pntvtded to us in confidence by our clients Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This iriprmalion is notfitmished in connection with a sale or c^fer to sell securities or in contiection with the solicitation of an
c^r to buy securities. This firm and its parent and/or their officers, stockholders, or members of their fivnilies may, fiom tune to time, have a long or short position in the securities
mentioned and may sell or buy stu^ securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE FRAGMENTATION OF KNOWLEDGE AND THE FAILURE OF ESTABLISHED FIRMS
• In "modular" innovation, some of the core concepts of the design are changed while the links
between them remain stable: The replacement of
analog with digital control in some instruments
is an example of this type of innovation.
• "Architectural" innovation is intermediate in
character between incremental and radical innovation: Much of the knowledge that a firm has
accumulated in its experience with incremental
innovation remains relevant, but its architectural
knowledge—^its knowledge about the relationship between components—becomes obsolete.
Architectural innovation in photohthography—^the introduction first of the proximity printer
and then of the scanner, the stepper, and the second
generation of stepper—created enormous problems
for established firms because they had allowed
their architectural knowledge to become embedded
in the tacit knowledge of the organization—^in their
communication channels, information filters, and
problem-solving strategies—^where it became difficult to observe and almost impossible to change.
The Case of Kasper Instruments
The case of Kasper Instruments and its
response to Canon's introduction of the proximity
printer illustrates some of these problems. Kasper
Instruments was founded in 1968 and by 1973 was
a small but profitable firm supplying approximately
half of the market for contact aligners. In 1973,
Kasper introduced the first contact aligner to be
equipped with proximity capability. Although
nearly half of all the aligners that the firm sold
fi^om 1974 onward had this capability, Kasper
aligners were only rarely used in proximity mode,
and sales declined steadily until the con^any left
the industry in 1981. The widespread use of proximity aligners oiHy occurred with the introduction
and general adoption of Canon's proximity aligner
in the late 1970s.
Canon's aligner was superficially very similar
to Kasper's. It incorporated the same components
and performed the same functions, but it performed
them much more effectively because it incorporated a much more sophisticated understanding of
the technical interrelationships that are fundamental
to successful proximity alignment. Kasper failed to
develop the particular component knowledge that
would have enabled it to match Canon's design.
More importantly, the architectural knowledge that
0011130
Kasper had developed through its experience with
the contact aligner had the effect of focusing its
attention away from the new problems whose solution was critical to the design of a successful
proximity aUgner.
Kasper conceived of the proximity aligner as
a modified contact aligner. Like the incremental
improvements to the contact aligner before it,
design of the proximity aligner was managed as a
routine extension to the p'oduct line. In particular,
the g^-setting mechanism that was used in the
contact aligner to align the mask and wafer with
each other was sUghtly modified, and the new
aUgner was offered on the market. As a result,
Kasper's proximity aligner did not perform well.
The gap-setting mechanism was not sufficiently
accurate or stable to ensure adequate {>erformance,
and the ahgner was rarely used in its proximity
mode. Kasper's failure to understand the obsolescence of its architectural knowledge is demonstrated graphically by two incidents.
The first incident was the firm's interpretation
of early complaints about the accuracy of its gapsetting mechanism. In proximity alignment, misalignment of the mask and the wafer can be caused
both by inaccuracies or instability in the gap-setting
mechanism and by distortions introduced during
processing. Kasper attributed many of the problems
that users of its proximity equipment were
experiencing to processing error, because it
beUeved that processing error had been the primary
source of problems with its contact aligner. The
firm "knew" that its gap-setting mechanism was
entirely adequate and, as a result, devoted very
little time to improving its performance. In retrospect, this may seem like a wanton misuse of
information, but it represented no more than a
continued reliance on an information filter that had
served the firm well historically.
The second illustration is provided by
Kasper's response to Canon's initial introduction of
a proximity aligner. The Canon aligner was evaluated by a team at Kasper and pronounced to be a
copy of a Kasper machine. Kasper evaluated it
against the criteria that it used for evaluating its
own aligners—criteria that had been developed
during its experience with contact aligners. The
technical features that made Canon's aligner a significant advance, particularly the redesigned gap
mechanism, were not observed because they were
not considered important. The Canon aUgner was
pronounced to be "merely a copy" of the Kasper
aligner.
®1991 Dataquest Incoiporated July-Reproduction Prohibited
SEMMS Newsletters 1991
THE FRAGMENTATION OF KNOWLEDGE AND THE FAILURE OF ESTABLISHED RRMS
Further Examples in the
Photolithography Industry
Similar problems show up in other episodes
of architectural innovation in the industry's history.
In one company, the engineers evaluating a
new technology—an architectural innovation—
accurately forecast the progress of iudividual components in the new system but failed to see how
new interactions in component development—
including better resist systems and improvements in
lens design—^would give the new technology a
decisive advantage.
Similarly, in another con:q)any, engineers were
organized by component, and cross-department
communication channels were all structured around
the architecture of the first-generation system.
Although the engineers were able to push the limits
of the component technology, they had great difficulty understanding die roots of the superior performance achieved by the next generation of equipment. A successful entrant in the market changed
aspects of the design—particularly the ways ia
which the optical system was integrated widi the
rest of the aligner—of which the established firm's
engineers had only limited understanding. Moreover, because these changes dealt with component
interactions, there were few engineers responsible
for developing this understanding. As a result, the
older firm's second-generation machines did not
deliver the kind of performance that the market
demanded. In both of these cases, other factors also
played a role in the subsequent loss of market share
for the older companies, but a failure to respond
effectively to architectural innovation was of critical importance.
01991 Dataquest Incorporated July-Repioductioii Prohibited
SEMMS Newsletters 1991
CONCLUSIONS
Is the concept of architectural innovation useful in thinking about the issues facing hightechnology industries today? The answer is a
resounding yes. Continuing research to support this
premise in the aerospace, pharmaceutical, and electronic instrument industries is currently under way
at MTT. The key to managing architectural innovation successfiilly ia aU of these areas seems to be
the e^Ucit recognition that a firm cannot afford to
let the knowledge of its design team become fragmented and bounded by information filters or communication channels diat reflect only the factors
that have made the firm successful historically.
Design teams that manage to survive architectural
innovation actively seek to reintegrate the knowledge of their designers through strong, integrative
team management; extensive cross training; and a
focus on the goals of the group as a whole rather
than the goals of any particular discipline. As semiconductor technology becomes increasingly complex and architectural innovation becomes more
pervasive, the firms that survive wiU be those that
have learned this lesson.
Rebecca Henderson, PhD.
Assistant Professor
Sloan School of Management
Massachusetts Institute of Technology
0011130
•TO
DataQuest
if'^^i.
acontpanycf
The Dun& Bradstrccl coq»nUDn
WM^
Xfr
^i v'.ii -i'c
Research Newsletter
FINAL 1990 SEMICONDUCTOR MARKET SHARES
INTRODUCTION
The number one rule of the semiconductor
game asserted itself forcefully in 1990: What goes
up must come down. Companies that derive large
portions of their revenue from highly volatile commodity products will eventually see their business
decline and their market share decrease as rigidly
as they had previously grown. The case in point is
MOS memory, which was responsible for much of
the semiconductor industry's growth in 1988 and
1989. The market share gains made by memory
suppliers in those years fell by the wayside (in
most cases) in 1990, allowing other companies to
move up in the ranking Hst.
MiCROCOMPONENTS: THE 1990 MARKET
DRIVER
Although the memory market did not disappear in 1990, steep price declines made it an
unpleasant market to be in. The top three semiconductor vendors in 1990—^NEC, Toshiba, and
Hitachi—each derived 35 percent or more of their
1990 semiconductor revenue from MOS memory
products. In 1990, this ratio slipped by 3 to
6 percentage points. Aldiough the same effect can
be seen for the fourth and fifth ranked
conqjanies—Motorola and Intel—these companies
received only 12 and 18 percent, respectively, of
their 1989 revenue from MOS memory. Figure 1
compares 1989 and 1990 reliance on MOS memory
by these companies.
In 1990, these top five companies increased
the percentage of total semiconductor revenue from
MOS microcomponents. In the case of Intel, which
went iq) in the rankings from number eight to
number five, the portion of total semiconductor
revenue that came from microcomponents grew
from 79 to 86 percent lliese conqjarisons can be
seen in Figure 2.
The effect of MOS microcomponents on the
industry as a whole can be seen in Figure 3, which
shows that were it not for the dramatic growth of
microcomponents, the senuconductor industry
would have declined by 2 percent in 1990, rather
than growing 2 percent.
DIFFERENCES IN REGIONAL COMPANIES
It is very clear that vastly different product
strategies are being followed by each regional
grouping of companies. The largest portion of
North American companies' 1990 revenue—
27 percent—came from microcomponents, while
the largest portions of Japanese companies' revenue were from MOS memory (28 percent) and the
combined grouping of bipolar digital, discrete, and
optoelectronics (29 percent). European companies,
on the other hand, are heavily dependent on bipolar
digital/discrete/opto (36 percent) and analog
(30 percent); their revenue from the fastest-growing
segments of the semiconductor industry (in the
long term)—MOS memory and microcomponents—totaled only 21 percent of their semiconductor revenue. Asia/Pacific companies' revenue was very heavily skewed in favor of memory,
with 63 percent of their revenue coming from that
product category.
Figure 4 illustrates the product portfolios by
regional company base. From this analysis, it
appears that die most evenly balanced portfolios
belong to the North American and J^anese companies, with percentage point spreads of only 13 and
16 percentage points, respectively, between the
largest and smallest categories.
Although it might initially ^pear that European companies' portfolios are less well balanced,
their targeted implication markets differ from North
American and J^anese companies in that a higher
percentage of their ou^ut is aimed at the consumer
0 1 9 9 1 Ditaqueit Incoiponted May-Kepioduction Proliibiled
00104S6
Semicraiductor Oroup Newiletten 1991-06
The content of this report represents our interpretation and analysis qfir^rmation 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^idence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This ir^rmation is not furnished in connection with a sale or offsr to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their cheers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
FINAL 1990 SEMICONDUCTOR MARKET SHARES
FIGURE 1
Top Five 1990 Semiconductor Suppliers' Reliance on MOS Memory
(Percentage of Total Semiconductor)
Percentage
50-
E 3 1989
1390
40
30-1
20
M
10-1
NEC
Toshiba
HitacN
Motorola
c\\
Intel
Source: Dataquest (May 1991)
FiGUBE 2
Top Five Semiconductor Suppliers' Reliance on MOS Microcomponents
(Percentage of Total Semiconductor)
Percentage
100
90
80
ES3 1989
1990
70 ^
60
50-1
40
30 ^
20
10
0
NEC
Toshiba
Hitachi
Motorola
Intel
Source: Dataquest (May 1991)
001O4S6
01991 DaUqiieit Incoiponted May-Repraductian Prohibited
Sendconductor Group Newiletten 1991-06
FINAL 1990 SEMICONDUCTOR MARKET SHARES
FIGURE 3
MOS Microcomponents: The Industry Driver
(Percent Growth in 1990)
Percentage
252015
10
5-
L:VVWW\XN
Total Semiconductor
MIcrocomponent
kWWWNXT
Semiconductor
(Excluding Microcomponents)
Source: Dataquest (May 1991)
Sooice: Dataqueit (May 1991)
FIGURE 4
1990 Product Portfolios by Company Base
(Percentage of Dollar Revenue by Ftoduct Category)
Percentage
MOS Memory
MOS Logic
MOS MIcrocomponent
Analog
I
I Bipolar DIgltal/Dlscrete/OptoelectronIc
100-
80-
60-
40-
20
North American
Companies
Japanese
Companies
European
Companies
Asla/Paclfic
Companies
Source: Dataquest (May 1991)
ei991 Dauqoeit Incaiponted liby-Reiaoduetion Prohibited
Semicaiductot Otaap Newdettan 1991-06
00104S6
4
FINAL 1990 SEMICONDUCTOR MARKET SHARES
and telecommunications industries, which use large
quantities of discrete and analog chips.
The Asia/Pacific company statistics are
skewed by Samsung, whose sdes account for
62 percent of this group. Because Samsung
has become one of the world's largest DRAM
suppliers, it is not surprising that MOS memoiy
accounts for 63 percent of Asia/Pacific company
sraniconductor revenue.
1990 RANKINGS AND MARKET SHARE
Table 1 is an analysis of the worldwide semiconductor market by regional supplier base and
regional consumption market. This table shows that
in 1990 North American companies held 37 percent
of the worldwide semiconductor market, Japanese
companies held 49 percent, European companies
held 11 percrait, and Asia/Pacific conq)anies held
4 percent.
Table 2 is a breakdown of the semiconductor
market by product category for 1989 and 1990. The
market grew a total of only 1.8 percent, but as
previously alluded to, MOS microcomponents grew
by 22.8 percent. Analog ICs, the second fastestgrowing market, grew 12.6 percent. (We have
included nuxed-signal analog/digital ICs in the
analog category.)
TABLE 1
Estimated Final 1990 Semiconductor Market Share Analysis
(Factory Revenue in Millions of U.S. Dollars)
Company:
Each Regional Base
Product:
Total Semiconductor
Region of Consumption:
Each
Distribution Channel:
NM
Application:
All
Specification:
All
Company Base
North America ($M)
Percent of Regional Maricet
Percent of Company Sales
Japan ($M)
Percent of Regional Market
Percent of Company Sales
Europe ($M)
Percent of Regional Maiket
Percent of Company Sales
Asia/Pacific ($M)
Percent of Regicmal Maiket
Percent of Company Sales
World ($M)
Percent of Regicxial Maiket
Percent of Company Sales
North
America
Regional Market
Asia/
PacificROW
World
Japan
Europe
11,942
2,402
4,492
2,701
21,537
69
55
11
11
42
21
35
13
37
100
3,777
19,825
1,814
2,961
28,377
22
13
88
70
17
6
39
10
49
100
1,074
164
4,117
851
6,206
6
17
1
3
39
66
11
14
11
100
593
117
238
1,157
2,105
3
28
1
6
2
11
15
55
4
100
17386
22,508
10,661
7,670
58,255
100
30
100
39
100
18
100
13
100
100
N M - Not ]
Souice: Dataqucft (May 1991)
0010456
01991 Dataqueit liocoipotated May-Repiodiictioa PtofaiUted
SemicaoduGtof Oroap Newdetten 1991-06
FINAL 1990 SEMICONDUCTOR MARKET SHARES
TABLE 2
Estimated Semiconductor Consumption
(Factory Revenue in Millions of U.S. Dollars)
Company:
Product:
Region of Consumption:
Distribution Channel:
Application:
Specification:
Total Semiconductor
Total Integrated Circuit
Bipolar Digital
Bipolar Memory
Bipolar Logic
MOS Digital
MOS Memory
MOS Microcomponent
MOS Logic
Analog
Discrete
Optoelectronic
AU
Each
Worldwide
NM
All
All
Percent
Change
1989
1990
57,213
58,225
46,924
4,510
540
3,970
33,024
16361
8,202
8,461
9,390
47,303
4,440
459
3,981
32,292
13,091
10,068
9,133
10,571
1.8
0.8
-1.6
-15.0
0.3
-2.2
-20.0
22.8
7.9
12.6
7,662
8,235
7.5
2,627
2,687
2.3
NM = Not meamiigfiil
Soince: Dataqoeit (Mqr 1991)
The top 40 semiconductor coii^)ames' worldwide rankings and revenue are shown in Table 3.
can pay off handsomely. We continue to believe
that conq>anies widi balanced product portfolios in
conjunction with volatile commodity e^qmsure will
gain market share over the long term.
DATAQUEST CONCLUSIONS
The memory market will recover, and companies with major commitments in this market will
have a chance to regain semiconductor market
share. However, in 1990 Dataquest saw that a
strong marketing strategy in other product areas
e i 9 9 1 Dataqueit Inccnpanted Mmy^^Repioduction Prohibited
Semieanductor Gitnip Newdetten 1991-06
Patricia S. Cox
Note: Detailed market share data books have
he&a. conpleted and mailed to binderholders of the
SIS, JSIS, ESIS, ASETS, and NASM services..
O0IO4S6
FINAL 1990 SEMICONDUCTOR MARKET SHARES
TABLE 3
Estimated Market Share Ranking
(Factory Revenue in Millions of U.S. Dollars)
Company:
Product:
Region of Consiunpti<on:
Distribution Channel:
Applicationi:
Specification:
Top 40
Total Semiccmductor
Worldwide
NM
All
All
Percent
Change
1990
Market
Share (%)
1990
Rank
1989
Rank
1
1
NEC
5,015
4,898
-2
8.4
2
2
Toshiba
4,930
4,843
-2
8.3
3
3
Hitaclii
3,974
3,893
-2
6.7
4
4
Motorola
3,319
3,694
11
6.3
5
8
Intel
2,430
3,171
30
5.4
6
5
Fujitsu
2,963
2,880
-3
4.9
7
6
Texas Instrumoits
2,787
2,574
-8
4.4
8
7
Mitsubishi
2,579
2,319
-10
4.0
9
10
Philips
1,716
2,011
17
3.5
10
9
Matsushita
1,882
1,942
3
3.3
11
11
National Semiconductor
1,618
1,719
6
3.0
12
13
SGS-Thomson
1,301
1,463
12
2.5
13
12
Sanyo
1,365
1,381
1
2.4
14
15
Sharp
1,230
1325
8
2.3
15
14
Samsung
1,260
1,315
4
2.3
16
16
Siemens
1,194
1,224
3
2.1
17
19
Soay
1,077
1,146
6
2.0
18
17
Oki
1,154
1,074
-7
1.8
19
18
Advanced Micro Devices
1,100
1,053
-4
1.8
20
20
AT&T
873
861
-1
1.5
21
21
Harris
830
800
-4
1.4
22
22
Rohm
740
774
5
1.3
23
23
LSI Logic
512
598
17
1.0
24
26
Saaaikea
387
407
5
0.7
25
NM
GEC Plessey
0
390
NM
0.7
26
28
Riji Electric
362
385
6
0.7
00104S6
1989
Revenue
1990
Revenue
01991 Dataqueit IncorpcmCed May.J(eprodiictiai Prohibited
Semicoaductor Group Newiletten 1991-06
FINAL 1990 SEMICONDUCTOR MARKET SHARES
TABLE 3 (Continued)
•
Estimated Market Share Ranking
(Factory Revenue in Millions of U.S. Dollars)
Company:
Product:
RegicHi of Consumpti<3n:
Distribution Channel:
Application:
Specification:
1990
Rank
1989
Rank
27
29
28
Tbp 40
Total Semiconductor
Worldwide
NM
AU
All
Percent
Change
1990
Market
Share (%)
1989
Revenue
1990
Revenue
Analog Devices
357
381
7
0.7
25
m
390
371
-5
0.6
29
31
VLSI Technology
286
324
13
0.6
30
30
Telefunken Electronic
299
295
-1
0.5
31
24
Micron Technology
395
286
-28
0.5
32
32
Hewlett Packard
269
279
4
0.5
33
35
Chips & Technologies
240
265
10
0.5
34
40
Intemational Rectifier
190
225
18
0.4
35
39
Cypress Semiconductor
196
223
14
0.4
36
42
Goieral Instrument
170
214
26
0.4
37
27
Seiko Epson
368
213
•Al
0.4
38
NA
Shindengoi Electric
NA
209
NA
0.4
39
33
NMB Semiconductor
247
201
-19
0.3
40
43
Rockwell
165
200
21
0.3
7,043
6,399
-9
11.0
North American Companies
19,978
21,537
8
37.0
Japanese Coiiq>anies
29,809
28,377
-5
48.7
European Companies
5,443
6,206
14
10.7
Asia/Pacific Companies
1.983
2,105
6
3.6
Total Market
57,213
58,225
All Odiers
100.0
NA - Not avaiUble
NM - Not meaiiiigbl
Somce: DabupiMt (M^ 1991)
01991 DaUipieft Incoiponled May-Reproduction Prohibited
Semiccoductor Oroi^ Newiletten 19914)6
00104S6
.^^g-
Dataouest
^ •, "^ •• r
^] ^_^' /-[:-
i m acompanyof
U 0 TheDun&BradsticetCorporatK
:,^=
f^W*.-"
<.^-L
Research Newsletter
WILL INDUSTRY ECONOMICS DEFY "MOORE'S LAW?
"/« r/ie /o«g run, financial pressures may
force semiconductor companies to slow down the
pace of technological development."
—^Dr. Gordon E. Moore
Dataquest's 1987 Semiconduaor Industry Conference
SUMMARY
Dataquest believes that the semiconductor
industry is facing a slowdown in the historical pace
of integrated circuit (IC) price/performance
improvements. The factors affecting this slowdown
have less to do with technical limitations than with
the costs of overcoming them. As the semiconductor industry enters the final decade of the
20th century, it faces the critical question of
whether the cost of contuiiiing its improvements in
device complexity will set the deciding limits to
future declines in the cost per function of semiconductor ICs.
A slowdown in IC price/performance
improvements has enormous implications: Such a
slowdown coiild ultimately slow the rate of semiconductor market growth, lower the industry's
return on investment, and reduce technical
improvements because of reduced R&D spending.
The net result of diminished price/performance
improvements could well be the lengthening
of product life cycles in leading-edge devices—a
phenomenon that wotild in turn have a profound
intact on the electronic systems industry! The best
analogy of this situation is that of a vicious circle
taking the form of a tightening noose.
This newsletter is the first in a series that
explores global industry issues. This iSrst installment reviews the basic tenets underlying "Moore's
Law" and examines cost issues that Dataquest
believes are fundamentally altoing the price/performance dynamics of the semiconductor industry.
MOORE'S LAW REVISITED
The electronics industry's notions of price/
performance improvements were best defined
almost three decades ago, when Intel Corporation's
cofounder and chairman of the board. Dr. Gordon
E. Moore, observed that the "complexity of
integrated circuits has approximately doubled every
year since their introduction, (and) cost per function has decreased several thousandfold" (author's
italics). Dr. Moore's observation has become a
generally accepted semiconductor industry canon
known as "Moore's Law." Over a nearly 20-year
period from the introduction of the first planar
transistor to the mid-1970s, the semiconductor
industry achieved a nearly 800-fold increase in
device complexity, which Dr. Moore attributed to
exponential progress in the following areas:
• Die size—^In a 20-year period, IC die size for the
most complex integrated circuits increased by a
factor of 20.
• Linewidth—^During the same period, reductions
in linewidth and space improved by a factor of
approximately 32.
In combining these factors. Dr. Moore was
able to account for an overall 640-fold increase in
device complexity, leaving a factor of about 100 to
account for. This factor, according to Dr. Moore,
was "clevemess"—^the contribution of circuit and
device advances to the achievement of higher
device density. "It is noteworthy," Dr. Moore
observed, "that this contribution to complexity has
been more important than either increased chip area
or finer lines."
OF TECHNOLOGY AND CEREAL PRIZES
Moore's law, it may be argued, is the most
fundamental observation of how the semiconductor
<D1991 Dataquest Incorporated April-Reproduction Prohibited
SEMMS Forecast/Business Conditions
0010052
The content of this report represents our irjerpretation and analysis ofinprmation 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 confidenee by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This information is not furnished in connection with a sale or offer to sell securities or in connection 'Adth the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of theirfiimiliesmay, from time to time, have a long or short position in the securities
mentioned and may sell or buy such .securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
WILL INDUSTRY ECONOMICS DEFY "MOORE'S LAW"?
industry has added value during the more than
40 years since the invention of the transistor. It is
the basis for bit-density improvements in semiconductor memories, gate-count improvements in logic
devices, and the basic explanation for how an
industry can deUver miracles on a timetable basis.
The semiconductor industry has made possible the
fact that consumers worldwide have an unshakable
expectation that electronics products will continuously deliver greater functionality and performance
at lower cost. Equally important, Moore's Law has
been the basis of profitabUity (however marginal at
times) for an industry that appears perpetually
poised to cannibalize itself. The fact that a leadingedge device such as the 4Mb DRAM can drop
from a price of $460 to less than $10 in a six-year
period ftiels, through increased application, a level
of demand that to some degree offsets such drastic
erosions in price.
A powerful example of price/performance
improvements in the semiconductor industry—one
that moves us from the realm of abstraction to
consumer reality—is the liquid crystal display
(LCD) electi-onic watch. In 1975, the LCD watch
represented electronics industry state of the art in a
number of areas—the electronic timing semiconductor chip, the CMOS processing that made
it possible, LCD technology, the small lithium
battery, and the quartz crystal oscillator. Less than
20 years later, these very technologies are being
delivered as a plastic giveaway in a box of breakfast cereal. Not only that, all of these impressive
technologies of 1975 are now routinely manufactured at low cost in Asia and assembled in mainland China at a likely cost of less than 60 cents.
The plastic watch case may be as much of a cost
factor as the semiconductors it contains!
WHERE ARE THE LIMITS?
Although the semiconductor industry has
granted the same veracity to Moore's Law that one
might accord to Newton's Third Law of Motion, it
must be remembered that Moore's Law is essentially an observation of the dynamics of an industry
based on past performance. Numerous industry
leaders, including Dr. Moore himself, have
cautioned against an extrapolation of device complexity trends through a strict linear progression
based on Dr. Moore's insight of three decades ago.
In short, there are limits beyond which Moore's
Law cannot logically apply. At Dataquest's 1987
Semiconductor Industry Conference, the coinventor
0010052
of the integrated circuit. Jack Kilby, commented
that such a straight-line extrapolation of device
complexity to the year 2027 would suggest a range
of 100 billion to 10 triUion transistors on a 12-inch
square chip—at a price of $3 per chip!
The mind-boggling impossibility of Jack
Kilby's extrapolation appears obvious even to an
industry accustomed to enormous technological
strides. During the 1987 conference. Dr. Moore
identified a more fundamental limit to silicon-based
semiconductor technology when he observed that
progress in dimension reduction, the most important aspect of the industry's technological progress,
"doesn't show any signs of abating until we reach
a physical limit probably near 0.1-micron"—a
limit that the industiy will most likefy encounter
20 years from now.
In an update of his original analysis 30 years
ago, Dr. Moore observed that packing efficiency in
ICs had progressed by a factor of four between
1959 and the mid-1970s. At that time, however, he
observed, "I am inclined to suggest a limit to the
contribution of circuit and device cleverness of
another factor of four (improvement) in component
density." With cleverness diminishing as a chief
contributor to device complexity. Dr. Moore speculated that the slope of the Moore's Law complexity
curve could slow to a doubling of complexity every
two years rather than every year. Although this has
not yet proven to be the case, there are clearly
limits to how far the industry can go in improving
such factors as defect density and jdeld.
THE PRICE OF THE FUTURE
While acknowledging the technical challenges
of following Moore's Law into the submicron era,
one must also acknowledge that the history of the
semiconductor industry is replete with examples of
overcoming technical limitations. However, there
are laws other than those of physics that affect the
future pace of the semiconductor industry's price/
performance improvements. A greater obstacle to
the continuation of Moore's Law may be the law of
economics—a law more harsh in its effects on
companies than any they may face in the R&D lab.
Put another way, a question facing the industry is:
"Which do we run out of first, cash or creativity?"
The economic forces arrayed against Moore's Law
have to do with the price of industry progress: the
costs of designing, marketing, and manufacturing
leading-edge ICs. Some of the most fundamental
cost issues currentiy facing the semiconductor
industry are reviewed in the following paragraphs.
01991 Dataquest Incoiporated April—Reproduction Prohibited
SEMMS Forecast/Business Conditions
WILL INDUSTRY ECONOMICS DEFY "MOORE'S LAW"?
Lithography Discontinuity
In a recent meeting with Dataquest semiconductor analysts. Eh-. Moore expressed his current
doubts that industry progress toward the limits of
device physics wovdd proceed unabated. His earlier
conviction that dimension reduction could follow
its current exponential slope until it reaches the
0.1-micron level has been tempered by a concern
that the industry will encounter serious obstacles in
getting past the 0.35- to 0.25-micron range,
the feature size requirement for production of
64Mb and 1-gigabit DRAMs! Production manufacturing of circuits with 0.25-micron geometries will
be feasible, from I>r. Moore's current perspective,
only if it can be done with optical lithography
technology. The industry faces an enormous discontinuity at the point that optical lithography is
abandoned. The issue, however, is not so much the
technical achievement of manufacturing devices
with feature sizes below 0.25 micron; it is the cost
of doing so given the investments that wiU have to
be made in the technological alternatives to optical
lithography.
per-wafer cost of semiconductor manufacturing.
The most critical of these are as follows:
• An increased nimiber of mask steps—from 5 for
a 16K device to between 25 and 30 for a
64Mb DRAM.
• An increase in interconnect levels—1Mb
SRAMs, 4Mb SRAMs, and 16Mb DRAMs will
use two levels of metal. ASICs, which lead
interconnect technology, will have four or
five levels of metal.
• An increase in the number of process steps—
each interconnect level involves many deposition and etch steps and drives up process complexity as well as requiring expensive equipment. On average, the total number of process
steps per wafer wOl increase from 200 for
the 1Mb DRAM to ^jproximately 800 for the
64Mb DRAM.
• The increase in process complexity wiU cause a
decrease in wafer throughput, resulting in lower
factory productivity. In addition, increases in
process complexity wiU mean higher workin-process inventory costs.
Design Costs
Dataquest beheves that design cost, whether
measured by per bit, transistor, or gate, now costs
about one-fortieth of what it cost in the early
1970s. Numerous factors lie behind this tremendous progress, most of these having to do with the
switch from physical layout and "hand analysis" to
computer-aided design (CAD) advances in
schonatic capture, auto-routing, and simulation. On
the other hand, chip density has increased
2,000-fold in the last 20 years, making the cost of
design higher on a per-device basis. As a rule of
thumb, design and/or development costs have gone
up with the square root of density despite the
offsetting benefits of CAD technology. Look at the
history of microprocessor development: In the past
decade, accordiag to Intel, development costs rose
from $25 million for the 8086 to $250 million for
the 80486. As a whole, Dataquest estimates that
design costs have risen 45 times during the last
two decades.
Manufacturing Costs
M die near future, Dataquest sees a number
of process trends that threaten to drive up the
01991 Dataqtiest Incoiporated April-Reproductiaii Prohibited
SEMMS Foiecast/Business Conditions
Marketing Costs
The costs associated with marketing ICs have
increased as a function of the increase in market
size and the industry's movement to worldwide
markets. Dataquest estimates that the former
accounts for about an 8 times increase in marketing
costs and the latter about a 3 times increase—or
about 25 times altogether. Increasing competition
among suppliers at the applications level wiU tend
to push these costs up still further.
Wafer Fab Costs
According to Dataquest's Semiconductor
Equipment, Manufacturing, and Materials Service
(SEMMS), the cost of a state-of-the-art fab has
risen from $30 nulMon in 1970 to $300 milUon
today, as illustrated in Figure 1. Equipment costs
for a single station routinely exceed $1 milUon and
are increasing rapidly. As a result, process equipment costs have risen from 40 percent of the total
fab cost to ^proximately 70 percent.
Based on discussions that SEMMS analysts
have had with several semiconductor manufacturers
in Japan and the United States, the cost estimates
0010032
WILL INDUSTRY ECONOMICS DEFY "MOORE'S LAW"?
FlGUKE 1
Building plus Equipment Costs for a High-Volume Fab Line
Source: Dataquest (April 1991)
for a 64Mb DRAM fab range from $500 million to
as high as $1 billion. Assuming the more conservative end of this range, we would still be looking at
fab costs rising at a compound annual growth rate
(CAGR) of 15 percent from 1990 to 1996 compared with a CAGR of 12 percent from 1970 to
1990. With both fabfinancingand fab productivity
becoming equally critical, a slow ramp in production would be disastrous to suppliers both in terms
of carrying cost and market prices. If this was true
in the past, it wiU be more so in the future.
THE LIMITS OF PRODUCTIVITY
Up until now, the semiconductor industry has
been doing a remarkable job of meeting the
manufacturing challenge of increasingly complex
devices with increased productivity. Clearly, the
industry has responded to manufacturing complexities and costs with increases in die and wafer size
and improvements in device jaeld. At Dataquest's
1987 Semiconductor Industry Conference, Dr.
Moore happily remarked that the semiconductor
industry had exhibited "the ability to improve
products in all dimensions simultaneously by making them smaller with essentially no trade-offs."
A nvimber of factors have contributed to the
industry's success in continuing its dramatic price/
performance improvements. CAD, for one thing,
has not only enabled the industry to design significantly more complex products widi greater efficiency but also to model "real-world" conditions
0010052
of advanced wafer manufacture without incurring
all of the attendant learning-curve costs. The industry has also witnessed a qualitative shift in technology development toward zero defects in the
sense that defect elimination is occurring more
rapidly than new generations of technology are
turning over. As a result of improvements in defect
density, there have been corresponding improvements in 5deld and therefore in process complexity
(in terms of the number of mask levels) and
wafer size.
Nevertheless, th^e are limitations to this progress as well. As Dr. Moore observed four years
ago, "as the defect density is reduced to a tenth of
a defect per square centimeter or lower, the yield
approaches 100 percent." Although the industry
has made great gains in improving average yields
from 20 percent to better than 80 percent, there
simply is not another factor of four left in jdeld
improvement Thirty years ago. Dr. Moore noted a
limitation to increases in die size: "Extension to
larger die size depends principally upon the continued reduction in the density of defects...(and)
their density can be reduced as long as such reduction has sufficient economic merit to justify the
effort." The operative words in this statement are
economic merit.
PAYING THE PIPER
With the industry reaching some serious limits
in the ability of productivity gains to offset rising
wafer capitzd and processing costs, it is clear that
01991 Dataquett Incoiporated April-Reproduction Prohibited
SEMMS Forecast/Business ConditioDs
WILL INDUSTRY ECONOMICS DEFY "MOORE'S LAW"?
chip costs wOl rise. The consequence of this
increase will be a marked slowdown in the rate of
price/performance improvement. Historically, data
show that when costs (and prices) stop falling, they
do so rather abruptly. This is true because of the
compounded effects of slower market growth,
lower return on investment, and reduced technical
improvements resulting from reduced R&D
spending—less market, less investment, less
opportunity.
The semiconductor memory market is already showing signs of slower growth, at least as
measured in compounded annual bit growth. From
1980 to 1985, compounded bit growth in semiconductor memories proceeded at a 114 percent annual
pace. Between 1985 and 1989, Dataquest observed
a slowdown in the rate of bit growth to 62 percent.
For die time being, the mad&et is continuing its
60 to 70 percent growth rate, but we believe that
growth win be far slower in the future.
This forecast does not imply that the market
wiU not continue to grow in dollar terms; the
01991 DaUquest Incoiponted Apiil-Reproductian Pndiibited
SEMMS Foiecasl/Busmess Conditioiu
semiconductor market will see healthy growth in
the foreseeable future. Dataquest does expect, however, that product lifetimes will lengthen and new
product introductions—^generation turnover—^wiU
come slower. As a speaker pointed out during a
speech at the 1990 Dataquest Semiconductor Industry Conference, "For 20 years I have been a proponent of the industry's experience curve. No longer.
Moore's Law is dead or dying. This wiU be
plainly evident in two to three years." The critical
questions now before the industry concem the consequences of a slowdown in the rate of price/
performance improvement and how the industry
might adjust to these consequences.
George Burns
Michael J. Boss
(This document is reprinted with the permission of Dataquest's Semiconductor Industry
Service.)
0010032
Dataoyest
acompanyof
The Dunk Bradsbeet Corporation
Research Bulletin
THE EFFECTS OF THE GULF WAR ON SEMICONDUCTOR
CAPITAL SPENDING
The eruption of the crisis in the Persian Gulf
in August sent a shock through the world's economic system. Consumer and business confidence
plummeted. The U.S. economy, already sinking
under the weight of a budget deficit, a trade deficit,
and the S&L crisis, finally submerged beneath the
waves of a recession.
EFFECTS OF THE GULF WAR
Dataquest has discussed the war's end with
several major U.S. semiconductor manufacturers.
We believe that, with flie war's end, the negative
consequences of the Gulf Crisis wUl now be
reversed and there wUl be positive short- and longterm consequences for the semiconductor industry.
Short-Term Consequences
From our discussions with industry representatives, we believe that the following short-term
consequences of the war's end are likely:
• The war's end will eliminate a major source of
instability and uncertainty.
• The war's end will increase consumer and business confidence.
Long-Term Consequences
Important as the short-term consequences are,
many in the industry believe that, because of the
spectacular success of high technology in the war,
the long-term consequences of the war's successful
conclusion will be even more substantial. The following long-term consequences are expected:
• Increased awareness on the part of the U.S.
government, including the executive branch, of
the strategic importance of high technology in
general and semiconductors in particular
• Stable level of military IC procurement, or at
least less of a decrease than had been originally
projected with the thawing of the cold war
• Increased support from the U.S. government for
high-technology and semiconductor R&D,
including direct funding, tax credits, and support
for consortia
• Increased likelihood of support for an industrial/
trade policy supportive of high technology
DATAQUEST CONCLUSIONS
We believe that increased confidence on the
part of both business and consumers could signal
an early end to die ciirrent recession. An end to the
recession would buoy up semiconductor capital
spending in 1991. Increased confidence and the
return of general economic growth would also lead
to continued growth in 1992 and 1993. At the end
of the last general economic recession (in November 1982) semiconductor ci^ital spending enjoyed
(in 1983 and 1984) two record years of growth,
pattiy due to the PC boom but also due to a
resurgence of business and consumer confidence.
Althou^ Dataquest does not expect capital spending to grow as much in percentage terms as it did
in 1983 and 1984, we do expect rebounding confidence on the part of business and consumers to
lead to vigorous growth in coital spending. (Our
current forecast looks to growth of 24 percent and
23 percent, respectively, in 1992 and 1993.)
®1991 Dataquest Incorporated March-ReproductioD Prohibited
SEMMS Newsletters 1991 Forecasts/Business Conditioiis
0009603
The corueni of this report represents our irtterpretalion and analysis cf information ^nerally 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 confiderwe by our clients Individual companies report^ on and analyzed by Dataquest
may be clients cfthis and/or other Dataquest services. This information is not fitmished in connection with a sale or offir to sell securities or in cormectitm with the solicitation of an
o^r to buy securities. This firm and its parerU and/or their officers, stockholders, or members of their families may, fiom time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE EFFECTS OF THE GULF WAR ON SEMICONDUCTOR CAPITAL SPENDING
High technology has been a key component
of U.S. defense strategy for decades. Because of its
success in the Gulf War, insiders in Washington
D.C. report that they have never seen enthusiasm
for high technology at such a high leveL The
likelihood that this enthusiasm will jeU into support
and funding for R&D in high technology and semiconductors is now stronger than it has been for
years.
There have been debate and questions for
years in the United States about whether or not the
U.S. government could support high technology
and, if so, how to go about this. A major consequence of the Gulf War may very well be that high
technology in the United States will win substantial
support for its long-term health and growth from
the U.S. government.
George Burns
0009603
01991 Dataquest Incorporated March-Reproduction Prohibited
SEMMS Newsletters 1991 Forecasts/Business Conditi<nu
Dataquest
t
«1
ifiompamai
'The Mn li BiadstiRt Corporation
Research Newsletter
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY
OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
INTRODUCTION
In spite of a U.S. recession and the threat of
war, the worldwide semiconductor industry grew in
the fourth quarter of 1990 in both bookings and
billings. The Persian Gulf war, which began on
January 16, 1991, when the allied forces started
bombing Baghdad, might be expected to cast a pall
over the entire world economy to the detriment of
the semiconductor industry. However, Dataquest
beheves that the industry will continue to grow,
albeit modestly, through 1991. We expect quarterly
growth to be stronger in 1992. Our annual growdi
forecast by region is shown in Figure 1. Overall,
we expect 9 percent growth in 1991 and 13 percent
growth in 1992.
FIGURE 1
Annual Semiconductor Industry Growth Rates
by Regional Market
(Percentage of Dollars)
Percentage of Dollars
19S9
1990
1991
1992
Source: Dataquest (January 1991)
0 1 9 9 1 Dataquett Incoiporated Imuaiy-Repioduction Piohilnted
0009372
Semiconductar Oioi^ Newdetten 1991-04
Jhe content of this report represents our interpretation and analysis ofii^rmation 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. Individual companies reported on and analyted by Dataquest
may be clients of this and/or other Dataquest services. This itifbrmation is notfiimished in connection with a sate or cffer to sell securities or in connection mth the solicitation of an
c^rto buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, ftom time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
The reasons for our relative optimism are as
follows:
• Our monthly survey of major OEM semiconductor procurement managers continues to support
improvement in the systems market outlook.
• A trade war, brought on by patriotic fervor in the
United States, could dimipt world economies
enough to adv^sely affect the semiconductor
industry. This possibility is avoidable if the U.S.
government and its allies actively control events
that might result in protectionist U.S. policies.
• Semiconductor inventories at OEMs are less
than 20 days and within 8 days of target.
OUTLOOK FOR 1991 AND 1992
• Many semiconductor manufacturers are reporting strong bookings for the month of January.
• WSTS statistics show both bookings and billings
on an upward trend through November, the last
worldwide actuals available.
• Increasing pervasiveness of semiconductors in
electronics and consumer goods and increasing
functionahty per chip will continue to raise chip
average selling prices and allow the semiconductor industry to grow faster than the electronic
equipment industries.
• Telecommunications equipment production continues to do well, due in part to d«nand firom
eastern Europe. This will continue to drive semiconductor consumption in Europe.
• There is evidence—^in the huge i^yproval rating
of U.S. President George Bush, the large U.S.
stock maiket rallies, and signs of inq)rovetnent
in the index of leading indicators—that U.S.
consumer confidence has increased dramatically
since the bombing of Baghdad began.
• U.S. allies have pledged $45 biUion toward die
cost of ttie war thus far, thereby alleviating a
potentially onerous financial burden on one
nation.
To be sure, there are also possible hazards on
the horizon:
• Protraction and/or major expansion of the
war in die Persian Gulf could sabotage world
economies.
• Increased political and economic instability in
the Soviet Union could become a very explosive
situation with worldwide repercussions.
• Lack of soundness of the U.S. financial system
could damage the U.S. economy if massive bank
failures were to occur. This possibility can be
averted by effective action on die part of the
Federal Reserve Board, Congress, and die Bush
administration.
0009372
We have looked at several different scenarios
for semiconductor industry growth this year and
next They range from highly optimistic to highly .
pessimistic. We believe that the most likely scenario is somewhere in between, with worldwide
growdi of 9 percent in 1991 and 13 percent in
1992.
In the final months of 1990, both bookings
and billings were well ahead of the same period in
1989, at 13 and 14 percent, respectively. The same
trend holds when looking at the three months ended
November 1990 versus the three months ended
November 1989. Because of this trend and because
of renewed confidoice levels since fighting began
in the Gulf, we believe that die first and second
quarters of 1991 are going to show growth, with
most of it in the second quarter. We are forecasting
modest growth in the third and fourth quarters of
1991. We diink quarterly growth will be considerably higher in 1992 for the following reasons:
• We believe that die war will have been resolved.
• We believe that the U.S. savings and loan and
banking crisis wUl be in the solution phase.
• We believe diat psychology will play a strong
role: Just as low consumer confidence contributed strongly to the U.S. recession in die
fourth quarter of 1990, a positive mindfiramein
the electronics industry can buoy up the
semiconductor industry.
Figure 2 shows our sequential quarterly
growth history and forecast worldwide. Figure 3
shows worldwide growth by quarter versus the
same quarter a year ago.
Regionally, we expect to see the following
trends in 1991:
• Nordi American maiket growth will be strongest
in the second quarter.
• European market growth will be strongest in the
first quarto:. European semiconductor consumption benefited in 1990fix>ma boom in TV and
VCR production, which we do not believe will
be repeated diis year.
01991 Dauqueft Luxaponled JamiaijF-Reproduction Probibited
Semicaodiictar Orai^ Newdetten 1991-04
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
FiGUKE 2
Worldwide Semiconductor Industry Growth
by Sequential Quarters
Percentage of Dollars
1412
1
10
8
\
6
4
S
S.
S
2-1
N
^h
I
-21Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 40 10 2Q 3Q 40
1
II
II
II
II
II
I
1987
1988
1989
1990
1991
1992
Source: Dataquest (January 1991)
FIGURE 3
Worldwide Semiconductor Industry Growth
versus the Same Quarter One Year Ago
Source: Dataquest (January 1991)
01991 Dataqueit IneoipQnted Jamuiy-Repnxliictian Pndiibited
Scmicandiictor Otmip Newiletten 1991-04
0009372
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
• Ji^an will show the weakest quarterly growth of
any region this year, largly due to the
unprecedented economic challenges it is facing
in its stock market and real estate market and die
slowdown in consumer spending in the United
States, upon which a large part of Japan's
semiconductor consumption depends.
• The Asia/Pacific-Rest of World (ROW) market,
which slowed in the fourth quarter of 1990 due
to a faUoff in clone demand, will resume growth
in the second quarter of 1991.
In general, the outlook for the Asia/Pacific
markets continues to be brighter than that for other
regions for the following reasons:
• The Asia/Pacific countries' GDPs in general
continue to grow at high single-digit rates.
• Much of the Asia/Pacific semiconductor demand
will come from products to be sold within the
coimtry of manufacture. In fact, Japanese
conq>anies are now producing goods in Asian
countries for sale there radier than for export to
other regions.
0009372
• This market is still the smallest, least mature
regional market; therefore, it can support a
higher percentage growth than can other regions.
Our 1992 outlook calls for Asia/Pacific-ROW
to remain the fastest-growing regional market,
followed, in order of growth, by North America,
Europe, and Japan.
DATAQUEST ANALYSIS
Never before has Dataquest forecast semiconductor industry growth during a global conflict that
could affect worldwide economic powers. We
believe that enough positive factors exist to result
in modest industry growth both diis year and next.
The war could have either a significantly positive
effect or, conversely, a significandy depressing
effect on the semiconductor industry. We have
chosen a scenario in which most volatile effects are
counterbalanced by other influences, and life continues on, though perlu^s not at the frenetic pace
of the 1980s.
Patricia S. Cox
01991 Diuquett locoiponied Januiy-Reproduetian Probiliited
Semicaaduettir Oioup Newiletten 1991-04
'mim
*%
S ^
mi
Dataoyest
aoHi^nyof
The DunSTBradsbcet Coq>oration
Research Bulletin
REGIONAL WAFER FABRICATION EQUIPMENT MARKET FORECAST
SUMMARY
An understanding of regional market trends
and driving forces is essential to forecasting the
growth of the worldwide wafer fabrication equipment market This newsletter provides Dataquest's
forecast for the regional wafer fabrication equipment ntiarkets. In our January 1991 newsletter, we
forecast the worldwide wafer fabrication equ^ment
market to grow at a compound annual grow^ rate
(CAGR) of 14.4 percent from $5.6 billion in 1990
to $10.9 billion by 1995. The different regional
wafer fabrication equipment markets have unique
characteristics and are expected to have widely
different growth rates during the next five years.
Dataquest believes that the Asia/Pacific-ROW
(Rest of World) and European wafer fab equ^nnent
market growth rates wiU substantially outpace the
U.S. and Japanese market growth rates during the
next five years. The continuing globalization of the
semiconductor industry, the emergence of vigorous
capital spending by the newly industrialized countries of the Pacific Rimi, and the race by semiconductor manufacturers to achieve a manufacturing
presence in each continent will have profound
effects on regional market trends within the wafer
fab equipment industry.
REGIONAL WAFER FABRICATION
EQUIPMENT MARKET FORECAST
Ikble 1 displays the regional wafer fabrication
equipment market forecast. Through sheer size and
momentum, the Japanese wafer fabrication equipment market is expected to continue to prevail as
the largest market during the next five years,
although it wiU grow at the slowest rate among the
four major geographical markets. Many Jeqpanbased device manufacturers are r^idly shifting
their ci^ital investments to offshore fabs in their
pursuit of globalization. However, leading-edge
technologies such as 8-inch 16Mb DRAM fabs will
continue to be built and ranq)ed up in Japan before
transfer to offshore clone fabs.
The U.S. market is e3q)ected to continue to be
the second-largest wafer fab equipment market in
the world. The influx of offshore Japanese fabs,
together with 8-inch submicron fabs built by leading U.S. captive and merchant semiconductor
manufacturers, wUl contribute to the momentum of
the U.S. equqnnent market. However, U.S. semiconductor manufacturers have largely retreated
from conmiodity-device markets such as the
DRAM and SRAM markets, histead, they have
focused on high-margin, value-added designs,
TABLE 1
Regional Wafer Fabrication Equipment Market Forecast
(Millions of Dollars)
North America
Japan
Europe
Asia/Pacific-ROW
Worldwide Total
1990
1991
1992
1993
1994
1995
1,691
2,414
689
768
5.562
1,835
2,651
834
890
6,211
2,239
3,171
1.068
1,211
7,689
2,700
3,785
1,340
1,657
9,482
2,936
4,029
1,517
1,935
10,417
3,053
4,216
1,655
1,977
10,901
CAGR (%)
1990-1995
12.5
11.8
19.2
20.8
14.4
Scuzce: Dataquest (Febniary 1991)
0 1 9 9 1 Dataquest bicorporated Febniaiy-Repxoduction Prohibited
SEMMS Newsletters 1991 Forecast/Business Conditioiis
0009398
The content of this report represents our interpretation and analysis (^information generally available to the public or released by responsible individuals in the subject companies, but
is not guaranteed as to accuracy or avr^leteness. It does not ojntain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This information is not furnished in amnection with a sale or offsr to sell securities or in connection with the solicitation of an
o^r to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
REGIONAL WAFER FABRICATION EQUIPMENT MARKET FORECAST
which fetch higher revenue streams per dollar of
capital investment because of their relatively higher
average device prices. Many new niche-oriented
U.S. device companies have elected to go the fabless route in order to escq)e the crippling expenses
of submicron fabs. Dataquest e?q)ects a significant
shift in the ownership mix of U.S.-based fabs
through the 1990s.
The European wafer fab equipment market
win enjoy healthy growth during the next five
years as Jiq)anese, U.S., and Asian semiconductor
companies set up Europe-based fabs to cater to a
unified European market as well as to large blocs
of Eastern European countries. In effect, European
semiconductor manufacturers will face steeply
escalating conq)etition in their own backyards.
The Asia/Pacific-ROW market is in a transition from being an export-driven market to becoming an inwardly focused, consumer-oriented market. South Korea and Taiwan continue their
strategic government-backed coital investments in
leading-edge DRAMs, ASICs, and foundry businesses. In addition, Japanese semiconductor companies are building fabs in other low-labor-cost
Asian countries such as Malaysia, Thailand, and
China in order to cater to more mature device
product demands.
Figure 1 illustrates the relative proportions of
the regional wafer fabrication equipment markets in
the world market between 1990 and 1995. The
rapid growth of the European and Asia/PacificROW equipment market between 1990 and 1995 is
readily apparent.
DATAQUEST CONCLUSIONS
Semiconductor manufacturing is emerging as
a truly global industry; nevertheless, important
regional differences are emerging between the
various geographical wafer fabrication equqnnent
markets. Dataquest believes that it is inqx>rtant to
understand the forces and end-use applications
driving the various regional wafer fab equipment
markets. The wafer fab equipment markets are
experiencing rapid change as they develop in order
to support global semiconductor producers. We
expect European and Asia/Pacific fab equipment
markets to grow at the fastest rates during the next
five years as those regions act swiftly to balance
chip demand-and-supply forces. The Japanese
equqnnent market is expected to continue its role
as a high-volume, DRAM-driven leading-edge
market Cs^tive U.S. ciap producers, together with
value-added merchant U.S. chip companies and
offshore plants, are expected to maintain the U.S.
equ^nnent market as the second largest in the
worM.
Krishna Shankar
FIGURE 1
Regional Wafer Fab Equipment Markets
IV%ri Japan
^ ^ 1 North America
p g ] AsIa/Paclflc-ROW
I
1990
$5.6 Billion
I Europe
1995
$10.9 Billion
Source: Dataquest (February 1991)
0009398
01991 Dataquest Incoipoiated Febnury-^epioductiaii PiDbiUted
SEMMS Newsletten 1991 Foiecast/Bunnen Conditiaai
^
Dataoyest
acompanyof
i h c Dwifli Bradsbcct CoqHHatKHi
Research Bulletin
1991 SILICON WAFER FORECAST
The worldwide market for silicon wafers
(prime polished, test, and epitaxial wafers) grew
11.9 percent in 1990. Dataquest's preliminary estimate of worldwide consumption is 2,028 million
square inches (msi) for the calendar year. Demand
grew in all regions except Europe; demand in
Europe remained flat (see Table 1).
The Asia/Pacific region continued its explosive demand for wafers, led primarily by semiconductor plants located in South Korea and Taiwan.
Demand for silicon wafers totaled 169 msi in 1990,
up 32.8 percent from 1989. Based on the silicon
wafer consumption trend, it is evident that South
Korean semiconductor manufacturers, which focus
heavily on MOS memory (it is about 60 percent of
their production on a revenue basis), increased their
device production in spite of falling memory prices
with the objective of winning more market share.
Dataquest expects future growA in demand
for silicon wafers in Asia/Pacific to outstrip growth
in other regions of the world. The five-year forecast compound annual grov/th rate (CAGR) for
Asia/Pacific is 19.8 percent. Consumption wUl
increase 248 msi, a unit growth surpassed only by
Japan. In the long term, China looks especially
promising. Dataquest believes that recent
announcements by Japanese semiconductor companies of plans to build front-end facilities on the
mainland signal the beginning of an investment
cycle that will result in rapid growth of silicon
wafer demand in China.
Total European consumption was flat in 1990
because several European semiconductor manufacturers saw flat growth in MOS and bipolar
products. Only the analog, discrete, and optoelectronic device markets grew in 1990; however,
TABLE 1
Silicon Consumption Forecast by Region
(Millions of Square Inches)
CAGR (%) Delta msi
1990-1995 1990-1995
1989
1990
1991
1992
1993
1994
1995
United States
Percent Growth
560
13.0
620
10.7
662
6.8
733
10.7
847
15.5
861
1.7
855
(0.7)
6.6
235
Japan
Percent Growth
909
17.7
1,022
12.5
1,101
7.7
1,221
10.9
1,400
14.7
1,434
2.4
1,447
0.9
7.2
425
Europe
Percent Growth
216
10.2
217
0.1
244
12.6
285
17.0
343
20.2
368
7.4
395
7.3
12.8
178
Asia/Pacific
Percent Growth
127
29.4
169
32.8
221
30.6
273
23.6
338
23.9
389
15.1
417
7.0
19.8
248
1,812
16.0
2,028
11.9
2.228
9.9
2,512
12.8
2,928
16.6
3,053
4.3
3,114
2.0
9.0
1,086
Total
Percent Growth
Source: Dataquest (Jamiaiy 1991)
®1991 Dataquest Incorporated January—Reproduction Prohibited
SEMMS Newsletters 1991 Forecasts/Business Conditions
0009324
The content cfthis report represents our interpretation and analysis cf information generally available to the public or released by responsible individuals in the subject companies, but
is not guaranteed as to accuracy or ojmpleteness. It does not contain material prcfvided to us in ctmfidence by our clients. Individual ampanies reported on and analyze by Dataquest
may be clients of this and/or other Dataquest services. This information is notfiirrnsh^ in amnection with a sale or offer to sell securities or in connection with the solicitation (rfan
qffi:r to buy securities This firm and its parent and/or tiieir officers, stockholders, or members of their fomilies may,fromtime to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
1991 SILICON WAFER FORECAST
TABLE 2
Semiconductor Production Forecast
(Millions of Dollars)
United States
1989
1990
1991
1992
1993
1994
1995
21,919
21,868
23,829
27,690
33,915
36,545
38,461
CAGR (%)
1990-1995
12.0
8.7
(0.2)
9.0
16.2
22.5
7.8
5.2
30,275
13.4
28,390
29,353
32,875
39,593
42,176
44,255
Percent Growth
(6.2)
3.4
12.0
204
6.5
4.9
Europe
Percent Growth
6,789
17.3
7,604
9,244
21.6
14,327
26.2
15,849
10.6
17,515
12.0
11,353
22.8
Asia/Pacific
2.472
4,097
5,116
29.3
24.9
6,466
26.4
7,667
18.6
8,286
8.1
21.2
32.3
3,169
28.2
61,455
61,031
66,523
77,034
94,301
102,237
108,517
12.2
8.4
6.1
Percent Growth
Japan
Percent Growth
Total
Percent Growth
12.7
(0.7)
9.0
15.8
22.4
9.3
18.2
10.5
Source: Dataquest (January 1991)
because these markets are small, the growth had
little impact on the demand for sUicon wafers.
Even so, demand is forecast to have a 12.8 percent
CAGR through 1995, the second fastest growth
rate, because of the increased level of investment in
fabs by U.S. and Japanese companies. E>emand for
silicon is forecast to be up 178 msi during the
five-year period.
The demand for silicon wafers was up
12.5 percait in Japan, reaching 1,022 msi in 1990.
The fact that demand for silicon wafers in Japan
grew at the same time that device revenue fell
2.1 percent is attributable to the severe pricing
pressures in MOS memory, which is a large segment of the J^anese market. Dataquest expects
silicon wafer demand in J{^an, currently the largest
market in the world, to continue to grow during the
next five years at a 7.2 percent CAGR. Although
the growth rate is modest, J ^ a n will far surpass
the other regions of the world in terms of unit
growth; consumption is estimated to increase by
425 msi during the forecast period.
Growth of 10.7 percent in the U.S. market in
1990 was a little below worldwide average growth.
However, demand for epitaxial wafers e^loded;
merchant shipments grew over 25 percent, reaching
0009324
70 msi. Much of the demand for epi wafers was
due to strong growth in nucroprocessor production
and MOS memory devices using trench technology.
Prime polished and test wafers accounted for most
of the 620 msi of silicon consiimed in the United
States; these products grew in the 7 to 9 percent
range. Dataquest's five-year forecast estimates that
silicon wafer demand will have a moderate
6.6 percent CAGR.
Dataquest is currently conducting its worldwide silicon study, which will be completed by the
beginning of second quarter. At that time, we will
publish a more detailed breakout of the 1990
worldwide silicon wafer market and an update to
our forecast. The silicon forecast is based on a
semiconductor device forecast (see Table 2), Please
note when coaspaiiDg the two forecasts that the
device forecast is reported in U.S. dollars and
therefore is impacted by exchange rates and
changes in device prices. Consequently, correlation
between the forecast trends in the two tables may
not be obvious.
Mark FitzGerald
01991 Dataquest Ihcoiporated January-Reproduction Prohibited
SEMMS Newsletters 1991 Forecasts/Business Conditions
Swfi
•m^X
••^-'rk
DataQuest
a ccwi^nyof
the Dun & BradstreetCorpcHation
I
••"H'V.,.
Research Bulletin
WAFER FAB EQUIPMENT MARKET NEAR-TERM FORECAST
SUMMARY
Dataquest is cautiously optimistic about
the near-term outlook for the wafer fabrication
equipment market. We expect the worldwide firontend equipment market to grow by 12,0 percent in
1991 compared with 1990. The short-tenn outlook
is clouded by current worldwide macroeconomic
and political uncertainties. In the long term, however, Dataquest believes that the wafer fabrication
equipment niarket will enjoy healthy growth at
a compound annual growth rate (CAGR) of
14.4 percent, from $5.6 billion in 1990 to
$10.9 biUion in 1995. Wafer fabrication equipment
companies with global presence, financial muscle,
and innovative customer-driven technology solutions can be optimistic about flieir long-term future
in an increasingly chip-pervasive world.
ASSUMPTIONS IN THE FORECAST
Dataquest's wafer fabrication equipment forecast assumes that major world economies do not
sUde into a major, sustained recession in 1991. We
also assume that the Middle East political crisis is
resolved without a shooting war. If either of these
events happen, capital spending on property, plant,
and equipment will almost certainly be scaled back
in the short term. Although the 1991 outlook is
&aught with poUtical and economic uncertainty,
Dataquest is buUish about the long-tsm growtti
prospects for the wafer fabrication equipment
industry. We will revisit our capital spending and
wafer fab equipment forecast in ftie second quarter
of 1991.
WAFER FABRICATION EQUIPMENT
MARKETS
Table 1 presents Dataquest's five-year forecast for wafer fabrication equipment by segment.
As the semiconductor industry pushes into the submicron era, process complexity and fabrication technology requirements condnue to escalate
dramatically. Lithography, deposition, and etch/
clean equipment continue to be technology drivers
that fuel the wafer fabrication equipment industry's
growth. The wafer fabrication equipment market
win exhibit robust growth over the next five years
because of both unit shipment increases to support
additional production and increased average selling
prices driven by increased process complexity.
In 1990, the regional maiicet growth trends
were varied. In 1990, the U.S. and J^anese wafer
fabrication equipment markets remained relatively
flat compared with 1989 levels. The European
market grew significantly because of offshore U.S.
and Ji^anese fabs being built in Europe. However,
the Asia/Pacific-ROW regional market, which
enjoyed spectacular growth in 1988 and 1989,
e3q)erienced a severe contraction in 1990. Dataquest expects wafer fab equipment spending in
Asia/Pacific to resume healthier growth in 1991.
WAFER FABRICATION EQUIPMENT
COMPANIES
Dataquest continues to observe consoUdation
and globalization in the worldwide wafer fabrication equipment industry. The large global companies are continuing to grow, while many small,
regionally focused companies are struggling to survive. In 1990, U.S. and European equipment companies reported a disparate mixture of positive and
negative growth rates relative to 1989. In contrast,
Japan-based equipment companies uniformly
reported flat or positive growth in 1990 due to the
relatively strong domestic Japanese economy and
offshore expansion of j£qpan-based device companies. The 1990s will see significant changes in the
balance of power between European, Japanese, and
U.S. equipment companies as they hasten to foUow
their increasingly globalized semiconductor
manufacturers.
Krishna Shankar
Peggy Marie Wood
®1991 Dataquest lacoiporated January-Reproduction Prohibited
SEMMS Forecasts/Business Conditions
0009226
IJie content of this report represents our interpretation and analysis qfir^rmation 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 provitled to us In confidence by our clients Individual companies reported on and analyzed by Dataquest
may be clieras of this and/or other Dtuaguest services. This information is not furnished In connection with a sale or offer to sell securities or in connection 'Aith the solicitation of an
(^er to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position In the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
TABLE 1
Worldwide Wafer Fab Equipment Market Forecast
(Millions of Dollars)
8 *
Lidiography
Contac^Proximity
Projection
Steppers
Direct-Write Lithography
Maskmaking Lithography
X-Ray
Total Lithogr^hy
Autom. Photoresist Proc. Eqp.
Etch and Dean
Wet Process
Diy Strip
Ehy Etch
Total Etch and Clean
Deposition
CVD
PVD
Silicon Epitaxy
MOCVD/MBE
Total E}eposition
Difiiision
Rapid Thermal Processing
Total Ion Implantatitxi
Process Control
CD (Optical and SEM)
Wafer Inspection
Other Process Control
Total Process Control
Factory Automaticm
Other Equipment
Total World Fab Equipment
Percent Change
Source: Dataquest (lanuaiy 1991)
1990
1991
1992
1993
1994
17
80
1,115
75
81
4
1,372
311
16
86
1,264
85
94
10
1,555
348
16
99
1,576
105
116
15
1.927
425
15
112
1,970
131
145
31
2,404
523
15
119
2,145
154
167
50
2,650
569
278
110
610
998
312
120
690
1,122
385
150
845
1,380
465
190
1,068
1,723
500
210
1,180
1,890
550
360
60
109
1,079
300
26
444
630
405
55
119
1,209
330
30
490
770
490
85
142
1,487
400
45
632
970
600
74
166
1,810
500
60
774
1,070
650
105
182
2,007
550
70
835
166
98
368
632
205
195
5,562
(6)
192
109
391
692
225
210
6,211
12
239
136
483
858
275
260
7,689
24
290
165
574
1,029
340
320
9,483
23
320
181
619
1,120
375
350
10,416
10
2
2
10
IK
pi?
Dataquest
a company of
The Dun & Bradsticct CcHpcnration
Research Newsletter
A GLIMPSE AT FUTURE 64Mb DRAM TECHNOLOGIES
SUMMARY
LITHOGRAPHY TRENDS
The IEEE International Solid State Circuits
Conference (ISSCC) held every February is a good
barometer of future trends in device technology and
applications. The 1991 conference featured several
experimental versions of 64Mb DRAM devices.
Although these devices are at least five years away
from volume production, they provide a glimpse of
future high-volume process technologies. In this
newsletter, Dataquest analyzes key implications of
these prototype 64Mb DRAM technologies for the
semiconductor equipment, manufactiu-ing, and
materials industries in the years ahead.
All of these 64Mb DRAMs were fabricated
with 0.4-micFon design rules using optical lithography tools. Fujitsu and Mitsubishi opted for i-line
steppers and Matsushita and Toshiba chose excimer
laser steppers. The astonishing progress of optical
lithography in combination with technology such as
phase-shift masks pushes X-ray lithography even
further out into the future. Semiconductor manufacturers have a huge installed base of investment and
experience in optical lithography that they are
reluctant to throw away. Japan-based DRAM companies are racing to convert development results in
phase-shift masks into commercially useful technologies to extend the lifetime of optical lithography tools through the 64Mb DRAM generation and
potentially to the 2S6Mb DRAM generation.
Issues such as global and local planarization,
depth of focus, wafer flatness, and intrafield focus
on large fields may yet force semiconductor
manufacturers to eventually migrate to X-ray
lithognqjhy, which has far higher depth-of-focus
latitude. However, X-ray Uthogre^hy has to contend with the challenges of IX maik technology.
Hie prohibitive costs associated with synchrotron
orbital rings (SORs) for X-ray lithography, together
with the technical challenges of IX mask materials,
mask fabrication, inspection, and repair, have
prompted 64Mb DRAM manufacturers to stay with
the evolutionary, incremental advantages of optical
lithography.
Dataquest believes that the extension of optical lithography usiog i-line and excimer laser steppers in combination with phase-shift mask technology may enable the 64Mb DRAM device to foUow
the traditional decrease in the cost-p>er-bit curve.
Given the extension of optical lithography to the
64Mb DRAM generation, Uthography equqnnent
companies need to focus on high-throughput, widefield steppers that can offer better productivity in
64Mb DRAM TRENDS
Table 1 illustrates the key features of
experimental 64Mb DRAMs unveiled by Fujitsu,
Matsushita, Mitsubishi, and Toshiba at ISSCC
1991. Dataquest believes that DRAM companies
will continue to push optical lithography to
0.4-micron geometries for the 64Mb DRAM. All of
the 64Mb DRAM devices were characterized by
multiple levels of poly/polycide and double-level
interconnect technology. Gate and capacitor dielectric thickness values are expected to be in the 50to 100-angstrom range. All four companies used
variations of a stacked-capacitor cell scheme.
U.S.-based DRAM manufacturers have traditionally favored a trench c£^acitor-based memory
cell. La contrast, Japan-based DRAM companies
favor the simple stacked capacitor scheme over the
more complex trench capacitor scheme with its
attendant problems of trench etch damage and
trench sidewall leakage currents. Toshiba i^pears
to have the most aggressive 64Mb DRAM design.
Toshiba's use of excimer laser hthogr^hy, together
with the asymmetric stacked trench capacitor
design, yields the smallest cell size (0.9 x 1.7 um^)
and the fastest speed (33ns).
01991 Dataquest Incorporated March-Reproduction Prohibited
SEMMS Newsletter! 1991 Technology Trends
0009622
The contenJ of this report represenls 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- Individual companies reported on and analysed by Dataquest
may be clients of this and/or other Dataquest services. This infitrmation is notpmished in connection with a sate or offer to sell securities or in connection with the solicitation of an
offer to buy securities This firm attd its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short posltitm in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
TABLE 1
Key Features of 64Mb CMOS DRAMs at ISSCC 1991
Company
Minimum
Feature size
(Microns)
i
3
^i
rf 1.
I
Metal
Levels
Gate Oxide
Thickness
(Angstroms)
Capacitor
Type
Cell Size
um X urn
Fujitsu
0.4
Mine Phase-shift
4
2
NA
Double-fin stacked
1.0 X 1.8
Matsushita
0.4
KrF excimer laser
3
2
120
Tkinnel stacked
1.0 X 2.0
Mitsubishi
0.4
I-line
3
2
120
Dual-cellplate
stacked
1.0 X 1.7
Toshiba
0.4
KTF excimer hsa
4
2
50
Asymmetric
stack trench
0.9 X 1.7
NA = Not sToliblc
Souice; ISSCXVDiuqueit (Maich 1991)
i
Lithography
Poly/
Polycide
Levels
A GUMPSE AT FUTURE 64Mb DRAM TECHNOLOGIES
spite of highCT average selling prices (ASPs). Significant opportunities exist for conq)anies to target
new business areas such as i-hne and excimer laser
photoresists, ancillary lithography chemicals,
phase-shift masks, mask coatings, niask etch, and
mask inspection/repair equipment.
ETCH/CLEAN TRENDS
Dataquest estimates that the number of mask/
etch levels will almost double between the 1Mb
DRAM (16 levels) and the 64Mb DRAM (about 30
levels). In fact, the number of wet clean/dry etch
processes will exceed the number of masking
processes because of the addition of more elaborate
wet/dry vapor cleans as well as blanket (maskless)
etchback steps such as trench refill etchback, LDD
spacer etchback, and contactMa plug etchback,
intermetal planarization etchback. The unique
requirements of the 3-D stacked or trench 64Mb
DRAM c^acitor offer extraordinary challenges to
the abihty of wet chemicalA^sqxn- phase cleans to
truly "clean" the wafer without adding additional
particles and contamination.
Dry etch equipment has to offer extremely
high selectivities, uniformity, critical dimension
(CD) control across 8-inch wafers, and low ionization damage in order to etch 0.4-micron gate features. A variety of plasma sources are being considered in order to handle the stringent processing
requirements of 64Mb DRAM dry etch processes.
New gas chemistries such as bromine, NF^, and
other non-fluorocarbon processes offer significant
processing challenges to gas supplio's and dry etch
equipment companies.
DEPOSITION TRENDS
DRAM manufacturers have already switched
fix)m single-level metal to double-level metal for
the 16Mb DRAM generation. The challenges
associated with metal step coverage dramatic^y
increase as contact and via dimensions approach
the 0.4-micron level. CVD titanium nitride, CVD
tungsten, and CVD polysiUcon are being examined
as viable candidates for contact plug processes.
Meanwhile, the efforts to inqirove the step coverage of sputtered aluminum and refractory barrier
metals such as titanium nitride continue vigorously.
Many opportunities exist for materials companies
to develop new sputtering materials and CVD
source materials for interconnect applications in the
64Mb DRAM generation.
01991 Datiqueit Incoiponted Much-Reproduction Prohibited
SBMMS Newdettert 1991 Technology Trends
The polysilicon CVD equipment market is
expected to grow dramatically over the next five
years in order to cater to muslurooniing j^pUcations
for high-quality polysilicon films at multiple levels
in the 64Mb DRAM process. For exan^le, Toshiba
is reportedly planning to use four levels of poly/
polycide films in its 64Mb DRAM process.
Stacked capacitors and trench c^acitors will use
multiple poly depositions to achieve the desired
cell cq>acitor area. Many new types of poly CVD
equipment such as improved vertical LPCVD poly
tubes and integrated cluster tools incorporating
rapid thermal oxidation/nitridation (RTO/RTN),
low-pressure poly CVD, and low-pressure tungsten
sUicide CVD may emerge in response to these
applications.
Interlayer dielectrics between poly and firstlevel metal and intermetal dielectrics between
metal levels need to be highly planarized because
of metal step coverage, bridging, depth of focus,
resist uniformity, and ovo'-etch considerations in
64Mb DRAM wafers. In addition to the famihar
spin-on-glass planarization schemes, Dataquest
believes that 64Mb DRAM companies will
examine other global planarization techniques such
as biased electron cyclotron resonance (ECR) CVD
techniques, chemical-mechanical polishing, l^OSbased plasma-enhanced CVD oxide fill/etchback,
and in-situ deposition/low-temperature reflow
oxides. Tungsten, poly, aluminum, and copper
CVD plugs are being e:q>lored for contact and via
fills. The choice of the optimum planarization and
back-end interconnect process will have profound
effects on 64Mb DRAM speeds, yield, and
reliability.
DIFFUSION/IMPLANT TRENDS
Vertical diffusion and LPCVD tubes will
probably be used for all diffusion and oxidation
processes on 8-inch 64Mb DRAM wafers. Vertical
furnaces offer high-quality thin oxides, thermal
nitride, and polysilicon. Vertical tubes are also
more conq)atible with the automation and film
uniformity requirements of 8-inch fabs. Loadlocked vertical diffusion fiimaces may be used to
implement tube-to-tube transfer between oxidation,
nitridation, and LPCVD poly/nitride processes.
The numbo' of implant steps continues to rise
significantly in order to precisely control the electrical behavior of 0.4-micron geometry transistors.
In addition to the traditional requir^nents for dose
uniformity and low particulates across 8-inch
wafers, continuously variable tilt angles and
0009622
A GLIMPSE AT FUTURE 64Mb DRAM TECHNOLOGIES
parallel beam scanning are expected to become the
norm for mq)lanting 3-D 64Mb DRAM device
structures.
PROCESS CONTROL TRENDS
CD and wafer-inspection equipment companies will enjoy major business opportunities at the
64Mb DRAM generation. The process of analyzing
variations in critical dimensions at the 0.4-micron
level across 8-inch wafers is a major challenge.
The move toward integrated processes will lead to
the loss of critical intermediate CD and wafercondition ioformation. Some equipment companies
are evaluating the incorporation of in-situ metrology tools such as CD SEM measurement chambers
and particle-detection/wafer-inspection chambers
onto cluster tool platforms.
Thin films and resistivity measurement
systems will face similar challenges in measuring
thin oxides and shallow doped junctions. Electrical
measurement techniques may be used to augment
physical thin-film thickness and resistivity
measurements. •
DATAQUEST CONCLUSIONS
Dataquest believes that DRAM process technology wUl continue its evolutionary progress
between generations. The extension of optical
lithography and the stacked capacitor cell structure
to the 64Mb DRAM devices are aimed at keeping
the DRAM cost per bit on its historical decline.
Dramatic increases in the complexity of lithography, interconnect, planarization, dry etch, and
process-control processes may push the price tag of
a 8-inch high-volume 64Mb DRAM fab to well
over $600 million. At the 0.4-micron 64Mb DRAM
level, interconnect process complexity and poformance will be the limiting factors that control the
device speed and cost per bit.
Krishna Shankar
The topics covered by SEMMS newsletters are selected for their general mteiest to SEMMS clients, which include wafer fab equqiment
siqiplieis, semiconductor materials conqianies, and semiconductor device mannfactums. Tlie topics selected indicate the broad range of research
that is conducted in the SEMMS group. Clients, however, often have specific information requirements that eittier go beyond the level of detail
contained in (he newsletters or beyond the scope of what is normally published in die newsletters. In order to provide oon^Iele decision support
to our clients, Dataquest has a consulting service available to handle these additional information needs. Please call Stan Broederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
0009622
01991 Dataquett Incorponted Mucb-ReproductioD Prohibited
SEMMS Newfletter* 1991 TBcfanology Trendi
i
i^p
WMm
'^tmWm
DataQuest
acDfiHanyDf
STBraddnxl CDrporHim
t ~rfKDun
^m^.'-mT
J ^ ; ' " "
•
"'
•
''
Research Newsletter
EQUIPMENT AND CONTAMINATION IN THE
Process equipment costs today can exceed
$2 million. Fab costs today are $300 million and
will reach $1 billion by the year 2000. These sums
of money are large, yet they can be jeopardized by
particles smaller than one-tenth of a micron in size.
This newsletter focuses on where these particles
come from and their effect on process equqnnent
and facilities in the future.
SOURCES OF PARTICLES TODAY
Eliminating particles as sources of contamination is iinportant in semiconductor manufacturing
because particles can cause "killer" defects in a
device's circuits. Killer defects are defects that
cause a device to malfunction. Particulates onetaath of the size of the miniTnum linewiddi of a
device can cause a killer defect. For today's
leading-edge devices (e.g., 4Mb DRAMs), this
means that particles smaller flian one-tenth of a
micron can cause a killer defect. Defects inq>act
jdelds, and yields impact the bottom line. Thus, an
investment of $1 billion can be put at risk by a
particle one one-thousandth the thickness of a
human hair.
Dr. Venn Menon of Sematech said at a recent
Microcontamination Conferoice that human beings
and the clean room itself accoimt for only ^rproximately 25 percent of particulate contamination in
today's state-of-the-art facility. Equipment and
process account for ^yproximately 75 percent This
percentage is very different from what it was
10 years ago. Then, human beings, clean room
practices, and the clean room itself were major
sources of contamination.
According to Dr. Menon, by 1995 the contribution of hiiman beings and the clean room to
particulate contamination will decline even further;
only 10 percent of particulate contamination will be
attributable to these sources. The reasons for this
shift in importance of people and the clean room as
1990s
sources of contamination are several: continuing
improvements in clean room practices, increased
use of automation, isolation of human beings from
the wafer, use of microenvironments, and the
advent of sub-Class 1 clean rooms.
PARTICULATE SOURCE BY EQUIPMENT
TYPE
At the same conference, Dr. Menon also
presented data on the source of particulate contamination by equipment type (see Figure 1). Although
these data were specific to Sematech's fab in
Austin, Texas, and therefore may not be representative of the industry as a whole, diey were nonetheless instructive.
One further caveat to the data in Figure 1 is
that particulate counts were measured by particles
per bare wafer pass. This type of measurement is
FIGURE 1
Equipment-Sourced Particulate Contamination
by Equipment Type
Plasma Etch/
Deposition
7%
Source: Sematech, Dataquest (February 1991)
01991 Dauquest Incoiponted Febniaiy-^epioductian Pndiibited
SEMMS Newsletteis 1991 Technology Ttends
0009327
The conlenl of this report represents our interpretation and analysis of in^rmation 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 tnaterial provided to us in confidence by our clients. Individual companies reported on and analysed by Dataquest
may be clients of this and/or other Dataquest services. This infomwtion is not furnished in connection with a sale or q^r to sell securities or in connection with the solicitation of an
c^T to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentitmed and may sell or buy such securities,
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
EQUIPMENT AND CONTAMINATION IN THE 1990s
FIGURE 2
Fab Costs
Equipment and Facilities
Percentage of Total Fab Costs
10090
60
70^
Facilities
Costs
Equipment
Costs
60
50
40-1
30
20
10
19B0
1963
1987
1990
1993
1996
Source: Dataquest (February 1991)
suitable for equipment used for wet cleaning;
however, it does not capture the particles formed
from process reactions during plasma etch and
PVD/CVD.
It is ironic that the cleaning process, which
after aU is supposed to remove particulates, is itself
a source of particulate contamination (i.e., IS percent). Because of this fact, dry-wafer-cleaning technologies, which are cleaner than wet-cleaning technologies, win be the technology of choice for
state-of-the-art fabrication by the mid-1990s.
Dr. Menon estimates that wet processing will
decline from over 80 percent of today's cleaning
steps in a state-of-the-art fab to less than 35 percent
by 1995. By the year 2000, all cleaning in a
state-of-the-art fab will very likely be dry.
Because equipment and process wiU account
for 90 percent of particulate contamination by
1995, in-situ equipment monitcning will become
more important and prevalent. It will not only
provide manufacturers with particulate measurement within the process chamber during a process,
it will also allow for the real-time correction of
process equipment problems.
DATAQUEST CONCLUSIONS
Contamination control engineers and semiconductor manufacturers wiU seek the sources of
particulate contamination less in people and the
clean room and more in equipment and process.
Because dry-wafer cleaning is cleaner than
0009327
wet-cleaning technologies, there will be a shift
from wet-cleaning processes to dry-wafer-cleaning
processes. By the end of the decade, virtually all
cleaning in state-of-the-art manufacturing will be
dry cleaning. Anotha: result of the prominence of
equipment and process as sources of particulate
contamination is that in-situ monitoring will
become increasingly important.
Equipment and process will continue to be
major sources of particulate contamination in the
1990s. Therefore, equipment vendors not only will
have to face the sub-half-micron challenges of
resolution, etch, and deposition, they also will have
to face the challenge of eliminating particles onetenth the size of the technology they were designed
to fabricate. This challenge will cost money.
To make their equipment cleaner, equq)ment
companies' R&D expenses will continue to rise, as
wiU the likelihood that the cost of manufacturing
the equipment wiU escalate. The result wiU be
cleaner equipment and processes—and more
expensive equipment.
Finally, because both people and clean room
contributions to particulate contamination will
decline relative to equipment and processes, we can
expect facility costs in the 1990s, while continuing
to rise at a faster rate than in the past, to rise at
rates slower than the growth of equipment costs for
a new state-of-the-art fab. Therefore, as shown in
Figure 2, equipment costs wiU continue to grow as
a percentage of total fab costs.
George Bums
01991 Dataquest Incoipoiated Febniaiy—Reproduction Prohibited
SEMMS Newsletters 1991 Technology Trends
•^Ji^^t^JJ.
^t^-J'it
•
^
Dataquest
acoiTi|»nyof
l l i c Dun & Bradstiect CcMporatKHi
; V / ' " - • • ' "•
". ^ - i
- I- - : j
;s!fS^
^iS^'^M'
Research Newsletter
A QUICK LOOK AT
TEXAS INSTRUMENTS
This newsletter provides a quick look at
Texas Instruments' (TI's) semiconductor capital
spending, R&D spending, capacity by line geometry, planned facilities, and recent company highlights as related to semiconductor manufactunng.
This newsletter is part of the "Quick Look" series
of newsletters from Dataquest's Semiconductor
Equipment, Manufacturing, and Materials Service
(SEMMS).
SEMICONDUCTOR CAPITAL SPENDING
AND R&D SPENDING
Figure 1 graphically illustrates TI's semiconductor capital spending and semiconductor
R&D spending expressed as a percentage of semiconductor revenue by year from 1975 through
1990.
MANUFACTURING FACILITIES
Additions to TI's manufacturing facilities in
1991 and planned additions beyond 1991 are shown
in Table 1. Figure 2 illustrates the percentage distribution of TI's existing worldwide semiconductor
capacity by line geometries.
COMPANY HIGHLIGHTS
The following discussion highlights significant events for TI.
FIGURE 1
Texas Instruments—Semiconductor Capital Spending and R&D Spending by Year
Capital Spending (IVIillions of Dollars)
R&D as a Percentage of Revenue
720 H
1-20
Capital Spending
Research and Development
640
560
-15
480
400
-10
320
240- 5
160--:
80
0
^
KN
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987
1990
Source: Dataquest (February 1991)
01991 Dataquest Incoiporated Fetnuaiy-Reproduction Prohibited
SEMMS New>letten 1991 Company Information
0009525
77K 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 Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This irifbrmation is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of their fiimilies may, from time to time, have a long or short position in the securities
rtientioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT TEXAS INSTRUMENTS
TABLE 1
Texas Instruments' Planned Facilities—1991-1993
Products
Location
Wafer Size
Year
Hiji, Japan
4Mb DRAMs, 1Mb SRAMs
8-inch
1991
Hsinchu, Taiwan'
4Mb DRAMs
6-inch
1991
Tsukuba, Japan
R&D facility
NA
1991
Avezzano, Italy
16Mb DRAMs
8-inch
1992
Hyogo, Japan
VLSI logic, ASICs
8-inch
1992
Hiji, Japan
16Mb DRAMs, 4Mb SRAMs
8-inch
1993
NA = Not available
1
Joint ventuie with Acer
2
Jaint ventuie with Kobe Steel
Source: Texas Instnuncots
FIGURE 2
Texas Instruments—Existing Capacity by
Line Geometry
(Percentage of Distribution)
Dallas, Texas
Recent events at TI's Dallas, Texas, facility
include 16Mb DRAM samples and development
and production in 1990 of a 0.8-micron
BiCMOS 100,000-gate array. Also, TI is shifting
some 1Mb DRAM capacity to digital signal
processing (DSP) devices and advanced logic.
Fresing, Germany
TI is in the process of upgrading its advanced
logic line to submicron capability.
Hsinchu, Taiwan
TI expects this joint-venture facility with Acer to
produce first silicon of 4Mb DRAMs in the third
quarter of 1991.
Hyogo, Ji^an
Note: Capacity is measured In square Inches
of silicon.
Source: Dataquest (February 1991)
• Avezzano, Italy
TI started pilot production of 4Mb DRAMs at its
Avezzano location and plans to produce the bulk
of its 4Mb DRAMs for the European market at
Avezzano for the next few years. Initial investment cost of the fab is about $250 million. Over
several years, $1.2 billion will be invested at
Avezzano. This investment wiU include an additional fab and R&D center. The Italian govemment will contribute about $670 miUion to the
total investment.
000952S
Official ground breaking for this joint-venture
facUity with Kobe Steel occurred in February
1991, and first siUcon is expected late in 1992.
Miho, Japan
TI wUl start production of the 0.8-micron,
100,000-gate bipolar CMOS gate array developed at the Dallas faciUty. This fab already
produces 1Mb and 4Mb DRAMs.
George Burns
01991 Dataquest Incorporated Febniary-Reproduction Prohibited
SEMMS Newsletters 1991 Company Information
t^^^;i^
Dataquest
m m acompnyof
W O TticDun&BiadsticctCorpoiatKMi
mMM
Research Newsletter
A QUICK LOOK AT
NEC
This newsletter provides a quick look at
NEC's semiconductor capital spending, R&D
spending, capacity by line geometry, planned facilities, and recent company highlights as related to
semiconductor manufacturing. This newsletter is
part of the "Quick Look" series of newsletters
from Dataquest's Semiconductor Equipment,
Manufacturing, and Materials service (SEMMS).
R&D spending expressed as a percentage of semiconductor revenue by year from 1985 to 1989.
MANUFACTURING FACILITIES
Planned additions to NEC's facilities from
1991 to 1993 are shown in Table 1. Figure 2
illustrates the percentage distribution of NEC's
existing worldwide semiconductor capacity by line
geometries.
SEMICONDUCTOR CAPITAL SPENDING
AND R&D SPENDING
Figure 1 graphically illustrates the dollar
amount of NEC's semiconductor capital spending
by year from 1977 to 1990 and semiconductor
COMPANY HIGHLIGHTS
The following discussion highlights significant events for NEC.
FIGURE 1
NEC—Capital Spending and R&D
Spending as a Percentage of Semiconductor Revenue
Millions of Dollars
Percentage of Revenue
700
630
560
r35
Capital Spending
R&D
490
420
350280
fS
210140-
V^
\
70
^
0
1977
1978 1979 1980
I
1981
$ M
1982 1983
N
f^
1984 1985 1986
1987
1988 1969 1990
Source: Dataquest (March 1991)
01991 Dataquest Incorporated Maich-Reproduction Pndiibited
SEMMS Newsletters Conq>aiiy Infojmation
0009671
The conlent of this report represents our iruerpretatUm and analysis qfinprmation generally available to the public or released by responsible individitals in the subject companies, but
is not guaranteed as to accuracy or completeness. It does not contain material provided to us in cortfidence by our clients, bidividuai companies reported on and analyzed by Dataquest
nwy be clients of this and/or other Dataquest services. This information is notfttmlshed in connection with a sale or o^r to sell securities or in connection with the solicitation of an
qffirto buy securities. This firm and its parent artd/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and truly sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT NEC
TABLE 1
NEC's Planned Facilities—1991 to 1993
Location
Wafer Size
Products
Year of
Production
Livingston, Scotland
4Mb DRAMs, 1Mb SRAMS,
MPUs
6-inch
1991
Roseville, California
4Mb DRAMs
6-inch
1991
Higashi-Hiroshima, Japan
4Mb DRAMs, 1Mb SRAMs,
EPROMS
8-inch
1991
Kiunamoto-Shi, Japan
16Mb DRAMs, 4Mb SRAMs,
MPUs, gate arrays
8-inch
1992
Yamaguchi, Japan
16Mb DRAMs
NA
1993
Higashi-Hiroshima, Japan
4Mb DRAMs, 16Mb DRAMs
8-inch
1993 or 1994
Beijing, China (Joint venture
with a Chinese company)
64K DRAM-level LSIs
NA
NA
NA = Not available
Source: Dataquest (March 1991)
FIGURE 2
NEC—Existing Capacity by Line Geometry
(Percentage of Distribution)
in additional equipment, which wiU double its
wafer starts to 18,000 6-inch wafers per month.
This facility is producing 4Mb DRAMs and
MCUs. NEC Semiconductor (U.K.) Ltd. will
also start production in 1991 at a new plant in
Livingston, Scotland, to produce 4Mb DRAMs
for the European market.
NEC's M-UNE fab at Roseville, California, wiU
begin 4Mb DRAM production this year using a
tool set similar to the one at the NEC Hiroshima
fab.
The NEC Hiroshima fab was completed at a cost
of $425 million. This facility will be used to
manufacture 4Mb DRAMs, 1Mb SRAMs, and
EPROMs. This facility will eventually manufacture 16Mb DRAMs using a 0.6-micron CMOS
technology.
Note: Capacity is measured
in square inclies of siiicon.
Source: Dataquest (March 1991)
• NEC and AT&T Microelectronics have signed a
five-year technology exchange agreement. The
agreement calls for NEC to provide production
technology for all of its CMOS gate array
products to AT&T. In return, AT&T will supply
CAD system technology to NEC.
• NEC is currently processing 9,000 6-inch wafers
per month at its plant in the United Kingdom.
NEC is in the process of installing $49 million
0009671
NEC will sell semiconductors to countiies in
eastern Europe for use in consumer goods and
automobiles from its Berlin-based subsidiary
NEC Electironic Gramany GmbH.
>\^thin several years, NEC plans to establish a
semiconductor plant in Beijing, China, to produce 64K DRAM-level ICs. The fab is a joint
venture between NEC and Shoudu Iron and
Steel Company.
Kunio Achiwa, Tokyo
Jeff Seerley, San Jose
01991 Dataquest Incoipoiated Maicb-Reproduction Prohibited
SEMMS Newsletters Company Lifomution
f^M^r^
hM^t^-'-'
Dataquest
iwm
The Dun & Bradsticct Corporation
^r-^-.'v^.-^'^
^0m
I^JJ^
Research Newsletter
A QUICK LOOK AT
MOTOROLA
This newsletter provides a quick look at
Motorola's semiconductor capital spending, R&D
spending, capacity by line geometry, planned
facilities, and recent company highlights as related
to semiconductor manufacturing. This newsletter is part of the "Quick Look" series of newsletters from Dataquest's Semiconductor Equipment, Manufacturing, and Materials Service
(SEMMS).
semiconductor capital spending and semiconductor
R&D spending by year from 1975 through 1990.
SEMICONDUCTOR CAPITAL SPENDING
AND R&D SPENDING
COMPANY HIGHLIGHTS
MANUFACTURING FACILITIES
Additions to Motorola's manufacturing facilities in 1990 and planned additions beyond 1990 are
shown in Table 1. Figure 2 illustrates the percentage distribution of Motorola's existing worldwide
semiconductor capacity by line geometries.
The following discussion highlights significant events for Motorola.
Figure 1 graphically illustrates Motorola's
FIGURE 1
Motorola—Capital Spending and R&D Spending by Year
Millions of Dollars
700-,
630
Capital Spending
Research and Development
560-1
490
420
350
280-1
210
,*.-.-^*'K *•«.-».-*'^*
„«.-^'»'"
140
_
^
^
• « . - « .
^
'
70
1975
1977
1979
1981
1983
1985
1987
1989
Source: Dataquest (November 1990)
O1990 Dataquest Incoipoiated No'vember-4lepioductian Prohibited
SEMMS Newsletters 1990 Equipment Business Index
0008925
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^idence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This ir^rmation is rtot furnished in connection with a sale or of^r to sell securities or in connection with the solicitation of an
q^r to buy securities. This firm and its parent and/or their officers, stodtholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or bity such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT MOTOROLA
TABLE 1
Motorola's Planned Facilities—1990-1992
Location
East Kilbride, Scotland'
Seremban, Malaysia
Hcmg Kong (Silicon Haibor)
Oak Hill, Texas
Sendai, J^an^
Sendai, Japan
Chandler, Arizona
Chandler, Arizona
Products
DRAMs, SRAMs, MFUs
Discretes, power
Design, assembly, and test
Memoiy, MFUs
4Mb DRAMs
Assembly and test
ECL and BiCMOS, gate aiiays
R&D line
Wafer Size
6-inch
6-inch
NA
8-inch
6-inch
NA
6-inch
6-inch
Year
1990
1990
1990
1991
1991
1991
1992
1992
^ A = Not available
jCapadty aqmiaan
Motofola-Toshiba joiiit ventme
Souioe: Moiorola
FlGUKE 2
Motorola—^Existing Capacity by Line Geometry
(Percentage of Distribution)
Chandler, Arizona (R&D line and Bipolar 6)
Bq>olar 6, an ASIC fab designed for small
lots and quick turns, will be built adjacent to
Motorola's new ASIC R&D line. Motorola
will be able to take devices from R&D to
pilot to manufacturing, aU within the same
facility.
Silicon Harbor
The Silicon Harbor facility, located in Hong
Kong, is a 326,000-square-foot facility that will
house Motorola's Asia/Pacific headquarters, an
advanced ASIC design center, and an assembly/
test center.
Plans for China facility
Note: Capacity Is measured by square Inches
of silicon.
Source: Dataquest (November 1990)
• Oak Hill, Texas (MOS 11)
This new fab is Motorola's first 8-iiich wafer
facUity. It is scheduled to begin production of
advanced memory and microprocessor devices
in the latter part of 1991. Initial cost of this
facility will be $250 million. Recent equqnnent
orders for MOS 11 include ^>plied Materials
CVD, etch, and inq>lanter systems; Canon steppers; Lam Research plasma etchers; Silicon
Valley Group track equipment and vertical
reactors; and Varian implanters.
00OS92S
Motorola is reported to be still interested in
building an assemblyAest facility in China—
perhi^ by 1992 in I^anjin.
Toshiba joint venture
Motorola's joint venture with Toshiba, Tohoku
Semiconductor, is scheduled to begin production
of 4Mb DRAMs at its Sendai plant in Japsa in
mid-1991. It is rumored that Motorola and
Toshiba also will start construction of a jointventure 4Mb DRAM plant in Europe in 1991.
George Bums
O1990 Dataquett bcoiponted Novembec-ReproductioD ProbiUted
SEMMS Newtletten 1990 EquipoieiU Buajneti Index
mm
hmA
mm^
'""•VWf'
Dataquest
aa»ii|»nYpf
The Dun Kwadsticct Corporation
Research Newsletter
A QUICK WOK AT
KLA INSTRUMENTS
This newsletter provides a quick look at the
financial performance and recent company activities of KLA Instruments, a major siqrplier of semiconductor wafer fabrication equipment. This newsletter is one of the "Quick Look" series of
newsletters from Dataquest's Semiconductor
Equipment, Manufacturing, and Materials Service
(SEMMS).
FINANCIAL REVIEW
Selected financial data for KLA are presoited
here with no comments or analysis. Figure 1
presents KLA's quarterly sales and net income
through the quarter ending September 30, 1990.
Table 1 provides a summary of fiscal year (ending
6/30) sales, net income, R&D expenditures as a
percentage of sales, and return on assets (ROA) for
KLA.
FIGURE 1
KLA Instruments—Sales and Net Income by Quarter
(Millions of Dollars)
Sales (Millions of Dollars)
Net Income (Millions of Dollars)
50
^ ^ ^ ^ Sales
*'*.'*.-•». i5> Net Income
40-
- 4
30-
- 3
yf'"'^
•^•-^
20-
- 2
J.'*
10-
4
\
- 1
ft
1962
TOM:
19B6
1986
1990
Source: KLA Instruments* Quarterly Reports, Dataquest (December 1990)
O1990 Dataquect Incoiporated December-JlepTDductioii Prohibited
SEMMS Newiletten 1990 Equipment Business bidex
0009137
The content of Ms report represents our interpretation and analysis cf information generally available to the public or released by respof\sible individuab in the subjea companies, but
is not guaranteed as to accuracy or completeness. It does not coraain material provided to us in confidence by <mr clients. Individual companies reportai on and analyzed by Dataquest
may be cliertis oftfus and/or other EkUaquest services. This ir^rmation is not furnished in cormection with a sale or offer to sell securities or in connection with the solicitation cfan
c^r to buy securities. This firm and its parent and/or their officers, stodcholders, or merrdjers of their families may, from time to time, have a long or short position in the securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT KLA INSTRUMENTS
TABLE 1
KLA Instruments—Fiscal Year Sales, Net Income, R&D as a Percentage of Sales, and ROA
(Thousands of Dollars)
Fiscal Year
(Ends 6/30)
Sales
Net Income
Ratio of Net
Income to Sales (%)
Ratio of
R&D to Sales (%)
ROA (%)
1982
1983
1984
1985
1986
1987
1988
1989
1990
16,162
23,396
42,873
62,878
82,526
88,194
112,851
165,459
167,916
1,510
2,928
5,664
8,802
9,854
7,489
8,827
11,678
9.380
9.3
12.5
13.2
14.0
11.9
8.5
7.8
7.1
5.6
15.0
13.1
14.3
17.1
12.3
9.8
12.2
15.2
18.4
8.4
8.4
8.7
10.9
10.0
6.5
6.6
7.3
5.2
Source: KLA Instiunieals' Annual Repoiu, Dataquest (December 1990)
COMPANY HIGHLIGHTS
The following discussion highlights significant events for KLA.
• KLA Instruments' business activities
Founded in 1975, KLA is a major manufacturer
of automated optical inspection equipment
primarily used by the semiconductor industry
and manufacturers of printed circuit boards. The
company's equipment product offerings are
organized into four divisions: the Reticle and
Photomask Inspection Division (RAPID), the
Wafer Inspection System Division (WISARD),
the Automated Ibst Systems Division (ATS),
and the KLA Scanning, Inspection, and Classification Division (KLASIC). KLA's productdevelopment activities are carried out in the
United States (California), Gemiany, Israel, and
Japan. Manufacturing operations are located in
California, Germany, and IsraeL
In the late 1970s, the company formed a relationship in J^an with Tokyo Electron Limited
(TEL) to sell and service its semiconductor
products. This early relationship in Japan proved
to be a significant conqxinent in building the
conqtany's international presence. Li 1980, international sales totaled 18 percent of total revenue,
compared with 7 percent in 1979. By 1990,
international sales accounted for 53 percent,
more than one-half of the coiiq>any's revenue.
The company's semiconductor equipment
products perfonn optical defect detection on
photomasks, reticles, and wafers; electrical
defect detection on wafers and individual
integrated dicuits, linewidth and overh^ measurements on wafers; and electrical probing on
0009137
finished wafers. KLA's high-speed imageprocessing technology was the first to provide
the semiconductor industry with automated
defect detection of photomask, reticle, and wafer
patterns. This achievement was significant
because it allowed the subjective judgement of
the operator to be eliminatedfi-omthe inspection
process.
Recent significant announcements
In October 1990, KLA Instruments' WISARD
division introduced its second-generation wafer
inspection system, the KLA 2110. The 2110 has
significantly increased througt^ut and in^oved
defect sensitivity compared with its predecessors
in the IOJIX family of automated defect inspection systems. Hie 2110 has been designed
specifically for defect inspection of repeating
pattern arrays such as DRAMs, SRAMs,
EPROMs, EEPROMs, and gate arrays. The
conq>any views the 2110 as complementary with
its other 20xx product offerings, which have
full-pattem ci^ability for all device structures
such as logic.
Tn September 1990, KLA and Nippon Mining of
Japan announced that they had signed an agreemeiU to form a new joint-venture conr^pany,
KLA AcTotec Conqiany, Ltd., for development
of flat-panel display inspection equipment The
new conqiany will be based in Japan and staffed
by enoployees of KLA and N^jpon Mining. Tlie
automated defect-detection technology of KLA
is weU suited for hquid-ctystal and flat-panel
display manufacturing, areas where it is critical
to find and repair defects before [socessing is
complete.
Joe Grenier
Peggy Marie Wood
O1990 Dataquest Incorporated December-Reproductiaii Prohibited
SEMMS Newiletters 1990 Equipment Bunneu lodex
DataQuest
1^
a company of
Tlie Dun & Bradstreet Corpoiation
Research Newsletter
A QUICK WOK AT
INTEL
This newsletter provides a quick look at
Intel's semiconductor capital spending, R&D
spending, capacity by line geometry, planned facilities, and recent company highlights as related to
semiconductor nunufacturing. This newsletter is
part of the "Quick Look" series of newsletters
from Dataquest's Semiconductor Equipment,
Manufacturing, and Materials Service (SEMMS).
SEMICONDUCTOR CAPITAL SPENDING
AND R&D SPENDING
Intel's semiconductor capital spending and
semiconductor R&D spending by year is shown in
Figure 1.
MANUFACTURING FACILITIES
A percentage distribution of Intel's existing
worldwide semiconductor capacity by line
geometries is shown in Figure 2. Additions to
FIGURE 1
Intel—Semiconductor Capital Spending and R&D Spending by Year
(Millions of Dollars)
Millions of Dollars
700
Capital Spending
Research and Development
1975
1977
1979
1981
1983
1985
1987
1989
Source: Dataquest (December 1990)
O1990 DauqueM Incorporated Deccmber-Reproductian Prohibited
SEMMS Newdetters 1990 Equipment Bunneu Index
0009102
The coruera of this report represetus our interpretation and analysis qfit^rmation 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. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This inpjrmation is notfitmished in connection with a sale or offer to sell securities or in connection with the solicitation of an
q^r to buy securities This firm and its parent and/or their officers, stockholders, or members of theirfamiliesmay, fiom time to time, have a long or short position in the securities
mentioned and tnay sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT INTEL
Intel's facilities in 1990 and planned additions
beyond 1990 are shown in Table 1.
COMPANY HIGHLIGHTS
The following discussion highlights significant events for Intel.
• Santa Clara, California (D2)
Intel's newest developmental fab, D2, is a memory and embedded control device development
fab with 25,000 square feet of clean room. It
was designed to develop the manufacturing
p-ocesses to support three different technologies
simultaneously: current generation down to
0.8 micron, 0.5 to 0.6 micron, and 0.2 to
0.3 micron. D2 was thefirstdevelopment fab in
the United States to use Dr. Ohmi's submicron
clean room concepts.
• Rio Rancho, New Mexico (Fab 9.1 and Fab 9.2)
Intel added significant microprocessor and
advanced logic c£5)acity in 1990 to Fab 9.1. It
also began equipping its submicron Fab 92 to
produce microprocessors and other advanced
logic devices. Intel e:q)ects to begin volume
production in Fab 9.2 by the second quarter of
1991.
Chandlrar, Arizona (assembly and test facility)
Intel con^leted an assembly and test facility in
Chandler, in 1990. It will be used to assemble
and test R&D and prototype devices and leading-edge production devices.
Livermore, California (Fab 3)
Intel stated that its Livermore fab (Fab 3), which
was going to be closed this year, will stay open
an extra year to meet 80383SX demand.
Leixlip, Ireland
Construction of a systems manufacturing plant
began at this site in 1990. Intel also plans to start
construction of a wafer fab in 1991, which
Dataquest ei^)ects to be an 8-inch facility. Chip
production will begin in 1993.
George Burns
FIGURE 2
Intel—^Existing Capacity by Line (xeometry
(Percentage of Distribution)
Note: Capacity is measured by square Inches of silicon.
Source: Dataquest (December 1990)
TABLE 1
Intel's Planned Facilities—1990-1993
Location
Santa Qaia, California (D2)
Rio Rancho, New Mexico (9.2)
Leixlip, Ireland
Products
Nonvolatile memoty technology development
MPUs, EPROMs
80486/80586 MPUs
Wafer Size
6-inch
6-inch
8-inch
Year
1990
1991
1993
Souice: Intel
0009102
01990 Dataqiwit Incoipanted December-ltepraductiaa Prohibited
SEMMS Newiletlen 1990 Equipment Butinen Index
p i
^^Jr^^iiV-J&ri
& V ^ J ^ ^ ^
Dataqyest
V r a accHiipanyof
l i o TheDunK^iadstreetCorporatK
Research Newsletter
A QUICK LOOK AT
ADVANCED MICRO DEVICES
This newsletter provides a quick look at
Advanced Micro Devices' (AMD's) semiconductor
capital spending, R&D spending, capacity by line
geometry, planned facilities, and recent company
highlights as related to semiconductor manufacturing. This newsletter is part of the "Quick Lxjok"
series of newsletters from Dataquest's Semiconductor Equipment, Manufacturing, and Materials
Service (SEMMS).
and semiconductor R&D spending e?q>ressed as a
percentage of semiconductor revenue by year from
1975 to 1990.
MANUFACTURING FACILITIES
Additions to AMD's facilities in 1990 are
shown in Table 1. Figure 2 illustrates the percentage distribution of AMD's existing worldwide
semiconductor capacity by line geometries.
SEMICONDUCTOR CAPITAL SPENDING
AND R&D SPENDING
COMPANY HIGHLIGHTS
Figure 1 graphically illustrates the dollar
amount of AMD's semiconductor capital spending
The following discussion highlights significant events for AMD.
FIGURE 1
AMD—Capital Spending and R&D Spending by Year
Millions of Dollars
320
Percentage of Revenue
Capita] Spending
Research and Development
280
240
200-J
160
1208040
0
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1567 1988 1989 1990
Source: Dataquest (January 1991)
©1991 Dataquest Incorporated January-Reproduction Prohibited
SEMMS Newsletters Cosopany Information
0009260
The coroent of this report represents our iruerpretation and analysis qfin^rmalion 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 cortfidence by our clients Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This irifijrmation is not furnished in connection with a sale or q^ to sell securities or in cormection with the solicitation of an
qffir to buy securities This firm and its parent and/or their officers, stockholders, or members of theirfimaliesmay, from time to time, have a long or short position in the securities
meruioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
A QUICK LOOK AT ADVANCED MICRO DEVICES
TABLE 1
Additions to Advanced Micro Device's Facilities—1990
Products
Wafer Size
Year
Sunnjrvale, California (SDC)
Technology development fab
6-inch
1990
Austin, Texas*
EPROM, logic, MPUs, PLDs,
and SRAMs
5-inch
1990
Location
*CM0S cq>acity expansion
Souzce: Advanced Micio Devices
FIGURE 2
AMD—^Existing Capacity by Line Geometry
(Percentage of Distribution)
Austin, Texas (Fab 10)
Fab 10, which produces 80286 MPUs and PLDs,
originally was an NMOS fab. AMD has
expanded capacity by 40 percent to produce
CMOS PLDs.
San Antonio, Texas (bipolar facilities)
AMD sold its bipolar facilities in San Antonio to
Sony for $55 million.
Bangkok, Thailand (assembly and test facility)
AMD began ramping up its new state-of-the-art
157,000-square-foot automated assembly and
test facility.
Capital spending in 1991
Note: Capacity is measured in square inches
of silicon.
Source: Dataquest (January 1991)
AMD expects its capital spending in 1991 to be
substantially less than in 1990—^£rom approximately $300 million in 1990 to $130 million
iQ 1991.
George Bums
• Sunnyvale, CaMomia (SDC)
AMD began processing product wafers at its
new Submicron Development Center (SDC) in
September 1990. The SDC is one of the most
advanced R«&;D lines in the world. The SDC is a
paperless fab with a wafer-start capacity of
3,000 wafers per week. Of these, 2,400 will be
product wafers and 600 will be R&D wafers;
therefore, the SDC is both an R&D fadlity and a
production Una. The facility was designed to be
able to process wafers down to 0.25 micron by
the end of the decade. Air cleanliness is
Class 0.1. AMD's investment in the SDC was
nearly $200 million.
0009260
ID1991 Dataquest Incorporated Januaiy-Reproduction Prohibited
SEMMS Newsletters Company Information
Dataoyest
Dun STBradstreet Cwporation
Research Newsletter
A MACRO VIEW OF MICROENVIRONMENTS
DEFINITION OF A MICROENVIRONMENT
SUMMARY
As semiconductor technology advances, the
cost of building a new fab also increases. Figure 1
shows the cost trend for building and equq>ping a
state-of-the-art fab. In 1970, fee initial cost of
building and equipping a high-volume, state-of-theart fab was $30 million. By 1990, diis cost had
risen to $300 million. As line geometries shrink
and process steps increase, the cost to build and
equip a fab will continue to increase. The clean
room cost represents a significant portion of the
total cost to construct a fab. Alternative clean room
procedures are being developed to lower clean
room costs and increase contamination control.
One such alternative is microenvironments.
A microenviromnait is defined as a relatively
small volume of controlled space surrounding and
isolating wafers firom contamination sources both
in the process tool and the room in general Different processes and material-handling mechanisms
require different microenvironm^t solutions.
MICROENVIRONMENT SUPPLIERS
Three Nordi American conq)anies are aggressively perusing microenvironment market share.
They are Asyst Technologies Inc., Briner/Yeaman
Engineering Inc., and Intelligent Enclosures
Corporation.
Figure 1
Building and Equipment Costs for a High-Volume Fab Line
Millions Of Dollars
320
280
KS\ Facility Costs
Equipment Costs
Total Cost
240H
200
160
120
SO
40
O-l
1970
1976
1980
1883
1987
1990
Source: Dataquest (August 1991)
01991 DaUqiieit Incoiponted Augu(t.4lepiodDctiaa Prahitnted
SEMMS Newiletten 1991
0011179
The coroent cfthis report represaits our interpretation and analysis cfi^rmation generally available to the public or released by responsible individuals in the subject aympeaues, but
is not guaranteed as to accuracy or amq?leieness. It does not amtain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This ir^^matiim is notfiimished in connection with a sale or offer to sell securities or in connection with the solicitation cfan
offer to buy securities. This firm and its parent and/or ^ir officers, stockholders, or members of theirfamiliesmay, from time to time, have a long or short position in the securities
meruioned and may sell or buy sudi seairities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Tfelex 171973 / Fax (408) 437-0292
A MACRO VIEW OF MICROENVIRONMENTS
Asyst Technologies
Asyst Tschnologies is located in MiJ^itas,
California. The coitq)any manufactures and markets
a product line designed around a standard mechanical interface (SMIF). The SMIF product line
includes a sealed wafer carrier (SMIF pods),
robotic transfer mechanisms (SMIF arms), and custom enclosures. A list of semiconductor manufacturers with SMIF installations follows:
• AMI
• Cypress Minnesota
minienvironments. "Minienvironment" is its tram
for microenvironment. The minienvironment
enclosures are modular and reusable, and they can
accommodate tool clustering as well as a wide
range of material-handling methods including manual load, SMIF pod/arm, or robotic automation.
Intelligent Enclosures is a second-source supplier
of minienvironments for IBM. It is currently working with a conq)any that has had a 10 percent
increase in product jdeld with the implementation
of its minienvironments. Bookings for projects in
1990 totaled more than $1 million, and the company expects business to quadruple in 1991.
• EM Microelectronics
• Hanis
• Hewlett-Packard
• IBM
TYPES OF MICROENVIRONMENTS
The following paragrq>hs discuss four different types of microenvironments.
• Intel
• LSI Logic
• NCR
• Phitq)s
• Siemens
• TSMC
• UMC
It is interesting to note that some level of
SMIF has been installed in every region of the
world—^North America, Japan, Europe, and Rest of
World.
Briner/Yeaman Engineering
Briner/Yeaman Engineering is located in
Santa Clara, California. It markets manual, SMIF,
and automation-compatible microenvironments.
Microenvtronments produced by Briner/Yeaman
Engineering achieve 10 to 100 times better contamination control than that of Class 1. In addition
to microenvironment design and manufacturing,
Briner/Yeaman performs wafer fab layout, design,
process equqnnent selection, maniifacturing simulation, and faciUty fitup/hookup design. Semiconductor manufacturers with microenvironments
installed by Briner/Yeaman include Fortrend, LSI
Logic, Motorola, NEC, and Pace.
Inteliigent Enclosures
Intelligent Enclosures is located in Norcross,
Georgia. The company manufactures and markets
0011179
Tool-Integrated Microenvironment
This type of microenvironmoit is the most
efOcient because it represents the minimum in
enclosure volume. It is also the most difficult to
design because of the needs of integrating the
normal equipment access requirements necessary
for normal operation and maintenance. Some tools
that have traditionally used integrated microenvironments are reduction wafer steppers, which
need precise control of temperature, humidity, and
particulates. Equqnnent conq>anies that are proactive with respect to providing microenvironment
control with flietr tool will opt for the toolintegrated solution.
Tool-Enclosing Microenvironment
The entke tool, with its own contamination
generating portions, is enclosed. The volume of
enclosed space is larger than fbe tool-integrated
enclosures, but some of the design and integration
difficulties are avoided.
Cassette-Enclosing Microenvironment
Run boxes are microenvironments for cassette
enclosing that have been used for many years and
are supplied by Fluoroware. This style of microenvironment is conq)atible with botti manual and
automated methods of microenvironment access.
More recently, the SMIF pod was developed as a
SMIF-style cassette-enclosed microenvironmenL
01991 Oaiaquen iDCOiponted Aiiguit-Reprodiictiai ProUbited
SEMMS Newdetten 1991
i
A MACRO VIEW OF MICROENVIRONMENTS
Robotic-Enclosing ii/licroenvironment
ADVANTAGES OF MICROENVIRONIMENTS
Robots are used in semiconductor manufacturing to move product ia a repeatable, controlled
manner. Microenvironments are used with robotics
to isolate the product from contamination sources
during transport and to keep the automationgenerated particles away fixnn the product. Automated diffusion furnace loading, inq>lanter end stations, automated stockers, and enclosed robots on
rails are exanq)les of this type of microenviromnent
in which manual operations have been replaced
with automated mechanisms for transporting wafo^
and cassettes of wafers.
One of the major advantages of a microenviromnent is that, fundamentally, a small volume is
easier to control than a large volume. More precise
control and better economies are both possible. In
the traditional clean room environment, a zone of
control is established for a relatively large volume,
enclosing multiple process tools along with
manufacturing personnel.
Some of the specific relative advantages of
microenviroimients over the traditional clean room
are better contamination control, less initial and
operating costs, reduced clean room protocol,
elimination of cross-contamination, allowance for
different control set points, facilities flexibility, and
ease of major upgrades.
ACCESS BETWEEN
MICROENVIRONIUIENTS
As microenvironments become more prevalent in the semiconductor industiy, one of the key
issues will be how material is moved between these
microenvironments while minimizing contamination. The material can be accessed manually or
moved with automation.
Manual access is the most cost effective, and
it is also the easiest to implement. The microenvironments are engineered to allow a clean load
area for opening the cassette carrier and for cassette
placement on the tool. Maintoiance access is easier, and no tool modifications are required. Btiner/
Yeaman Engineering has developed a manual
access port that interfaces easily to a tool-integrated
microenviromnent and allows for the manual,
contamination-free transfer of cassettes from a
Fluoroware-type run box to the equipment
Several con^anies provide automation compatible with microenvironments. These compaaiss
include Accufab, Asyst Technologies, Daifiiku,
Precision Robots Inc., Proconics, and Programmation. The Asyst-SMIF system uses clean-isolation
technology to protect the integrity of wafers during
processing, storage, and transportation within the
facility. Wafers are isolated in sealed, ultraclean
cassette containers, and specialized robotic arms
transfer the sealed wafers into and out of
enclosures.
01991 Dataqueit Incocponted Aiignit-Repiodiietioa Prohibiled
SEMMS Newilstten 1991
DATAQUEST
PERSPECTIVE
Dataqtiest believes that demand for microenvironments will increase, especially with die growing number of 5- to 10-year-old fabs that will
require upgrades to manufacture semiconductors
with competitive yields. Today, semiconductor
manufacturers can easily and ine:q>ensively add
this new level of contamination control. For the
average cost of about $12,000 per tool, a custom
manual microenviromnent can be engineered, bult,
and installed with minimal interruptions to production.
Considering that &e only other options for
increased contamination control are to construct a
cleaner facility with increased contamination control or shut down the existing facility for renovation, microenvironments are definitely a viable
alternative to achieving increased contamination
control (This article was written by Jeff Seerley in
conjunction widi Don Briner of Briner/Yeaman
Engineering Inc. For furdier information, please
contact Jeff Seerley.)
Jeff Seerley
Don Briner
0011179
Dataquest
^ S ^ BB^i^^S^,
TT^tCQqMHTtnfl
^"/^^
Research Newsletter
ALLIANCES: LARGE COMPANY RIVALRIES SPAWN START-UPS
Small start-up semiconductor companies, conceived in response to exogenous technological
changes, are bom of the conq)etitive rivalry among
large inciunbent companies. This view is contrary
to the popularly held belief that large companies
stifle the emergence of start-ups. However, according to Dr. Bruce Kogut, professor in the Department of Management of Wharton School, University of Pennsylvania, it is exactly the acceptance of
strategic technological alliances by large companies
that fosters the emergence of start-ups.
Dr. Kogut further believes that major companies strategically encourage start-up companies as a
way to promote proprietary technologies into
industry standards. In this manner, they may even
consciously extend their competition with their
major rivals.
Under Dataquest's sponsorship. Dr. Kogut and
doctoral candidate Dong-Jae Kim, also of the
Wharton School, have been analyzing data that
have been collected from as long ago as 29 years
on semiconductor company alliances. Between
1961 and 1989, there were about 1,975 alliances
involving semiconductor companies, with 1,765
occurring after 1979 (see Figure 1). About
63 percent involved U.S. companies. Data for the
study were compiled from Dataquest databases and
unpubUshed studies by the Electronic Industries
Association of Js^an (EIAJ, 1987) and New York
University (NYU, 1986).
Dr. Kogut initially noted that the relationships
implied some sort of network structure in the
global semiconductor industry. But upon further
examination he found that the companies tended to
pursue strategic, hence long-term, relations. This
tendency made the structure of the industry look
more like an interrelated web. "A typical structure
consists of one or more central players and
afGliated sateUite partners," he said.
He found that the central roles were most
commonly played by the large, incumbent companies, while satellite roles were taken by the small,
and often new, companies. Guided by this discovery, I>r. Kogut began exploring the role of large
incumbent semiconductor companies in regard to
the entry process of start-ups and the relationship
between large and small semiconductor companies.
The traditional belief about the relationship
among large and small companies competing in the
same industry is characterized as antithetical. Many
analysts beUeve that competition for market share
between companies inhibits start-up activities.
The assumption of implicit hostility between
large and small companies is reflected in policy
debates on the very factors that sustain the health
of the American economy. As per their debate in
the Harvard Business Review, George Gilder, who
supports small companies, and Charles Ferguson,
FIGURE I
Alliances by Region (1961-1989)
/
/\.
\
Japan
22%
'Europe
9%
\
JV Asia/
\
\
\
\
PaclflcROW
6%
United States
63%
Source: EIAJ, NYU. Dataquest (February 1991)
01991 Dataquest Incorporated February—Reproduction Prohibited
SO Newsletters 1991-03
0009451
The content of this report represents our interpretation and analysis cf 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. Individual c<mipanies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This information is not furnished in connection with a sale or q^r to sell securities or in connection with the solicitation of an
c^r to buy securities This firm and its parent and/or their officers, stodiholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
ALLIANCES: LARGE COMPANY RIVALRIES SPAWN START-UPS
who believes that large organizations are needed to
sustain heavy capital and R&D requirements, represent the typical positions.
Mr. Gilder affirms that die American strength
comes not from the law of complexity, but from the
law of microcosm; the case in point being the
down-scaling and technological development in the
computing industry. He believes that the
entrepreneur remains die driving force of economic
growth in all vibrant economies, especially the
U.S. economy.
In counterpoint to Mr. Gilder, Mr. Ferguson
argues that high-technology industries require
increasingly capital-intensive cost structures, dominated by R&D, computer networks, highly flexible
production systems, and global marketing and customer siqjport organizations. But based on what
I>r. Kogut sees in the strategic alliance data, he
says the argument between large and small is a
mistake. "This mistake is by no means minor in
light of the importance and magnitude of the decisions facing Eastern Europe and the Soviet Union
regarding the dissolution of their large enterprises
into thousands of small companies."
The notion that competing large companies
deter small companies from entering the industry
ignores evidence that the rivalry among large companies differs qualitatively from that between large
and small companies. Small and large companies
often do not compete directly, and competition
within an industry is partitioned by size categories.
However, the entrance of small companies does
dramatically impact maturing but innovative industries such as the semiconductor industry. Both
small and large companies appear to benefit substantially from each other. The innovative activities
of the new small companies pushes out the boundaries into new subfields and even to new major
branches.
Start-up companies seek protective alliances
with large incumbent companies in order to
enhance their chances for growth and survival. The
large companies seek to estabUsh central positions
in brokering and sharing knowledge in the development of new subfields.
The evolution of the semiconductor industry
has displayed a cyclical pattern of new entries to
the industry. The &st burst occurred in the 1950s
and early 1960s after the invention of the transistor.
The dominant vacuum tube companies failed to
extend their positions into this new technology, so
start-up companies such as Fairchild Semiconductor and Texas Instruments expanded rapidly
through the creation and adoption of new processes
and product innovations.
0009451
But internationally, the semiconductor industry displayed a different pattern during this period.
Large Japanese companies gained early access to
key patents and proprietary technologies needed to
bttild a domestic industry. Because technology went
to the large companies of Japan and there was a
tendency against technology sharing at that time,
few start-ups entered the market in Japan. In
Europe, most of the initial semiconductor production occurred by U.S. companies' subsidiaries.
Gradually, European companies entered the market,
although mostly in niche areas.
During the late 1960s and 1970s, start-up
entries were slow. However, the computer industry
in the United States grew, and the proliferation of
desktop computers came later, which fostered a
host of new specialty semiconductor opportunities.
These opportunities led to a rash of start-up companies in the late 1970s and 1980s (see Figure 2). To
a lesser extent, start-ups entered the semiconductor
market in Europe and Japan during the same period. Many of the start-ups of this era were design
centers without fabs.
As start-ups entered the subfields, they developed features that complemented existing products
manufactured by established companies. For example, some of die semiconductors could increase
microprocessor clock speeds. In order to take
advantage of this feature, start-ups need access to
proprietary information from the major companies.
In establishing technology agreements with the
major established companies, both the start-up
company and the major company benefited.
Start-up activity usually preceded the building of
alliances by two or three years (see Table 1).
Major companies such as Intel or Motorola
shared proprietary information with small companies because the technology of the small companies
enhanced the performance of their own products.
The large company could harness the higher niche
R&D productivity of the smallo- company, while
the small company enhanced its chances for growth
and survival. Both were betting on increased market share with the prospects of becoming an kidustty standard.
In strategic alliance development, it should be
noted that not only must microcomponents be compatible with the microprocessor they support, but
so must application-specitic integrated circuits
(ASICs) and various memory products such as
SRAMs and EPROMs. This compatibility requires
cooperation from the dominant company in order
to acquire the proprietary knowledge and legal
rights that result in a strategic alliance. The
&1991 Dataquest Incorporated Febiuaiy—Reproduction Prohibited
SG Newsletters 1991-03
ALLIANCES: LARGE COMPANY RIVALRIES SPAWN START-UPS
FIGURE 2
Start-Ups by Device T^pe
Number
16
•
•
X
•
1412-
ASIC
MBfTiory
Micro
Others
/
A
\
/
\
/ A \
10?
Br:
1
/ i t
6-:
^
I t
/
4i
2l
'y<^
019 79
1981
i^
^s.
1983
•
y'
1986
" \ * /
^
1967
1989
Source: EIAJ, N Y U , Dataquest (February 1991)
TABLE 1
Alliances Versus Start-Ups by Year
Allimces
'80
'81
'82
'83
'84
'85
'86
'87
'88
'89
21
35
52
88
153
232
300
420
325
139
47
27
18
9
6
9
3
Start-Up Entries
10
Source: EIAJ, NYU, Dataquest (Febniaiy 1991)
proliferation of these supporting products moves
the primary product ever closer to becoming an
industry standard, as have the Motorola 68XXX
microprocessors/microcontrollers and Intel's 80X86
microprocessors.
Given the depth of data, I>r. Kogut theorizes
that alliances, through cooperation and centrality,
have played a role in the evolution of the semiconductor industry as evidenced through the entry of
start-ups. He tracked the entry of start-up companies by product type: ASIC, microcomponents,
memory, analog, optoelectronics, discrete semiconductors, and gallium arsenide (GaAs).
By further tracking the pattem of alliances by
type of product and establishing a network of centrality, Dr. Kogut captures the extent to which
companies connect other companies to each other.
He notes that in the 1980s he saw a combination of
01991 Dataquest Incorporated February-Reproduction Prohibited
SG Newsletters 1991-03
new microprocessor-related technologies that
focused on customized integrated circuits (see
Figure 3). He believes that this trend spawned
an innovative wave that generated new opportunities for start-ups and established companies alike.
He suggests that the competitive uncertainty
among the established companies spilled over into
a race to encourage start-ups to innovate technologies compatible with their proprietary standards.
By establishing cooperative relationships, incumbent companies signaled their willingness to share
critical technologies or help new covapaia&s survive
by providing manufacturing capacity. This, he says,
induces the birth of new companies.
Dr. Kogut further suggests that the need for a
major compaay to seek a strategic alliance diminishes as its product gains acceptance in the market
and becomes an industry standard. For example,
0009451
ALLIANCES: LARGE COMPANY RIVALRIES SPAWN START-UPS
FiGUKE 3
Device-Specific Alliances by Year
Number
30-
•
25-
ASIC
•
Memory
X
A
Micro
others
20-
15-
10
5-
1
1979
M_^— — • ^ i S : ^
1981
•—n
1
1963
1987
1985
1969
Source: EIAJ, N Y U , Dataquest (February 1991)
Intel and Motorola both have market acceptance of
their microprocessors to the level of an industry
standard. He says that these con^anies' dominant
positions virtually have the industry locked into
their standards, which is why their rate of strategic
alliances is falling off. Barring new technological
developments, the rate of strategic alliances will
not increase (see Table 2).
National Semiconductor, by contrast,
introduced a 32-bit microprocessor in the same
time frame as did Motorola and Intel, but has not
yet received the same level of market acceptance.
National's rate of strategic alliance is still high, but
there is a decreasing incentive for small companies
to establish an alliance with National in this
product area.
TABLE 2
Selected Companies' Alliance Shares
(Percentage)*
Year
Intel
1979
Motorola
National
0
1.77
1.33
1980
0
1.87
1.49
1981
0.88
2.07
1.48
1982
1.81
2.04
1.13
1983
2.59
2.10
1.62
1984
3.46
2.06
2.81
1985
3.46
2.09
2.67
1986
2.87
2.01
2.57
DATAQUEST ANALYSIS
1987
2.72
1.77
2.97
Dataquest concurs with Dr. Kogut's analysis
that large semiconductor companies do stimtdate
new company birth and development through the
acceptance of strategic alliances. However, the relationship between small and large companies is
delicate. Cooperation with start-ups and small
1988
2.50
1.61
3.42
1989
2.53
1.65
3.22
00094S1
'Number of allimces of the company (cmmilativeVtoUl number of
alliances (cunulative) a 100.
Source: EIAJ. NYU, Dataquest (Febnuuy 1991)
®1991 Dalaquett Incorporated February-Reproduction Prohibited
SO Newalettcrs 1991-03
ALLIANCES: LARGE COMPANY RIVALRIES SPAWN START-UPS
companies as an extension of rivalry among large
companies can be stable only as long as the rivalry
persists under conditions of technological growdi
and innovation. As a dominant company emerges,
cooperation is no longer a potent strategy for either
the leader or the stragglers.
It also appears that the health of an industry
and economy is the result of a delicate and evolutionary balance between cooperation and competition, between innovation and diffusion, and
between small and large companies. Entry, growth,
and exit are elements of the revival and persistence
of iadustries and their subfields. These processes
also reflect die ecological balance achieved through
entrepreneurship of small companies and the
cumulative knowledge and assets of larger companies. The policy position, then, cannot be for the
promotion of any size of company but only for the
appropriate and dynamic mixture of both large and
small companies.
Marc Elliot
Dataquest oGfers consulting services to analyze strategic alliances or ptoq>ective alliances. Dataquest has
conqnled an extoisive wnldwide database of semiccnductor alliances and has stnictuied it to allow for
full analysis.
01991 Dataquest Incorporated February-Reproduction Prohibited
SG Newaletters 1991-03
0009451
mm
^^Mif-
Datapiuest
acAFtunyof
The Dun ii Bndnimcorporation
'iW^^W^M
m^
.•cW^i
Research Newsletter
U.S. COMPANIES TOP LIST FOR STRATEGIC ALLIANCES
INTRODUCTION
Strategic alliances among companies have
become commonplace in the semiconductor industry around the world. But the United States tops the
list for the number of strategic semiconductor alliances with 63.3 percent of the 1,875 strategic
alliances instituted worldwide since 1961 (see
Table 1). This should not be a surprise because the
majority of individual semiconductor companies
are in ttie United States. What is significant are the
types of agreements instituted in the different
regions.
at Whorton. The foUowiag seven types of alliances
were identified:
• Acquisition (includes merger) (AC(^
• Joint venture (JV)
• Equity investment (EQT)
• Licoising (LIC)
• Second sourcing (SCND)
• Cooperative agreement (COAG)
• Technology transfer (TECH)
The recorded agreements covo: 20 countries
and have been grouped into four regions: the
United States, Xs^an, Europe, and Asia/Pacific-Rest
of World (ROW). These data come fiwn Dataquest
and from unpublished studies conducted by the
Electronic Industries Association of Japan (EIAJ,
1987) and New Ywk University (NYU, 1986). The
study covoed the period from 1961 to 1989. In a
few cases, the agreements include more than two
conq>anies.
THE STRATEGIC ALLIANCE STUDY
Because alliances have a direct effect on the
direction of the industry, Dataquest sponsored a
study and analysis of the alliances by Whorton
School of Business, University of Pennsylvania.
Data on semiconductor con:q>any alliances, going
back 29 years, were conq}iled and analyzed by
Professor Bruce Kogut and Doctoral Candidate
Dong-Jae Kim of the Department of Management
TABLE 1
Total Number of Alliances
(1961-1989)*
United States
Japan
Europe
Asia/Pacific-ROW
Total
United States
Japan
Europe
Asia/PacificROW
1,538
506
242
61
31
214
61
49
25
120
31
25
13
2,378
506
214
120
2,378
840
349
189
3,756
Total*
840
349
189
'Double coumed: each paiticipmt in an alliance ii counted aa <
Somce: EIAJ, NYU, Dataquett (Pebniaiy 1991)
e i 9 9 1 Dauquett Incoiporated Febniaiy-Repioductian PrabiUted
00093S5
Semicaaductar Oroup Newdettera 1991-02
77K coramt of this report represertts our iraerpraation and armlysis of infinrmlion gemrally available to the public or released by responsible individuals in the subject companies bm
IS not guaranteed as to accuracy or completeness It does not contain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Datapiuest services. This injbrmatian is not furnished in connection with a sale or offer to sell securities or in connection with the solicitalion cfan
offer to buy securtbes Vus firm and its parent and/or their officers, stockholders, or members of their families may. from time to time, have a long or short position in the securities
menlioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
U.S. COMPANIES TOP LIST FOR STRATEGIC ALLIANCES
Since the early 1980s, the most accepted form
of alliance worldwide is the cooperative agreement
(53.3 percent) followed by technology transfers at
10.9 percent (see Table 2). Licensing accounted for
only 8.4 percent, second-sourcing for 8 percent,
acquisition for 7.8 percent, and joint ventures for
6.3 percent Equity investment was the least-used
fcnm of alliance. Although the acquisition form of
alliance is considered "noncooperative," it is still
popular, growing from 3 percent of die alliances ia
1986 to 15.1 percent in 1989 (see Table 3).
Alliances increased dramatically in the 1980s,
with an average annual growth rate of 33.3 percent,
but reached the peak in 1987. It is likely that the
decline since 1987 indicates the increasing density
of relationships within the industry because of the
limited amnber of companies in the industry.
United States
In the United States, 49.6 percent of the alliances are cooperative agreements (see T^le 4).
This is fairly close to the worldwide industry average of 53.3 percent. But the second-highest type of
strategic all^ces in the United St^Ues is acquisition at 13.1 percent of the total—nearly twice the
TABLE 2
Worldwide Alliance Type IVends
Year
ACQ
JV
EQT
Lie
SCND
COAG
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
10
7
8
24
9
10
33
33
22
1
0
1
2
12
18
18
39
15
6
4
4
4
7
6
7
11
21
13
14
2
5
6
16
11
23
18
30
16
22
0
2
4
10
24
26
35
27
9
4
3
13
17
34
67
117
169
240
220
62
1
4
12
17
29
32
39
30
19
9
21
35
52
88
153
232
300
420
325
139
112
6.3
91
5.1
149
8.4
141
7.9
942
53.3
192
10.8
1,765
Total
Pocentagc
138
7.8
TECH
Total
Somce: EIAJ. NYU. ;IHtaqimt (Febnuny 1991)
TABLE 3
Alliance Shares by Region
(Cumulative Percentages)
-
Year
United States
Japan
Europe
Asia/Padfic-ROW
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
75.4
76.3
75.3
73.1
71.2
66.1
62.1
62.0
62.7
63.3
4.9
4.7
7.9
10.2
13.3
18.2
22.4
23.4
23.1
22.4
13.4
12,2
11.6
10.7
10.5
10.2
9.8
9.4
9.4
9.3
6.3
6.8
5.2
6.0
5.0
5.5
5.7
5.2
4.8
5.0
Somce: EIAJ, NYU, 1DataquHt (Petanuiy 1991)
000938S
e i 9 9 1 Datmqneit IiKoiponted Febmary-Ileproduetian Prohibited
Sendcooductar Oraup N«wdettan 1991-02
U.S. COMPANIES TOP UST FOR STRATEGIC ALUANCES
industry average. The odier forms of alliances participated in by U.S. conq>ames are technology
transfer (9.8 percent), second-sourcing (8.3 percent), licensing (8.2 percoit), equity investment
(6.3 percent), and joint venture (4.3 percent).
Japan
Japan pursued strategic alliances more
actively in tbe latter half of the 19805, gadienng
20 percent of the total number of agreements
by 1989 (Jiblt 5). The cooperative agreement
appears to be die preferred form of alliance for
Japanese companies (60 percent as opposed to
49.6 percent for the United States). Jsqpanese companies use acquisition (2 percoit) far less than do
U.S. conq)anies (13 percent). Joint venture is the
second-most popular type of alliance in Japan
(112 percent), followed by technology transfers
(9.5 percent), licensing (6.9 percent), secondsourcing (6.3 percent), and equity investment
(3.9 percent).
It is not coincidraital that the Japanese have
chosen U.S. partners most frequently (60 percmt)
for strategic alliances, because die majority of technology developments and innovations have
occurred from U.S. research. Alliances between
Japanese companies account for 29 percent of Ji^anese alliances, while these conq)anies partner with
European companies 11 percent of the tune.
Asia/Pacific-ROW
One interesting phenomenon of industry
dynamics is die emerging role of conqianies firom
the Asia/Pacific Rim, specifically those from
Soudi Korea and Taiwan (see Figure 1). Taiwanese
companies have demonstrated a use of strategic
TABLE 4
U.S. Alliances by l^pe
Regional
Partner
ACQ
JV
EQT
Lie
SCND
258
11
25
16
39
49
3
11
106
23
16
4
132
36
8
19
130
32
28
8
310
13.1
102
4.3
149
6.3
195
8.2
198
8.3
United States
Japan
Europe
Asia/Pacific-ROW
Total
Percentage
TECH
Total*
740
293
101
46
126
60
31
16
1,538
506
214
120
1,180
49.6
233
9.8
2378
COAG
*DoUble cnuntwl' each paitici]mt m tbe alluace u ocmted ai ooe. TotaU inehide 11 uoideatified natknalitie*.
Somoe: EIAJ, NYU. Dataqueit (POaaay 1991)
TABLE 5
Japanese Alliances by T^pe
Regiona]
Partner
Japan
United States
Europe
Asia/Pacific-ROW
Total
Percentage
ACQ
JV
EQT
UC
SCND
COAG
TECH
Total*
2
11
2
1
32
49
9
4
8
23
0
2
10
36
6
6
10
32
10
1
170
293
27
14
10
60
7
3
242
506
61
31
16
1.9
94
11.2
33
3.9
58
6.9
53
6.3
504
60.0
80
9.5
840
'Double oauriad: each paitieqiaDt ki die alUince i i counted ai ooe. Totals ischide two ^identified natinnaliliwi
Somoe: EIAJ. NYU, Dtfaqueit (Febiuaty 1991)
e i 9 9 1 Dataqueit Incoiporatsd Febniary.^e{iiadueticMi PrahiUted
Semieooductoi Oroi^ N e w d e t t m 1991-02
000938S
U.S. COMPANIES TOP UST FOR STRATEGIC ALUANCES
alliances since the first two licensing agreenoents in
1979. South Korean conq)anies have actively participated in strategic alliances since 1980. South
Korean companies reached their peak in alliances
in 1986, with 2.5 percent of total worldwide alUances. Taiwan con^anies peaked in 1985 with
about 1.1 percent of total alliances.
South Korea
More than 41 percent of South Korea's
36 alliances are cooperative agreements (see
Table 6). Technology transfer (20.8 percent) and
licensing (18 percent) were the second and third
most frequently used types of alliances. Second
sourcing (9.7 percent) has also been actively pursued. Their heavy reliance on technology transfers,
licensing, and second sourcing indicates that South
Korean conqjanies have, so far, been technology
innovation followers. However, this may be gradually changing as they sink more money into R&D.
More than 76 percent of South Korean companies* alliances are with U.S. companies. But it is
interesting to note that few (only seven) of the
alliances have been with Jigianese companies. This
small portioa of J^>anese parmers mig^t indicate
the basis of an old argument agamst Japanese
companies' policies toward developing countries:
they are unwilling to share or transfer technological
know-how.
FIGURE 1
Alliance Shares—South Korea and Taiwan
Percentage
3
U South Korea
# Taiwan
2-
1-
0-
]
19 60
19S4
1982
isaa
1386
Soiirce: BIAS, N Y U , Dataquest (February 1991)
TABLE 6
South Korean Alliances by Type
Regional
Partner
South Korea
United States
Japan
Europe
Asia/Pacific-ROW
Total
Percentage
ACQ
JV
EQT
Lie
SCND
COAG
TECH
Total*
0
1
0
1
0
0
2
0
0
0
0
3
0
0
0
0
11
1
0
1
0
7
0
0
0
6
20
4
0
0
0
11
2
2
0
6
55
7
3
1
2
2.8
2
2.8
3
4.2
13
18.0
7
9.7
30
41.7
15
20.8
72
*Doable counted: eacb pitticiiant in die igieeuieut ii counted i
Souioe: EIAJ, NYU, DatM]uert (Febnuny 1991)
0009385
01991
btcpipontad Febniaiy-Repioductioa Prohibited
Semicondiictor Oraup Newiletten 1991-02
U.S. COMPANIES TOP UST FOR STRATEGIC ALLIANCES
TABLE 7
Taiwanese Alliances by Type
Regional
Partner
JV
EQT
Lie
0
1
0
0
0
0
1
0
0
0
0
0
0
2
1
4
4
0
0
1
0
1
1
0
0
0
10
2
1
0
0
2
0
0
0
4
19
3
3
2
1
3.2
1
3.2
3
9.7
9
29.0
2
6.5
13
41.9
2
6.5
31
100.0
ACQ
Taiwan
United States
Japan
Europe
Asia/Pacific-ROW
Total
Peicoitage
SCND
COAG
TECH
Total*
*DouUe counted: each pnHcipant in an alliancet is counted i
Scuice: EIAJ, NYU, DataqueA (Febniaiy 1991)
Taiwan
Licensing and technology transfers e7q)lain
35 percent of the 15 Taiwanese strategic alliances
(see Table 7). The Taiwanese companies are technology followers, similar to Soudi Korean conqpanies. However, unlike the South Korean companies,
the Taiwanese are more diversified in partner
nationality. About 13 percent of the agreements are
with domestic con:q>anies. The rest of the agreements are with offshore conq>anies as follows:
61 percent, United States; 10 percent, Europe;
10 percent, Japan; 6 percent, other countries.
DATAQUEST PERSPECTIVE
A View of the Giants
A quick look at the tables shows diat the
United States and Jiq)an dominate the number of
strategic alliances. However, differences exist in
the type of agreements used. Japmess con^Kinies
generally use cooperative forms of alliances, and
U.S. corqpanies typically take a more aggressive
posture. More than 60 percent of the alliances in
Jq>an are cooperative agreements, conq)ared with
49.6 percent in the United States. However, acquisition is used 13 percent of the time in U.S. agreements, conpared with 1.9 percent in Japan.
Actually, U.S. conq)aiues use mergers and
acquisitions more firequenfly dian do conqianies
from all the other countries in die world combined.
0 1 9 9 1 Dataqueit Ineoiponied Febnuuy-RepiDductiao Prahibited
Semicandiictor Q n n 9 Newdetten 1991-02
This is a fundamental difference between U.S.
business practices and those of con:q)anies in other
countries.
A second basic difference is that J^anese
coiiq>anies have farmed alliances with foreign partners more firequently than have U.S. coiiq>anies.
U.S. conq>anies tend to rely on domestic partners
for alliances. Japanese conq>anies select domestic
partners 29 percent of the time, whereas U.S. companies select domestic partners 65 percent of the
time. This trend might be e3q>lained by a higher
prop(Htion of technology leadership in tbs United
States; however, the leadershq> in several technologies has shifted to Jq>an.
Emerging Infiuences
Although Soudi Korean and TEUwanese companies lagged behind their counterparts in technology, diey do exhibit strong competitive qualities.
T h ^ have been acquiring technology through their
strategic alliances, ^^th growing domestic economies and increasing domestic and worldwide semiconductor demand, Dataquest believes that the
conq>anies of ttiese two countries will place an
increasingly stronger emphasis on technology
development As a result, we expect to see more
important roles played by these two emerging
countries in die future.
Marc Elliot
000938S
PSeMS^'Si'*'
K^'-h
^hm^i:
msi^m
ife^
Dataqyest
acom|»nyof
The Dun & Bradsticct Ccxporation
Research Newsletter
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY
OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
INTRODUCTION
In spite of a U.S. recession and the threat of
war, the worldwide semiconductor industiy grew in
the fourth quarter of 1990 in both bookings and
billings. The Persian Gulf war, which began on
January 16, 1991, when the allied forces started
bombing Baghdad, might be expected to cast a pall
over the entire world economy to the detriment of
the semiconductor industiy. Howevo*, Dataquest
beUeves that the industry will continue to grow,
albeit modestly, through 1991. We expect quarterly
growth to be stronger in 1992. Our annual growth
forecast by region is shown in Figure 1. Overall,
we expect 9 percent growth in 1991 and 13 percent
growth in 1992.
FIGURE 1
Annual Semiconductor Industry Growth Rates
by Regional Market
(Percentage of Dollars)
Percentage of Dollars
1983
1990
1991
1992
Source: Dataquest (January 1991)
01991 Dataquest lacoiporated Januaiy^Repnxtuction Prohibited
Semicanductor Oroup Newtletten 1991-04
0009372
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 corifidence by our clients Individual companies reported on and analyzed by Dataquest
may be clients cfthis and/or other Dataquest services This information is not famished in connection with a sale or o^r to sell securities or in connection with the solicitation c^an
q^r to buy securities This firm and its parera and/or their officers, stockholders, or members t}f their fomilies may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
The reasons for our relative optimism are as
follows:
• Our monthly survey of major OEM semiconductor procurement ntianagers continues to siq>port
improvement in the systems market outlook.
• A trade war, brought on by patriotic fervor in the
United States, could disnq)t world economies
enough to adversely affect the semiconductor
industry. This possibility is avoidable if the U.S.
government and its allies actively control events
that might result in protectionist U.S. policies.
• Semiconductor inventories at OEMs are less
than 20 days and within 8 days of target.
OUTLOOK FOR 1991 AND 1992
• Many sraniconductor manufacturers are reporting strong bookings for the month of January.
• WSTS statistics show both bookings and billings
on an upward trend through November, the last
worldwide actuals available.
• Increasing pervasiveness of semiconductors in
electronics and consumer goods and increasing
functionality per chip will continue to raise chip
average selling prices and allow the semiconductor industry to grow faster than the electronic
equipment industries.
• Telecommunications equipment production continues to do well, due in part to demand from
eastern Europe. This will continue to drive semiconductor consumption in Europe.
• There is evidence—^in the huge dpproval rating
of U.S. President George Bush, the large U.S.
stock market rallies, and signs of iinprovement
in the index of leading indicators—that U.S.
consumer confidence has increased dramatically
since the bombing of Baghdad began.
• U.S. allies have pledged $45 billion toward the
cost of the war thus far, diereby alleviating a
potentially onerous financial burden on one
nation.
To be sure, there are also possible hazards on
the horizon:
• Protraction and/or major expansion of the
war in the Persian Gulf could sabotage world
economies.
• Increased political and economic instability in
the Soviet Union could become a very e:q)losive
situation with worldwide repercussions.
• Lack of soundness of the U.S. financial system
could damage the U.S. economy if massive bank
failures were to occur. Hiis possibility can be
averted by effective action on the part of die
Federal Reserve Board, Congress, and the Bush
administration.
0009372
We have looked at several different scenarios
for semiconducttn- industry growth this year and
next They range from h i ^ y optimistic to highly ;
pessimistic. We believe that the most likely scenario is somewhere in between, with worldwide
growth of 9 percent in 1991 and 13 percent in
1992.
In the final months of 1990, both bookings
and hillings were well ahead of the same period in
1989, at 13 and 14 percent,respectively.The same
trend holds when looking at the three months ended
November 1990 versus the three months ended
November 1989. Because of this trend and because
of renewed confidence levels since fighting began
in the Gulf, we believe that the first and second
quarters of 1991 are going to show growth, with
most of it in die second quarter. We are forecasting
modest growdi in die third and fourth quarters of
1991. Wefliinkquarterly growth will be considerably higher in 1992 for the following reasons:
• We believe that the war will have been resolved.
• We believe that the U.S. savings and loan and
banking crisis wiU be in the solution phase.
• We believe that psychology wiU play a strong
role: Just as low consumer confidence contributed strongly to the U.S. recession in the
fourth quarter of 1990, a positive mind frame in
the elecfronics industry can buoy up the
semiconductor industry.
Figure 2 shows our sequential quarterly
growth history and forecast worldwide. Figure 3
shows worldwide growth by quarter versus the
same quarter a year ago.
Regionally, we expect to see die following
trends in 1991:
• North American maiket growth wiU be strongest
in the second quarter.
• European maiket growth will be strongest in the
first quarter. European soniconductor consuiiq>tion benefited in 1990 from a boom in TV and
VCR production, which we do not believe will
be repeated this year.
01991 Dataqneit Incotponted Januuy-Reproductiaa PrabiUted
SenncoDductar Oroiq) Newdetten 1991.04
-t
FIRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
FIGURE 2
Worldwide Semiconductor Industry Growth
by Sequential Quarters
Percentage of Dollars
1412
lo^
s
6
ra
4
s
2
H
L^
%
5''/'''
\
RS J
1 S
I
$
-^1
1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q 1Q 2Q 3Q 4Q
I
II
II
II
II
II
I
1987
1988
1989
1990
1991
1992
Source: Dataquest (January 1991)
FIGURE 3
Worldwide Semiconductor Industry Growth
versus the Same Quarter One Year Ago
Percentage
40-
30A
:S
20-
I
10
M
-10
N
K\
ra
\'
is
'W
^Q 2Q 3Q 4Q^^1Q 2Q 3Q 4Q^^1Q 2Q 3Q 4Q^^1Q 2Q 3Q 4Q^^1Q 2Q 3Q 4Q^^1Q 2Q 3Q 4Q^
1987
1988
1989
1990
1991
1992
Source: Dataquest (January 1991)
61991 DaUqueft Incatparated Jannuy^llepioductiao Pndubitcd
SemicoDdiietor Onnq> Newiletten 1991-04
0009372
RRST QUARTER 1991 WORLDWIDE SEMICONDUCTOR INDUSTRY OUTLOOK: EMBATTLED, BUT NOT BOMBED OUT
• Js^an will show the weakest quarterly growth of
any region this year, largly due to the
unprecedented economic challenges it is facing
in its stodc market and real estate noarket and the
slowdown in consumer spending in the United
States, upon which a large part of Japan's
semiconductor consumption depends.
• The Asia/Pacific-Rest of World (ROW) market,
which slowed in the fourth quarter of 1990 due
to a fallofif in clone demand, will resume growth
in the second quarter of 1991.
In general, the outlook for the Asia/Pacific
markets continues to be brighter than that for other
regions for the following reasons:
• The Asia/Pacific countries' GDPs in general
continue to grow at high single-digit rates.
• Much of the Asia/Pacific semiconductor demand
will come from products to be sold within the
country of manufacture. In fact, Japanese
con^anies are now producing goods in Asian
countries for sale diere rather than for export to
other regions.
0009372
• This market is still the smallest, least mature
regional market; therefore, it can support a
higher percentage growth than can other regions.
Our 1992 outlook calls for Asia/Pacific-ROW
to remain the fastest-growing regional market,
followed, in order of growth, by North America,
Europe, and Japan.
DATAQUEST ANALYSIS
Never before has Dataquest forecast semiconductor industry growth during a global conflict that
could affect worldwide economic powers. We
believe that enough positive factors exist to result
in modest industry growth both diis year and next.
The war could have either a significantly positive
effect or, conversely, a significantly depressing
effect on the semiconductor industry. We have
chosen a scenario in which most volatile effects are
counterbalanced by other influences, and life continues on, though perhaps not at the fienetic pace
of the 1980s.
Patricia S. Cox
ei991 Dauqueft Incaparated Jamiuy-Reiiiodiictian Prohibited
Semicaodiictar Group Nowiletten 1991-04
Dataquest
jam^nvof
The Pun eTBcHbtTCetCoqxjnttion
•T fit«
)'-r"J,-.,'"r.y-'-"j'
f*-
••
'^
M Research AfeH«fe/ter
D R A M PRODUCT LIFE CYCLES—A VIEW FROM THE INDUSTRY
INTRODUCTION
m
At the Dataquest Semiconductor Industry
Conference held in October in Monterey,
California, a half-day session was dedicated to the
subject of DRAMs. Six speakers discussed DRAM
topics that included DRAM product life cycles,
manufacturing costs, average selling prices (ASPs),
capaci^, packaging trends, technical characteristics, and applications. This newsletter focuses on
DRAM product life cycles, one of the DRAM
topics presented at the conference.
The DRAM product life cycle for each new
generation of DRAMs is a subject of great interest
to Semiconductor Equ^nnent Manufacturing and
Materials Service (SEMMS) clients. A secure
knowledge of DRAM product cycles he^s participants at all levels of the electronics industryincluding equipment and materials suppliers,
device manufacturers, and device end users—better
plan their future coinpany activities.
Three of the invited speakers presented their
views of DRAM product life cycles, which, in the
interest of our cUents, we present in this newsletter.
David Sear, vice president, Standard Products
Operations, Integrated Circuits Division, Fujitsu
Microelectronics, gave Fujitsu's view; Dr. Tsugio
Makimoto, director and general manager, Semiconductor Design and Development Center, Hitachi,
presented Hitachi's view; and Robert J. Brown,
senior vice president and group executive, Semiconductor Operations Group, Toshiba America
Electronic Components, presented Toshiba's view.
We also augment the speakers' viewpoints with
Dataquest's perspective.
D R A M PRODUCT LIFE CYCLES
Figures 1, 2, and 3 show Fujitsu's, Hitachi's,
and Toshiba's views of DRAM product life cycles,
respectively. The figures show Has life cycles of
each DRAM generation and the projected worldwide unit demand throughout the DRAM's life
cycle. The three figures probably were meant to
present general industry projections radier than
official company forecasts, and Dataquest does not
want to manipulate these data to arrive at decisions
unwarranted by the accuracy of die data. Nevertheless, significant differences exist among the general
viewpoints of these companies, such that the reader
should examine the cluuts in more detail. (The
charts also use different scales, so the reader should
be careful with direct con^arisons of the charts.)
For Fujitsu, worldwide peak unit production
for each new generation of DRAMs surpassed diat
of the previous generation. Toshiba's view is
similar to Fujitsu's, at least for the 256K, 1Mb, and
4Mb generations. Hitachi's view is somewhat
different: It projects that peak production, at least
for the 1Mb, 4Mb, and 16Mb DRAMs, essentially
is the same for each new generation and lower than
the peak production for 256K DRAMs. Table 1
shows some rough data that Dataquest extracted
from Figures 1 through 3. Using the 16Mb DRAM
as an example, Fujitsu forecasts peak unit production to be 1,500 ndUion units, twice Hitachi's estimated peak production of 750 million units.
Fujitsu estimates that worldwide DRAM
demand will exceed 500 million units for each
DRAM generation from 1Mb through 64Mb for six
to seven years, while Hitachi estimates that worldwide deinand will exceed 500 million units for
only diree to four years for the 1Mb, 4Mb, and
16Mb DRAM generations.
When Fujitsu's higher peak unit production
and longer time for the DRAM generation to be
above 500 million units is conq>ared with Hitachi's
lower peak production and shorter time above
500 million units, it is clear that Fujitsu forecasts a
much larger niunber of units for each DRAM
generation than does Hitachi.
O1990 Dataqiieft Incoipoiated December-Reproduction Prohibited
SEMMS Newiletters 1990 Indiutiy bnies
DOOaOTO
The content of this report represents our inlerpretation and analysis cf irtformation generally avaiUtbte 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 corr/tdence by our clients. Individual companies reported on and analyz&i by Dataquest
may be clients of this and/or other Dataquest services. This ir^jrmation is notfitmished in connection with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities This firm and its parent and/or their cfficers, stockholders, or members of their fiimilies may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
DRAM PRODUCT LIFE CYCLES—A VIEW FROM THE INDUSTRY
FIGURE 1
Fujitsu's Projected DRAM Life Cycle
3000 - r -
2500 -4-
Units
(M)
2000
+
1500 4 -
1000 H—>.--.^..j.—
500 4 -
8 4 ' 8 5 ' 86- 8 7 ' 8 8 ' 8 9 ' 9 0 ' 9 1 ' 9 2 ' 93" 9 4 ' 9 5 ' 9 6 ' 9 7 ' 9 8 ' 9 9 ' 2 0 0 1 ' 0 2 ' 0 3 ' 0 4 ' 0 5 ' 0 6 '
00'
Source: Fujitni
FIGURE 2
Hitachi's DRAM IVend
PER CAPITA BITS
8.000K
Million
PCS
_ 10'
10«
160K_„„„-
'
to'
-
• 10*
OJBK„„-
'
"
10'
10^
1.000
UNIT SHIPMENT
A \/>
/
\
(
/ 64K \
1
256K
\ /
I M \
f^6M \
/AM
y
SOO
64M>
0
1980
'
1990
y^
256r4/^
/
>
2000
Souice: Hitachi
0009070
O1990 Dauqueit Incoiponted December-^epiodiictioii PiohiWted
SEMIwIS Newiletten 1990 Induitry Inuei
DRAM PRODUCT LIFE CYCLES—A VIEW FROM THE INDUSTRY
FIGURE 3
Toshiba's DRAM Market Forecast
Units (Millions)
1,400
1,200
1MEG
1,000
-»/
800
4MEG ''
600
400
200 -
'87
leUEQ
'88
•89
•90
•91
•92
•93
•94
Soiuce: Toshiba
Although these are only a few conq)anies'
views of the future, the fact that they are disparate
leads to caution. The semiconductor industry in
general needs to have a fairly consistent view of
the future in order to avoid the cyclicaHty that has
characterized the industry. For instance, if the consensus forecast is for h i ^ worldwide demand, then
capital expansions are more likely to be made to
meet the projections, and the coiiq>anies making
expansions will be able to share in die rising
market. On the other hand, if actual demand turns
out to be lower than was originally forecast, the
industry will have an overcapacity situation with all
die concomitant problems diat such a situation
brings to the semiconductor manufacturers and the
equipment and material suppliers.
Similarly, if the forecast is for a lower worldwide demand, then fewer coital expansions will be
made, and the companies moderating their e}q)ansions will not find themselves in an overci^acity
situation. However, if the actual demand proves to
be much higher than originally forecast, then die
industry will find itself in an undersupply situation
with all of the problems that entails for the end
users of die devices.
Table 1 shows that the dnee companies agree
on the year of peak unit production for each
O1990 Dttaqueit Inccnpotated Deceiiiber4lepiodiictiaii Prohibited
SEMMS Newdetten 1990 Induiby bnies
DRAM generation~1991 for die 1Mb, 1995 for die
4Mb, and 1999 for the 16Mb. Fujitsu estimates that
the 64Mb DRAM will reach peak production
between die years 2002 and 2003. The figures also
show that DRAM generation life cycles for the
1Mb, 4Mb, and 16Mb devices are q)proxJniately
nine to ten years or more.
DATAQUEST PERSPECTIVE
Dataquest's projections for peak unit production of the 64K dnough 4Mb devices also are
shown in Table 1.
Dataquest estimates that the 1Mb DRAM will
reach a peak unit production of 1,075 million units
during 1991, which is similar to Fujitsu's and
Toshiba's projections, but considerably higher than
Hitachi's forecast of 750 million units. We believe
that 1Mb suppliers currentiy are in an oversupply
and overcf^acity situation, which continues to
drive down prices for 1Mb DRAMs. We also
believe that major European and Japanese DRAM
suppliers are cutting back production volumes of
1Mb devices. The number of cutbacks and the
timeliness of these cutbacks may not be adequate
to reduce the estimated oversupply for the
0009070
TABLE 1
DRAM Maricet Tk%nds—DifTerent Views
(Millions of Units)
64K
Peak
Year
Units
at Peak
256K
Peak
Units
Year
at Peak
Peak
Year
1Mb
Units
at Peak
4Mb
Peak
Units
Year
at Peak
16Mb
Peak
Units
Year
at Peak
Fujitsu
NA
NA
1988
800
1991
1,000
1995
1,200
1999
1,500
Hitachi
1984-1985
900
1988
875
1991
750
1995-1996
750
1999-2000
750
Toshiba
NA
NA
1987-1988
950
1991
1,150
NA
NA
NA
NA
1984
852
1988
963
1991
1,075
1995
uoo
NA
NA
Dataquest
NA - Not Kv«lUUe
Source: Pi^itni, HilKin, Todnba, Dataipiejt (Deceniber 1990)
2
DRAM PRODUCT UFE CYCLES—A VIEW FROM THE INDUSTRY
remainder of fliis year and 1991. ^^th the increasing availability of and demand for 4Mb DRAMs,
many end users may consider purchasing more
1Mb devices to protect their investments in
products that require the 1Mb device.
Considering the 4Mb DRAMs, we estimate
that peak unit production will occur in 1995 and
will be more than 1,200 million units.
A number of semiconductor manufacturers
have announced availability of 16Mb DRAM
engineering samples starting in eidier the fourth
quarter of 1990 or the first quarter of 1991. TTie
manufacturers include Fujitsu, Hitachi, Matsushita,
NEC, Samsung, Siemens, Texas Instruments, and
Toshiba. Engineering complications stiU exist for
the 16Mb DRAM, including the trench or stacked
capacitor cells and standardized packaging dimensions. We believe that volume shipments of the
16Mb DRAM will not occur until 1992 and that
peak production occurring in the years 1999 to
2000 is realistic.
Hitachi announced late last summer that it has
a prototype of a 64Mb DRAM and is rs^idly
working toward a fiiUy functional sample. It is
unlikely that diis sanq)le will be available until
after 1992.
DATAQUEST CONCLUSIONS
For equipment and materials suppliers,
DRAM product life cycles indicate when each new
generation of equipment and materials is required.
Dr. Graydoc Lanabee of Texas Instruments has
estimated that development of manufacturing
processes and production equipment for each new
generation of semiconductor devices must begin
about eight to ten years before the year in which
volume sh^ments of the device first occur. For
instance, the projections shown in Figures 1 and 2
indicate that volume shipments of the
64Mb DRAM will commence around 1997. This
start date means that 64Mb DRAM processes and
equipment must already be under substantial
development, and, indeed, this is the case indicated
by Hitachi's 64Mb DRAM announcement.
The eight- to ten-year development cycle also
means that equipment and material suppliers need a
long-torn ifjproach to R&D that includes both
near-term and long-term projects. The challenge is
how can the smaller equipment and materials
suppliers generate sufQcient R&D funds to accomplish such a strategy? The challoige extends to the
smaller semiconductor manufacturers as well—are
they making su£Bcient R&D investments in longterm projects to ensure their future survival? U.S.
wafer fab equipment manufacturers and semiconductor manufacturers already are sustaining an
R&D investment rate of about 14 percent of sales,
which is among the highest of any industry.
Joe Grenier
lone Ishii
Tbc topics coveted by SEMMS newsletters are selected for their general interest to SEMMS clients, vitich include wafer fab eqiiq)ment
siq^liers, semicondnctor materials companies, and senucondnctor device mamifiictnrets. t h e topics selected inaicMK tfae broad range of research
that is conducted in the SEMMS groq>. Clients, however, often have specific infonnatian requirements that either go beyond the level of detail
contained in die newsletters or beyond die scope of vibat is tmmally puldished in the newsletters. In order to provide con^lete decision sqiport
to onr clients, Dataquest has a consulting semce available to handle these additional infomiation needs. Please call Stan Bmederle at (406)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
O1990 Dataqueit Incotpcnated December-RepriMfaietian PioIiiUted
SEMMS Newiletlan 1990 Induitiy bniei
0009070
DataQuest
-'• •'•A
^:s
^sS
acompaiwof
Corporabon
The Dun ft'eradsticct
Research Newsletter
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR
MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
for Intel, Motorola, Texas Instruments, National
Semiconductor, and Philips.
INTRODUCTION
Dataquest has completed its preliminary
analysis of 1990 semiconductor market shares for
more than 150 semiconductor vendors worldwide.
We have reached the followiag conclusions based
on surveys of these vendors and our analysis of the
market:
MOS memory revenue dropped by 17 percent
worldwide, resulting in market share losses for
the companies that participate in this market.
For the first time since 1982, Japanese companies lost share of the worldwide market, droj)ping fiom a 52.1 percent market share in 1989 to
49.5 percent m 1990.
• In a worldwide sotniconductor market that grew
only 2 percent, MOS microcomponents grew a
whopping 23 percrait, paying off in a big way
FIGURE 1
Regional Shares of Worldwide Semiconductor Market
(Percentage of Dollars)
Percentage of Worldwide Market
100
197a
1980
1962
19B4
1986
1968
1990
Source: Dataquest (January 1991)
61991 DiUqueit bcoiponted Jamuiy-Jlepiadactiaii Prohibited
s o Maiket Share Newilettecs 1991-1
0009146
Tlie content of this report represents our interpretation and analysis of itiformation 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 conlain molerial provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This iri^rmation is notfitmished in connection with a sale or offer to sell securities or in connection with the solicitation c^an
c^T to buy securities. This firm artd its parent andlor their officers, stodcholders, or members of theirfittniliesmay, fiom time to time, have a kmg or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
• For the first time since 1979, North American
cotc^anies gained market share, growing from
34.9 percent of the worldwide market in 1989 to
36.5 percattt in 1990.
• "While Asian companies held steady at 3.5 percent market share, European companies reversed
a downward slide and gained a percentage point
of worldwide market share at 10.5 percent. We
e^qpect this upward trend to continue in the
future.
• Japanese companies now control 22 percent of
the North American semiconductor market,
down from 24 percent in 1990; at die same time,
North American conq)anies have increased dieir
Japanese market share to 10.4 percrait, up from
less than 10 percent in 1989.
Figure 1 shows the worldwide market share
held by each regional company base.
RANKINGS
Table 1 lists the top 20 semiconductor suppUexs worldwide. Although the semiconductor market
grew only 2 percent in 1990 (iu line with our
forecast of a flat year), the growth rates of
individual players varied widely, depending on
product portfolio. Among the top 20, Intel's
29 percent growth because of its strength in
microcon^nents was by far the highest, catapvlting Intel to the number five position worldwide, up
from number eight in 1989.
By the same token, although the top four
players remained the same as in 1989 (NEC, Toshiba, Hitachi, and Motorola), the first three each
experienced revalue declines of 1 percent because
of the heavy proportion of MOS memory in dieir
product portfolios. Motorola, on the other hand,
was able to grow 11 percent because of its strong
microconq}onent growth.
The bipolar digital IC market is still shrinking; it declined 1 percent in 1990. Although most
players in this market are suffering because of a
shift to CMOS, Fujitsu was able to buck this trend
with its stq)er ECL gate arrays, which are used in
its new mainframe computer (among other
products), which had very strong growth this year.
This high-ASP product enabled Fujitsu to surpass
Texas Instruments to become tlie top-ranked
bipolar digital supplier iu 1990.
0009146
MOS memory, the largest semiconductor
product category, sufferedfromfirce-faUingDRAM
prices and a slowdown in nonvolatile memory
demand, with revenue declining 17 percent from
1989. The Japanese companies clearly were the
most severely hit because of their dominance in
this market. Samsung's DRAM revenue actually
grew in 1990, however, because of its shift to 1Mb
production. In 1989, most of its production had
been 256K. Sharp was able to grow its MOS
memory revenue on the strength of its swift entry
into the SRAM market and strong demand from the
game market for its 8Mb ROMs.
As mentioned previously, MOS microcomponents was the fastest-growing product category,
increasing 23 percent from 1989. Intel retained its
position as the number one supplier, growing
41 percent; in fact, it strengthened its lead over the
number two supplier, NEC, quite sigmficanfly.
North American and European companies succeeded in taking market share in microcomponents
from Japanese companies.
MOS logic grew by 5 percent in 1990. The
top 5 players remaiaed the same—NEC, Toshiba,
Motorola, Fujitsu, and LSI Logic. Among the top
20, VLSI Technology showed strong growth at
25 percent and moved from number 18 in 1989 to
number 14 in 1990. Siemens also showed strong
growth at 24 percent and moved from number
20 to number 17.
In the analog market, European companies
showed extremely strong growth over 1989,
because of the strength of the telecommunications
market in Europe. European telecom conq)anies did
very weU in 1990, winning projects not only in
Europe but in third-world countries as well as
Eastern Europe. Philips, the analog revenue of
which grew by 17 percent and which junq)ed to the
number one spot in the rankings, also profited from
its own very strong consumer electronics business.
The analog growth rate in 1990 also is influenced
by a change in our definition of mixed-signal IC
revenue, some of which was reported previously in
the MOS logic category.
The discrete and optoelectronic markets both
showed growth in 1990. Particularly strong growth
was shown by International Rectifier, which grew
30 percent in discrete and went from the number
14 position to number 12.
Tables 2 through 9 list the top 20 suppliers in
the product segments of total integrated circuit,
01991 Ditaqueit Incoiponted lamiaiy-jleproduetion PiohiWted
SG Market Share Newdetten 1991-1
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 1
Preliminary Estimated Worldwide Market Share Ranking
Total Semiconductor
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share {%)
1
1
NEC
5,015
4,952
(1)
8.5
2
2
Toshiba
4,930
4,905
(1)
8.4
3
3
Hitachi
3,974
3,927
(1)
6.7
4
4
Motorola
3,319
3,692
11
6.3
5
8
Intel
2,430
3,135
29
5.4
6
5
Fujitsu
2,963
3.019
2
5.2
7
6
Texas Instruments
2,787
2,574
(8)
4.4
8
7
Mitsubishi
2,579
2,476
(4)
4.2
9
9
Matsushita
1,882
1,945
3
3.3
10
10
Philips
1,716
1,932
13
3.3
11
11
National Semiconductor
1,618
1,718
6
2.9
12
13
SGS-Thomscm
1,301
1.463
12
2.5
13
12
Sanyo
1,365
1,381
1
2.4
14
15
Sharp
1,230
1.360
11
2.3
15
14
Samsung
1,260
1.315
4
2.3
16
16
Siemois
1,194
1,221
2
2.1
17
19
ScHiy
1,077
1.172
9
2.0
18
17
Oki
1,154
1,074
(7)
1.8
19
18
Advanced Micro Devices
1,100
1,067
(3)
1.8
20
20
AT&T
873
830
(5)
1.4
57,213
58,414
2
100.0
Total Market
Source: Dataqueit (laminy 1991)
bq)olar digital, MOS memary, MOS microcomponent, MOS logic, analog, discrete, and opto.
llie following notes apply to the tables in this
newsletter:
Some revalue reported in 1989 as MOS logic
was reported in 1990 as analog (mixed signal),
because of a change in our definitions.
• Our company base has grown from apptoidmately 125 in 1989 to 155 in 1990.
NA K Not available
01991 Dataqueit Incoipaiated JaiuuijMleiirodiictiaD Piobibited
SO Market Shan Newaletten 1991-1
NM = Not meaningful
000914«
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 2
Preliminary Estimated Worldwide Market Share Ranking
Total Integrated Circuit
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
NEC
4,321
4,263
(1)
9.0
2
2
Toshiba
3,774
3,684
(2)
7.8
3
3
Hitachi
3,218
3,195
(1)
6.7
4
7
Intel
2,430
3,135
29
6.6
5
6
Motorola
2,519
2,851
13
6,0
6
4
Fujitsu
2,738
2,777
1
5.9
7
5
Texas Ihstnunents
2,691
2,488
(8)
5.2
8
8
Mitsubishi
2,185
2,092
(4)
4.4
9
9
Nati(»ial Semiconductor
1,548
1,645
6
3.5
10
10
Hiilips
1,250
1,416
13
3.0
11
11
Matsushita
1,244
1,285
3
2.7
12
12
Samsung
1.182
1,238
5
2.6
13
15
SGS-Thomson
1,019
1,148
13
2.4
14
14
Advanced Micro Devices
1,100
1,067
(3)
2.2
15
13
Oki
1,111
1,031
(7)
2.2
16
17
Shaip
902
1,021
13
2.2
17
16
Sanyo
975
979
0
2.1
18
18
Sinnens
847
833
(2)
1.8
19
19
Sony
732
817
12
1.7
20
20
AT&T
716
681
(5)
1.4
46^24
47,426
1
100.0
Total Market
Somce: DiUqueit (laauny 1991)
0009146
01991 Dataqueit Incoipotaled Januaiy^Repioduetioa Ptahibited
SO Market Share Newiletten 1991-1
PREUMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 3
Preliminary Estimated Worldwide Market Share Ranking
Bipolar Digital
(Millions of Dollars)
.;;
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
2
Fujitsu
617
710
15
15.9
2
1
Texas Instniments
671
663
(1)
14.8
3
3
Hitachi
479
510
6
11.4
4
5
National Semiconductor
458
440
(4)
9.8
5
6
Motorola
369
408
11
9.1
6
4
Advanced Micro Devices
474
380
(20)
8.5
7
7
Philips
306
299
(2)
6.7
8
8
NEC
302
292
(3)
6.5
9
9
Mitsubishi
125
121
(3)
2.7
10
11
Toshiba
102
113
11
2.5
11
12
Sanyo
67
67
0
1.5
12
16
Harris
50
60
20
1.3
13
13
AT&T
56
59
5
1.3
14
14
Raytheon
55
54
(2)
1.2
15
15
Siemens
54
53
(2)
1.2
16
17
Oki
48
47
(2)
1.1
17
MM
0
40
NM
0.9
18
18
Goldstar
32
32
0
0.7
19
19
CSiips & Technologies
24
25
4
0.6
20
20
Applied Micro Circuits Ccnp.
20
24
20
0.5
4,510
4,472
(1)
100.0
GEC Plessey
Total Market
Source: Ditaqiieit (lanuaiy 1991)
01991 DaUqueit Incarponted Jamuiy^Reproductian Piofaitrited
s o Market Share Newdetten 1991-1
0009146
PREUMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 4
Preliminary Estimated Worldwide Market Share Ranking
MOS Memory
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
Toshiba
1,918
1,681
(12)
12.3
2
2
NEC
1,739
1,453
(16)
10.7
3
3
Hitachi
1,534
1346
(12)
9.9
4
4
Fujitsu
1,265
1,114
(12)
8.2
5
5
Mitsubishi
1,161
997
(14)
7.3
6
7
Samsung
935
971
4
7.1
7
6
Texas Instruments
1,095
741
(32)
5.4
8
8
Sharp
476
547
15
4.0
9
12
Motorola
407
409
0
3.0
10
9
Oki
473
392
(17)
2.9
n
10
Intel
433
344
(21)
2.5
12
11
Siemens
416
344
(17)
2.5
13
14
Matsushita
370
319
(14)
2.3
14
15
SGS-Thomson
269
299
11
2.2
15
13
Micron Technology
395
286
(28)
2.1
16
16
Advanced Micro Devices
258
280
9
2.1
17
18
Sony
228
252
11
1.9
18
17
NMB Semiconductor
247
201
(19)
1.5
19
21
Cypress Semiconductor
149
159
7
1.2
20
23
National Semiconductor
138
147
7
1.1
16.361
13,612
Total Market
(17)
100.0
Somoe: Dmaifialt Qmaiay 1991)
0009146
01991 DaUqueit Incoiponied InHiaiy-ReprDductiaa Pndiibited
SG Mericet Shue Newdetlen 1991-1
PREUMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 5
Preliminary Estimated Worldwide Market Share Ranking
MOS Microcomponent
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
Intel
1,929
2,718
41
27.0
2
2
NEC
937
1,083
16
10.7
3
3
Motorola
803
1,002
25
9.9
4
4
Hitachi
554
648
17
6.4
5
5
Mitsubishi
435
462
6
4.6
6
6
Toshiba
407
449
10
4.5
7
7
Texas Instnunents
252
320
27
3.2
8
8
Matsushita
217
240
11
2.4
9
10
Fujitsu
211
239
13
2.4
10
11
National Semiconductor
172
237
38
2.4
11
9
Chips & Technologies
216
230
6
2.3
12
12
Advanced Micro Devices
172
200
16
2.0
13
17
Philips
131
189
44
1.9
14
13
SGS-ThomscHi
161
175
9
1.7
15
16
Western Digital
135
148
10
1.5
16
14
Oki
149
147
(1)
1.5
17
15
AT&T
141
145
3
1.4
18
19
Sharp
112
134
20
1.3
19
31
Cirrus Logic
29
129
345
1.3
20
18
Harris
115
110
(4)
1.1
8,202
10,076
Total Market
23
100.0
Source: Ditiqiieit (Tammy 1991)
01991 Ditiqueit loGinpanled Jamuiy-ReiiiDduetian ProhiUted
SO Market Share Newdetten 1991-1
0009146
8
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 6
Preliminary Estimated Worldwide Market Share Ranking
MOS Logic
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
NEC
928
1,036
12
11.7
2
2
Toshiba
775
832
7
9.4
3
3
Motorola
495
553
12
6.2
4
4
Fujitsu
482
550
14
6.2
5
5
LSI Logic
445
504
13
5.7
6
6
Oki
406
410
1
4.6
7
7
Hitachi
319
354
11
4.0
8
8
Matsushita
267
309
16
3.5
9
10
Texas Instnimotts
256
306
20
3.4
10
11
Sharp
249
271
9
3.1
11
9
AT&T
257
267
4
3.0
12
12
Philips
231
235
2
2.6
13
13
National Semiconductor
222
219
(1)
2.5
14
18
VLSI Technology
169
211
25
2.4
15
14
Harris
210
201
(4)
2.3
16
17
Sanyo
178
194
9
2.2
17
20
Siemens
133
154
16
1.7
18
22
Samsung
123
153
24
1.7
19
21
Yamaha
130
145
12
1.6
20
15
Seiko Epson
201
128
(36)
1.4
8.884
5
100.0
Total Market
8,461'
Somoe: DoaqoMt (Jnauny 1991)
0009146
0199I Dataquett Incnporaled Jumaiy^Rqnadiictioii Piohifeited'
SO MaifcBt Shate Newiletten 1991-1
PREUMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 7
Preliminary Estimated Worldwide Market Share Ranking
Analog
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Cliange
1990
Market
Share (%)
1
4
Philips
522
613
17
5.9
2
1
Toshiba
572
609
6
5.9
3
2
National Semiconductor
558
602
8
5.8
4
8
SGS-Thomson
393
554
41
53
5
3
Sanyo
530
541
2
5.2
6
5
Motorola
445
479
8
4.6
7
6
Texas Instruments
417
458
10
4.4
8
9
Mitsubishi
384
434
13
4.2
9
10
Matsushita
376
403
7
3.9
10
11
Sony
361
401
11
3.9
11
7
NEC
415
399
(4)
3.8
12
12
Analog Devices
337
360
7
3.5
13
13
Hitachi
332
337
2
3.2
14
15
Rolim
277
282
2
2.7
15
14
Harris
280
260
(7)
2.5
16
16
AT&T
249
197
(21)
1.9
17
151
0
195
NA
1.9
18
24
Silicon Systems
112
180
61
1.7
19
19
Siemens
152
175
15
1.7
20
17
Fujitsu
163
164
1
1.6
9390
10382
11
100.0
GEC Plessey
Total Market
Soutce: Dtuqiieit (lanuiy 1991)
61991 DaUquut Ihcaipanted Januaij^-Reproductian PndnUted
SO Maiket Shoe Newilotton 1991-1
0009146
10
PREUMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIQNS
TABLE 8
Preliminary Estimated Worldwide Market Share Ranking
Total Discrete
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
Toshiba
848
910
7
11.0
2
2
Motorola
775
814
5
9.9
3
3
Hitachi
690
662
(4)
8.0
4
4
NEC
574
565
(2)
6.8
5
5
Philips
442
494
12
6.0
6
6
Mitsubishi
364
352
(3)
4.3
7
7
Matsushita
332
351
6
4.2
8
8
Rohm
301
320
6
3.9
9
10
SGS-Thomson
282
315
12
3.8
10
9
Fuji Electric
287
312
9
3.8
11
11
Siemens
232
260
12
3.1
12
14
Interoatioaal Rectifier
187
243
30
2.9
13
12
Sanyo
230
232
1
2.8
14
13
Sankoi
213
224
5
2.7
15
NA
Shindengen Electric
NA
177
NA
2.1
16
15
General Instrument
170
173
2
2.1
17
16
rrr
155
161
4
1.9
18
17
AT&T
147
135
(8)
1.6
19
18
Harris
120
130
8
1.6
20
19
Fujitsu
109
117
7
1.4
7,662
8;262
8
100.0
Total Market
Somoe: Dataqueit (Tamaiy 1991)
0009146
e i 9 9 1 Dataquect Incoipacatcd J•lIllatJ^4leIlIoducti<n Prohibited
SO Maricet Share Newdetten 1991-1
11
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
TABLE 9
Preliminary Estimated Worldwide Market Share Ranking
Total Optoelectronic
(Millions of Dollars)
1990
Rank
1989
Rank
1989
Revenue
1990
Revenue
Percent
Change
1990
Market
Share (%)
1
1
Shaip
328
339
3
12.7
2
2
Toshiba
308
311
1
11.6
3
3
Matsushita
306
309
1
11.5
4
4
Sony
249
270
8
10.1
5
5
Hewlett-Packaid
213
223
5
8.3
6
6
Sanyo
160
170
6
6.4
7
9
Siemens
115
128
11
4.8
8
8
Fujitsu
116
125
. 8
4.7
9
7
NEC
120
124
3
4.6
10
10
Rohm
96
105
9
3.9
11
11
Telefunkoi Electronic
78
90
15
3.4
12
13
Hitachi
66
70
6
2.6
13
12
Optek
77
66
(14)
2.5
14
14
Quality Technologies
38
35
(8)
1.3
15
15
Texas Instraments
36
33
(8)
1.2
16
16
Oki
33
33
0
1.2
17
18
Mitsubishi
30
32
7
1.2
18
19
Motorola
25
27
8
1.0
19
17
Hmeywell
31
25
(19)
0.9
20
20
Philips
24
22
(8)
0.8
2,627
2,676
Total Market
2
100.0
Sauce: Dittqueit (Imuaiy 1991)
01991 Ditaqueit Inccnponled Jaiiuaiy.^einDductiaa Probibited
SO Marimt Shue Newdetten 1991-1
000914«
12
PRELIMINARY 1990 WORLDWIDE SEMICONDUCTOR MARKET SHARE ESTIMATES: THE MICROPROCESSOR REIGNS
DATAQUEST ANALYSIS
The year 1990 was a flat one for the worldwide semiconductor industry. Although the first
half of the year showed some surprising strength in
end markets, particularly personal con:^uters, this
demand slacked off as the year progressed. This
slackening, combined with severe downward pricing pressure on memories (in spite of unit demand
growth), an uncertain economy, industry layoffs,
and the unrest in the Middle East, caused a
depressed state of mind in the semiconductor
industry.
The bright spots in 1990 were high
microcomponent demand and the moderately positive growth in the stable analog, discrete, opto, and
MOS logic markets.
0009146
The Asia/Pacific con^anies, while unable to
increase market share, were at least able to hold
steady. North American and European companies
should be h^Fpy with their gain^ market share.
Japanese companies' revenue declined, but they
still control more of the semiconductor market than
any other company base. TTie uncharted waters of
1991 may demonstrate if die strategy of relying
heavily on DRAMs is still the way to maintenance
or growth of market share in (he semiconductor
market of the 1990s.
Patricia S. Cox
ei991 Ditaquest Incaiponled Jamaiy^epRiductioa PiohiUted
SO Muktt Sbue Newdetlen 1991-1
S 8
I. .' J
Dataqyest
,." • '•• V n n acompanvof
• M O TheDunarBiadsticetCorporation
Research i\fevwfe/ter
EQUIPMENT COMPANY OWNERSHIP—AT THE TURNING POINT
Figure 1 shows the persistent advance in market share by Japanese wafer fab equipment coinpanies and the concomitant loss of share by U.S.
equipment conq)anies for key segments of wafer
processing equipment. It is clear from Figure 1 that
Japanese equipment companies have become dominant in the overall wafer fab equipment market, but
what is h:^>pening in the individual key equ^nnent
segments? This newsletter examines Qxe shift in
regional company market shares for the key segments of the wafer fab equqnnent industry. It identifies the turning points and discusses the reasons
behind die transfer of dominant share from one
group of regional companies to another.
In the 1960s and 1970s, the wafer fab equipment industry was dominated by U.S. companies.
However, as the industry matured in the 1980s, one
of the major trends that emerged was the steady
gain in worldwide market share for Japanese
equ^nnent con^anies. This steady gain has been
due to the growth of a vigorous don^stic semiconductor device manufacturing industry in J^an,
which in 1990 accounted for $2,943 miUion of the
$5,813 million worldwide wafer fab equipment
market. A dominant position in their home market,
coupled with a growing export business, have contributed to establishing Je^anese equipment companies as market share leaders in the worldwide wafer
fab equqnnent industry.
FiGUKE 1
Worldwide Market Share of Regional Companies for
Key Equipment Segments*, 1982-1990
(Percentage of Dollars)
Percentage
Japanese Companies
WM
U.S. Connpanles
I
European Connpanles
I Joint Venture Companies
100
80
60
40-
20-
S
l^SS
1962
1S83
^ ^
1SB4
1985
1966
1987
1968
1989
^
1990
Source: Dataquest (April 1991)
C1991 Dataqueft Incoipoiated April-Repioduction Pndubited
SEMMS Newtletten 1991 Equipment—Wafer Fab Equipment/Inihutry Tieads
0010257
^ = ^ £ # g s s i r i s s l ^ ••
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
EQUIPMENT COMPANY OWNERSHIP—AT THE TURNING POINT
MAINTAINING OR GAINING SHARE
Table 1 identifies the dominant share held by
U.S., Japanese, and European suppliers for severed
key segments of wafer fab equipment in 1990. The
1984 percentages are include as a point of comparison. E>ominant share is measured on the basis of
unit shipments rather than revenue because doing
so removes the effect of currency appreciation during this time period. Dominant share represents the
largest value of market share held by a group of
regionally owned companies. For exanq>le, U.S.
conq>anies held dominant share of medium- and
high-onrent implanters in 1990 with 47 percent of
worldwide shipments, while Jq)anese and jointventure companies accounted for 29 and 24 percent
share, respectively. In this analysis, specific segments of wafer fab equipment were chosen with
two criteria in mind: technological importance in
advanced wafer processing and clearly definable
trends in dominant share held by regional
companies.
Two clear trends are evident in Table 1. When
the percent share in 1990 is con^ared with the
1984 level, U.S., Ji^anese, and European companies have all maintained dominant share in certain
segments of wafer fab equipment, but only Japanese companies have gained dominant share in any
given segment of equipment.
Strengths Maintained
U.S. Companies
Nontube (dedicated) CVD reactor technology
(APCVD, LPCVD, PECVD) continues to be one of
the major strengths of U.S. wafer fab equipment
TABLE 1
Dominant Share of 1990 Unit Shipments by Company Regional Ownership for
Selected Segments of Wafer Fab Equipment
(Percentage of Worldwide Unit Shipments)
1984
1990
61
100
50
66
82
56
62
76
77
47
72
54
Japanese Companies
Vertical tube difiiision
CD SEM
71
*
78
84
European Companies
HorizcHital tube PECVD
Molecular beam epitaxy
78
63
94
53
39
27
29
24
24
80
51
55
49
54
Maintained Dominant Share
U.S. Companies
APCVD
Nontube LPCVD
Nonmbe PECVD
High- and medium-currrait implanters
Rapid themial processing
Optical CD
Gained Dominant Market Share
Japanese Companies
Steppers
Diy strip
Diy etch
Horizcmtal and vertical tube LPCVD
Sputtering
*lUf mnket emnged after 1984.
Somoe: DaUqueet (i^nil 1991)
00102S7
61991 Dataqiiett Licoipatated Aptil-Repioductian PraUUted
SEMMS Newiletten 1991 Equipment—Wafer Fab Equipinmt/Indnitiy Ttendi
EQUIPMENT COMPANY OWNERSHIP—AT THE TURNING POINT
coirpanies. They maintain their number one position in this area because of their significant
installed base of knowledge in reactor design and
process technology. Dataquest believes that to culture dominant share away from U.S. companies in
this area would require far more eflfort than in any
other segment of wafer fab equ^iment because of
the process-intensive nature of CVD.
U.S. companies continue to maintain their
dominant share position in the medium- and highcmrent tmplanter equipment segmmts. Two longtime suppliers totibteindustry, Eaton and Varian,
and a more recent player. Applied Materials,
account fox all of the medium- and high-current
inq)lanter shipments by U.S. conq)anies. Dataquest
believes that the U.S. companies have continued to
maintain dominant share in this segment because
all three vendors have a strong international
presence, particularly in Japan, through joint venture con^anies (Sumitomo/Eaton Nova and TEL/
Varian) or overseas operations (Applied Materials
Japan).
European Companies
Table 1 identifies two segments of the wafer
fab equtpment industry in Mtoch European companies have established and maintained market
strength throughout much of the 1980s. In the case
of horizontal tube PECVD systerm, a single European vendor, ASM International, dominates this
market application. European companies have
maintained a strong position in the molecular beam
epitaxy (MBE) market because of an historically
strong base in high-vacuum technology.
At the Turning Point—Dominant
Share Gains
As shown in 'Mile 1, Jf^anese equqnnent
companies have gained tlie position of dominant
share in five key areas of wafer fab equipment in
which U.S. companies had previously dominated
the market. Table 2 identifies the turning point, or
specific year, when the ^lift in dominant share
occurred. In this analysis, a difference of five or
fewer percentage points in share of unit shipments
is considered a tie in determining dominant share.
Thus, for steppers, the turning point was during the
1985/1986 time frame when U.S. and Japanese
equipment companies held essentially equal share.
From 1987 to the presoit, however, Jqianese companies have continued to increase their share of
worldwide stepper unit shipments. The turning
point in dominant share for dry strq> and tube
LPCVD systems was in 1988, the turning pomt in
dry etch was in 1989, and tiie turning point in
sputter systems appears to have occorred in 1990.
It should be noted that in our analysis,
dominant ^lare in unit sb^nnents was tmuched by
dominant share as measured on a revenue basis,
Japanese Companies
As shown in Table 1, Japanese con:q>anies
have been the driving force behind the development and implementation of vertical diffusion
furnace technology and critical dimension (CD)
SEMs in semiconductor manufacturing. These two
relatively new categories of wafer fab equipment
emerged in the mid-19S0s as next-generation
replacement technologies for the more traditional
segments of horizontal diffusion furnaces and
optical CD measurement. Js^anese equipment companies have successfully maintained a dominant
position in vertical diffusion and CD SEM systems
since their introduction of this equqnnent to the
market.
TABLE 2
Ibming Point in Gaining Dominant Share
(Percentage of Worldwide Unit Shipments)
#
1982
1983
1984
1985
1986
1987
1988
1989
Steppers
X
X
X
X/O
X/O
Dry Strip
X
X
X
X
Ibbe LPCVD
X
X
X
Dry Etch
X
X
Sputter
X
X
1990
0
0
0
0
X
X
X/O
0
0
X
X
X
X/O
O
O
X
X
X
X
X
X/O
0
X
X
X
X
X
X
O
X • U.S. oompaiaei hdd *»•••"•"' diaie, O •E Jipanwe caapnin held domnuot ihaic
SauiGe: Dataqiieit (>^ail 1991)
01991 Dataquett Incoiponted April-ReptDdoctiao Prohibited
SBMMS Newdetten 1991 Eqinpment.—Wafer Fab Bquipment/tndiistiy Trendi
0010257
EQUIPMENT COMPANY OWNERSHIP—AT THE TURNING POINT
with the exception of one category. In the area of
dry etch, Japanese equipment con:q)anies held dominant share in unit shipments in 1990; however, on
a revenue basis, U.S. and Jiq)anese conripanies had
approximately equal share of the $683 million market. Jiq)anese etch conipanies offer a broad base of
equipment encompassing leading-edge tools as well
as older technology, whUe U.S. companies are
focused primarily on leading-edge systems. This
translates to an overall average seUiag price for
J^anese etch units that is lower than fhat for U.S.
etch systems.
Japanese companies established dominant
share in these equ^>ment categories through a combination of domestic technology developmoit and
acquisition of overseas companies and their technology. Powerhouse optics companies Nikon and
Canon chose to manufacture fuU stepper systems
rather than just supply sophisticated lens optics to
other stepper vendors. The Japanese stepper companies have attained a dominant share of die world
market today because of their en:q)hasis on equipment reliability in coihbination with leading-edge
lens design and manufacturing. Tokyo Election's
buyout of the TEL/Lam and TEiyrhermco joint
ventures added dry etch and tube LPCVD technology, originally developed at U.S. coiiq)anies, to an
already flourishing technology base in Japan. In
1990, dominant share in the sputter market shifted
from U.S. con^anies to the Japanese, ia large part
because of the acquisition of Materials Research
Corporation by Sony.
Table 2 ^ows that Jq)anese equipment companies have assumed dominant share of key segments of the wafer fab equipment market at Hoe
rate of iq)proximately one segment per year for die
0010ZS7
past five years. Clearly, the question is whether this
trend will continue, and if so, what will be the next
segments of the wafer fab equipment industry
where donunant share will transfer from one group
of regional conq)anies to another.
DATAQUEST PERSPECTIVE
The establishment of dominant share in key
segments of wafer fab equipment has propelled the
Japanese equipment industry into its number one
position. Dataquest believes that it is unlikely the
loss of dominant share by U.S. conqianies in key
segments of equipment can be regained without
tremendous ci^ital investment, particularly in those
segments with a turning point more tiian two years
old. One important aspect that our analysis suggests is a reevaluation of the role that SENIATECH
should play inrevitalizingsemiconductor manufacturing in &e United States. With limited resources
partly dependent on the whims of the congressional budget cycle, Dataquest believes that
SEMATECH should consider keeping its efforts
tightly focused on siqjporting only a few key areas.
These areas should be leading-edge equipment
technology and process development programs in
which U.S. conqianies still have an opportunity to
either successfully establish or capture a dominant
share position in the worldwide market. The time
for "catch up" in some segments of the wafer fab
equqiment business is over; the turning point is
past.
Peggy Marie Wood
01991 Dataqueit Inooipnated ApiU-Reproduetioii Pndiibited
SEMMS Newtletten 1991 EquiFmnit—Wafer Fab Equpmeat^ndiiitiy Trendi
mm
-.,- ^SiS^^?'
•cy"'
:..'.:t'-.'i-v
Dataoyest
I
acprnpanyaf
The Dun & Bradstrect c o r p o n t n n
fes^^?
i-f-t.^J-^^i/r
:.:!'•.•
^.bjiiyvW-Sf:
y-S''^^ 'V--k - f - i/^.W.^
>^:?;i?v'^-;5Y
Research Newsletter
T H E BRANSON/IPC AND GASONICS MERGER: A GROWTH OPTION
FOR SMALL COMPANIES WITHIN THE WAFER FAB EQUIPMENT
INDUSTRY?
SUMMARY
Gasonics' acqmsition of Branson/IPC from
Emerson Electric (St. Louis, Missouri) may result
in the emragence of a global equipment industry
player with a well-balanced portfolio of products
and the critical mass necessary for sustained
growth within the wafer fabrication equipment
industry. The combined entity, with revenue of
approximately $40 miUion in 1990, is better positioned to become a key player in the dry etch, dry
strip, and diffusion segments of the wafer fab
equipment market.
Dataquest believes that this merger illustrates
the trend toward rq)id consoUdation in the highly
fragmented North American equipment industry as
it struggles to cope with the escalating costs of new
product development and market penetration in an
increasingly globalized industry. Emerson Electric's
divestiture of Branson/IPC may also indicate ihat
many diversified North American industrial conglomerates do not wish to handle the volatile,
technology-driven nature of the fragmented semiconductor equipment industry.
PRODUCTS AND MARKETS SERVED BY
GASONICS AND BRANSON/IPC
Table 1 displays Dataquest's estimates for the
1990 wafer fabrication equqnnent market revenue
of Gasonics and Branson^PC in the dry strip, dry
etch, and diffusion market segments. The combined
entity has good synergy in its (ty etch, dry strip,
and diffusion equipment market presence. Branson/
IPC specializes in RF-based plasma technology,
while Gasonics specializes in microwave-based
plasma technologies. Dataquest beheves that semiconductor manufacturers wlU use a combination of
single-wafer and batch dry str^ processes utilizing
RF-plasma or microwave plasma technologies for
various device mask levels. The combined expertise of Gasonics and Branson/IPC in RF/microwave
strip technologies will enable the companies to
offer a one-stop worldwide solution to dry strip
market needs. The merged entity will have a significant presence in all key wafer fabrication
regional equipment madcets.
In addition, Branson/IPC's growing line of
industrial RF-plasma cleaning systems will provide
TABUE 1
Branson/IPC and Gasonics Products and 1990 Market Shares
(Millions of Dollars)
1990
Revenue
Compan;
Products
Branson/IPC
L3200, L3300
Series 9000
L2200
RF plasma/single-wafer dry strip
RF plasma/batch-barrel dry strip
RF plasma/single-wafer isotropic etch
10.0
4.5
3.5
Gasonics
Aura Series
AE2001
mPOx Series
Microwav^single-wafer dry strip
Microwave/single-wafer isotropic etch
High-pressure oxidation/diftusion
8.2
1.5
8.0
Market Segment
Source: Dataquest {AptH 1991)
0 1 9 9 1 Dataquett Incoipoiated April—Reproduction Prohibited
SEMMS Bquipnnnt—Wafer Fab Eqaipment/lDduttiy Tiendi
0010094
The corUeni cfthis report represents our interpretation and analysis of irt^muttion 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. Individual companies reported on and aruilyzed by Dataquest
may be clients of this and/or other Dataquest services. This in^rmation is not furnished in connection with a sale or of^r to sell securities or in connection with the solicitation t^an
ager to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, flvm time to time, have a Itmg or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE BRANSON/IPC AND GASONICS MERGER
a Stable revenue stream that wiU complement the
more volatile revenue streams from the semiconductor equipment operation. Branson/IPC's industrial cleaning systems are being e^lored for a wide
range of ^qppUcations in diverse areas such as the
aircraft, automotive, medical, and electrical
machinery industries.
COMPETITION IN THE MARKETS SERVED
BY BRANSON/IPC AND GASONICS
Figure 1 shows Dataquest's forecast for the
growth of die dry strip, isotropic dry etch, and
diffusion equipment markets between 1990 and
1995. The combined markets, which Branson/
Gasonics plan to address, will have a compound
annual growth rate of 14.3 percent, growing from
$497 million in 1990 to $968 million in 1995. The
dry strip market has assumed more importance
with the move to submicron processes. Issues such
as ionization damage due to dry strip plasma
exposure, resist removal difficulties due to ion
implant and reactive ion etch (RIE) high-energy
damage, and thin oxide charge buildup have
increased the size and importance of the dry strip
market.
Table 2 shows the worldwide 1990 dry strip
equipment company rankings. The Branson/Gasonics merger will catapult the new conq)any close to
the top of the worldwide dry strip company
rankings among such Japanese competitors as
Alcan Tech, Plasma Systems, Ramco, and Toltyo
Ohka Kogyo.
The isotropic dry etch market is a relatively
small niche within the much larger total dry etch
market. Gasonics and Branson/IPC hope to address
the standalone isotropic dry etch market as well as
the cluster tool module market, which will incorporate dry etch, dry strip, chemical vapor deposition (CVD), and physical vapor deposition (PVD)
processes. Competition, however, is heating up in
the isotropic dry etch market as leading dry etch
equipment companies develop their own versions
of isotropic etch modules for integration with their
cluster tools.
Gasonics' high-pressure HIPOx diffusion/
oxidation system occupies a unique niche within
the diffiision/oxidation equipment market. All other
diffusion equipment companies offer atmospheric
process solutions. Gasonics, with its high-pressure
system, owns a very small fraction of the overall
diffusion/oxidation equipment market. However,
Dataquest believes that high-pressure, lowtemperature diffusion/oxidation systems will find
increasing applications in submicron processes,
which attenq)t to minimize device exposure to high
temperatures for long periods of time. The combined company will have greater resources to
develop new applications and market the HIPOx
product family more aggressively worldwide.
FIGURE l
Worldwide Dry Strip, Isotropic Dry Etch, and Diffusion Equipment Market
Forecast (Million of Dollars)
Millions of Dollars
700630
$618
Sj
Dry Strip
560
490
Isotropla Dry Etch
M j ^
Diffusion
420
$322
350
280
210
140
70
0
$50
NXXvW
1990
1995
Source: Dataquest (April 1991)
0010094
01991 r>ataquest Incmponted ApiU-Rcpioduetioii Prahibited
SEMMS Eqaipmeiit—Wafer Fab Equipmmt/Indiistiy Trendi
i
THE BRANSON/IPC AND GASONICS MERGER
TABLE 2
Worldwide 1990 Dry Strip Equipment Company Rankings
(Millions of Dollars)
Company
Plasma Systems
Tokyo Ohka Kogyo
Ramco
Branson/IPC
Alcan Technology
Matrix
Gasonics
Ulvac
Tegal
Hitachi
Others
Worldwide Market Total 125.3 100.0
Revenue
Percent
Share
24.9
23.9
16.0
14.5
11.7
9.0
8.2
4.2
3.0
2.7
7.2
19.9
19.1
12.8
11.6
9.3
7.2
6.5
3.4
2.4
2.2
5.7
Note: Rankings are caleodar year lysiem levanw only; no ipaies or mvice aie included
Soutce: Dataqoeat (.^iiil 1991)
IMPLICATIONS OF THE MERGER FOR THE
FAB EQUIPMENT INDUSTRY
Dataquest believes that the Bianson/Gasonics
merger illustrates the $25 million to $30 million
annual revenue barrier that equipment coiiq)anies
have to surmount in order to generate a sufficirat
income stream to fund future product development
This becomes especially inqxntant in an industry
characterized by ever-^ortening design-in market
windows for future device generations, high R&D
costs, and high costs due to the need for complete
global customer support The North American
wafer fab equipment industry abounds in small,
entrcpieneural companies with creative products,
single-maiket focus, and limited R&D and global
expansion resources. Such small companies are
also constrained by their inability to raise cheap,
long-temi capital in the short-term-oriented North
American equity and debt markets.
m
The options for die survival of diese small,
entrepreneurial North American companies are
limited—seek a buyout or a cash infusion from a
larger partner and risk losing the flexible
entrepreneural edge and control over the technology or seek a marriage of equals with another
srnall entrepreneural con:q)any and attain critical
mass that way. Dataquest believes that the
marriage-among-equals option pursued by Branson/
IPC and Gasonics may have a good chance for
success within the wafer fab equipment industry.
The semiconductor equipment industry wiU closely
watch the combined Branson/Gasonics entity's
future market activities in order to gauge the longterm success of this growth strategy.
Krishna Shankar
Hie topics covered by SEMMS newsletters are selected for their general intoest to SEMMS clients, vbick indnde wafer fab equqiment
siqipliers, jcmicoaductcr materials coiiq>anies, and semiconductor device mamifoctiiiers. The topics selected iiwtiratc the broad range of research
diat is conducted in the SEMMS group. Clients, however, often have specific infonnation teqaitements fiiat eittKi go beyond tfie level of detail
contained in the newsletters or beyond die scope of what is nonnally published in the newsletters. In order to provide coiq>lete decision support
to onr clients, Dataquest has a consultiiig service available to handle these additional infonnation needs. Please call Stan Braederle at (408)
437-8272 or Joe Greoier at (408) 437-8206 to discuss your custom leqniiemeots.
01991 Dataquest Incoipoiated April-Reproductioa Piohilnted
SEMMS Equipmeat—Wafer Fab Eqnipnmt/Iiidustiy Treads
0010094
-^t'V^S^-''*'^
- * 4 - ^ - %'^""?f4:
^
s^t
^
Dataqyest
acotfiKinvof
BrvfaucclcorpontaDn
t itKlMn&E
;T*.r,:-<^:^^-ri.i,
BPf*
•fsyifiCfl'v.'^-;
^vj-'^j
•
Research Newsletter
1990 WAFER FAB EQUIPMENT MARKET: JAPAN ACCOUNTS FOR
MORE THAN HALF OF WORLD'S SUPPLY AND DEMAND
The Japanese market for semiconductor wafer
fab equipment accounted for more than 50 percent
of the worldwide market in 1990. At the same
time, Japanese equipment companies supplied more
than half of the world's wafer fab equipment needs.
These are just a few of the results from the annual
survey of flie wafer fab equipment industry recently
completed by Dataquest's Semiconductor Equipment, Manufacturing, and Materials Service
(SEMMS).
1990 IN REVIEW
In 1990, the world market for wafer fab
equipment was $5.8 biUion, down 3.1 percent from
its 1989 level of $6.0 billion. While Japan and
Europe both enjoyed modest growth of 5.2 and
5.3 percent, respectively. North America was down
3.8 percent. TTie Asia/Pacific market for wafer
fab equipment was hit especially hard in 1990
by the reduced spending levels of the major
Korean manufacturers, and it was further exacerbated by the woes of the Taiwanese financial market. These factors contributed to a heady decline of
37.3 percent in spending for wafer fab equipment
in Asia/Pacific in 1990.
Modest growth in Japan coupled with contracting markets in North Ainerica and Asia/Pacific
contributed to a 3.9 percent increase in 1990
regional market share for Japan. As shown in
Figure 1, Japan accounted for more than half
of the world's demand for wafer fab equipment
with 50.6 percent of the $5.8 million world market.
FIGURE 1
1990 Wafer Fab Equipment Market by Region
(Percentage of Dollars)
$5.8 Billion
Note: Parts do not add to 100% because of rounding.
Source: Dataquest (April 1991)
®1991 Dataquest Lacoipotated April-Reproduction Prohibited
SEMMS Newsletters 1991 Equipment—Wafer Fab Equipment/Industry Trends
0010059
The content of this report represents our inteijjretation and analysis afir^nnation 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 rrtaterial provided to us in confidence by our clierjts Individual companies reported on and analysed by Dataquest
may be clients of this and/or other Dataquest services. This irtfitrmation is notfiimtshed in cormection with a sale or o^r to sell securities or in connection with the solicitation of an
<^r to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
1990 WAFER FAB EQUIPMENT MARKET: JAPAN ACCOUNTS FOR MORE THAN HALF OF WORLD'S SUPPLY AND DEMAND
Europe, also propelled by modest growth, exhibited
a 1 percentage point increase in 1990 in its regional
share of the world equipment market. North
America maintained the same regional percentage
share in 1990 as in 1989, while Asia/Pacific
declined almost 5 percentage points.
Throughout the 1980s, Japanese equipment
companies have steadily gained market share in the
wafer fab equipment industry; 1990 was no exception. Japanese equipment company share of the
worldwide wafer fab equipment market was
51.1 percent in 1990, an increase of 2.8 percentage points from its 1989 level. Japanese
equipment companies increased share in all
regions of the world except Asia/Pacific. In particular, Japanese equipment vendors garnered a significant increase of market share in Europe in 1990
and now account for almost one-fourth of aU wafer
fab equipment spending in that region. The
increase of Japanese equipment company share in
0010039
Europe is a direct result of the European fab expansion plans of Japanese semiconductor manufacturers that tend to rely on J^anese equipment and
materials companies to supply their new overseas
fabs.
FUTURE ANALYSIS
Detailed analysis of the wafer fab equipment
market by equipment category, regional emphasis,
and company ownership will be discussed in a
series of newsletters by SEMMS analysts during
the next several monflis. In the meantime, the
SEMMS wafer fab equipment database, including
detailed company market share estimates by region,
by vendor, and by equipment category, is available
immediately to SEMMS clients via electronic
dehvery of Dataquest's On-Line Service.
Peggy Marie Wood
01991 Dataquest Incorporated April-Reproduction Prohibited
SEMMS Newsletters 1991 Eqaipment—Wafer Fab Equipment/Industty Trends
pilS
mM^'
DataQuest
) acsmrarnol
I thcIMn&EHadsUcetcorporsiiion
Research Newsletter
STEPPER EQUIPMENT MARKET—1990 MARKET IN REVIEW
SUMMARY
Worldwide stepper shipments totaled
775 units in 1990, down almost 19 percent from
the 1989 level of 954 steppers. The 1990 stepper
market, measured on a revenue basis, was
$1,067 million, down only 10 percent from its
1989 level. The difference in percentage change
between units and revenue clearly reflects the pervasive trend of increasing average selling prices
(ASPs) for advanced wafer processing equipment.
Higher ASPs for steppers are being driven by new
lenses with high niunerical apertures and wide
fields, as well as a continuing shift in the stepper
product mix toward i-line stepper systems.
Dataquest expects worldwide stepper shipments in
1991 to be 800 units, up a modest 3 percent from
1990, while stepper revenue is expected to be up
13 percent to $1,210 noillion.
REGIONAL MARKETS
AB regions of the world had reduced stepper
shipments in 1990 compared with 1989, with the
exception of Europe, which experienced modest
growth from 86 units in 1989 to 94 in 1990. Asia/
Pacific-Rest of World (ROW), in particular,
experienced a severe decline in stepper shipments
from 157 units in 1989 to only 73 in 1990. This
drastic decline in stepper shipments was of a magnitude similar to the decline observed in other
equipmrait segments as semiconductor manufacturers in South Korea and Taiwan reevaluated their
capital spending plans in light of softening DRAM
prices, die specter of overcapacity, and faltering
finandal markets.
Japan represents the largest regional madcet
for wrfer steppers and, with 410 steppers,
accounted for 53 percent of worldwide stepper
shipments in 1990 (see Figure 1). Activity in i-hne
stepper lithography has escalated dramatically in
Japan in the past several years as DRAM manufacturers have settled on their 16Mb and 64Mb lithoffs^Q strategies. I-hne unks accounted for only
6 percent of the product mix in Japan in 1988, yet
by 1990, i-line systems represented 37 percent of
all steppers sold in this region of the world. Phase
shift masks, now in development, offer the prospect
of extending existing i-line lens technology to
smaller line geometries without any significant loss
in depth of focus. Dataquest anticipates that if
phase shift mask technology proves both successful
and cost-effective, i-hne steppers will be used for
DRAM processing well into the 256Mb generation.
North American stepper shipments in 1990
totaled 198 units, or about one-fourfh of the world
market. North America traditionally has heesa the
strongest market for i-line steppers supported by a
well-established vendor base for both steppers and
photoresist. I-line systems represented 42 percent
of the North American stepper mix in 1990; however, on an absolute basis, i-line shipment levels in
North America ranked second behind Japan. I-line
lithography in Europe and AsiaA'acific-ROW has
yet to be embraced widi the same level of activity
as in J^an and North America. I-line shqnnents in
Europe and Asia/Pacific-ROW were 28 and
25 percent, respectively, of total shipments. In 19SK)
overall, i-iine units represented 36 percent of
worldwide stepper shipments, up from only
12 percent of the stcfiper product mix two years
earlier.
REGIONAL OWNERSHIP
The stepper equipment market continues to be
dominated by Japanese suppliers (see Figure 1).
Together, Nikon, Canon, and Hitachi controlled
close to 80 percent of 1990 worldwide stepper
shipments and over 98 percent of stepper shipments
in their home market of Japan. Historically, all of
OI991 Ditaqoest Incoiponted Aptil-Reproductian Prohibited
SEMMS Newiletters Equipment—Lithognphy
0010222
The content of this report represents our iraerpretmion and analysis of information generally amiloble to the public or lileased by responsible individuals in the subject companies but
IS not guaranteed as to accuracy or completeness It does not contain material provided lo us in confidence by our clients. Individual companies reported on and analyzed by Dat^uest
may be clients cflhis and/or other Dataquest services This information is not fitmished in conneclion with a sale or o^r lo sell securities or in connection with the solicitation
of an
of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of theirfitmiliesmay. from time to time, have a long or short position in the
securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
STEPPER EQUIPMENT MARKET—1990 MARKET IN REVIEW
FIGURE 1
1990 Stepper Regional Markets and Ownership
(Units)
North American
Companies
100
12.9%
European
Companies
58
7.5%
Markets
Total = 775 Units
Ownership
Total = 775 Units
Note: Pie chart segments do not total 100 percent because of rounding.
Source: Dataquest (April 1991)
Hitachi's stepper shipments have been in Japan;
however, 1990 marked the first year for Hitachi
stepper exports: a handful of systems were shq)ped
to Hitachi's semiconductor operations in the United
States. In contrast to Hitachi, Nikon and Canon
have been aggressive in estabUshing a strong presence outside of Japan. Together, these two companies in 1990 represented 43 percent of stepper
shipments in North America, 74 percent of stepper
slupments in Europe, and 73 percent of stepper
shqnnents in Asia/Pacific-ROW.
COMPANY RANKING
As shown in Table 1, Nikon maintained its
number one ranking in the stepper equipment market in 1990, accounting for one-half of the world
market with 384 steppers. E>ataquest estimates that
approximately 30 percent of Nikon's stepper mix
were i-line systems in 1990, a significant change in
product mix cotnpsFed with 198S whoi all of
Nikon's stepper shipments were g-Une systems.
Canon ranked second in the stepper maket in
1990, a strong position for a vendor that was not
yet shipping i-line systems. In mid-1990, the company announced its entry into the i-Une maiket widi
its model FPA-2000il, ^ e first announcement of an
i-line stepper with a wide-field lens (20mm x
20mm). Shipments of Canon's new i-line system
are expected to begin in 1991.
0010222
Hitachi ranked third in stepper shqnnents in
1990 with an estimated 83 units, essentially all of
which were i-line systems. Dataquest estimates that
iq)proximately 30 percent of Hitachi's stepper shipments are for internal use at Hitachi semiconductor
facilities. One of the advantages that Hitachi enjoys
in process equipment development is the close ties
it has to its sCTuconductor manufacturing parent
corporation. At the same time, this can be viewed
as a disadvantage in the marketplace because some
semiconductor manufacturers are reluctant to buy
Hitachi equipment and thus expose their clesm
room and processing activities to Hitachi engineers.
ASM Lithognq)hy, the only European vendor
of optical steppers, continues to be a shining star in
TABLE 1
1990 Worldwide Stepper Company Rankings
Company
Nikcn
Canon
Hitachi
ASM Lithogiaidiy
GCA
Ultiatech
SVG IJthograidiy
Total
Units
384
150
83
58
52
39
9
775
Market
Share (%)
49.5
19.4
10.7
7.5
6.7
5.0
Samce: Dttaqimt (>^>iil 1991)
01991 DaUqueR Licoiponted Ajtrit-Reptt.
SEMMS Newdmen Bqoipm.
1
STEPPER EQUIPMENT MARKET—1990 MARKET IN REVIEW
the European wafer fab equipment industry. Over
the past several years, the company has successfully positioned itself as a leading vendor of
advanced i-line systems. At the SPIE Syn^sium
on Microlithography in March 1991, ASM lithography introduced a new wide-field i-line stepper
with sub-0.5-micron capability. Dataquest believes
that this new system places ASM in a strong
coinpetitive position in this highly technical segment of the wafer fab equipment industry. ASM
lidiography ranked fourth in stepper shipments in
1990 witii an estimated 58 units.
GCA ranked fifOi in 1990 stepper shipments
with an estimated 52 systems, die majority of
which were i-line systems. GCA was the first vendor to start i-line stepper shipments back in 1985.
At SPm in March 1991, GCA described its most
advanced systems for i-line and excimer laser
lithography. Its new family of XLS steppers was
developed in conjunction with Sematech and its
member companies. The relationship with
Sematech and its member conq)anies has been
inqKfftant to GCA because more than three-fourths
of its unit shipments the last several years have
been in its home market of the United States.
Qearly, 1990 was a year of uncertainty for
Ultratech. In May 1990, General Signal, the parent
corporation of GCA and Ultratech, announced that
it would consolidate the operations of its regionally
dispersed stepper businesses. (GCA is in
Massachusetts; Ultratech is in California.) However, before fbis decision was put into motion, a
management buyout of Ultratech was announced.
By the end of summer 1990, however, the plans for
the buyout had fallen ^ a r t over disagreements
regarding financing. At the end of 1S>90, General
Signal was back where it began the year, the parent
corporation of two stepper companies located in
different parts of the United States and pursuing
different stepper technologies. Dataquest believes
that the uncertainty regarding Ulttatech's fiitute,
coupled with the widespread acceptance of reduction lithognq)hy as the de facto standard in the
industry, was responsible for the erosion of the
company's position in the maricet.
SVG Lithography (SVGL) was a new name
in the stepper market in 1990. SVGL was established as a subsidiary of SUicon Valley Groiq>
(SVG) after it acquired the former optical lithography group of Peikin-Elmer in May 1990, Wilh the
acquisition, SVGL acquired the complex, advanced
lithography technology known as "step-and-scan"
because of its combined csqf>ability of both projection and step-and-iepeat lithography. Dataquest
estimates that in calendar year 1990, Peikin-Ehner
ei991 Dataqueit Incaipanted Apiii-Reprodiictiaa Prohibited
SEMMS Newilettm Bquipment—IMmgupbf
and SVGL had combined shipments of nine
Micrascan systems, of which all but one system
went to IBM for its advanced DRAM production.
DATAQUEST PERSPECTIVE
The stepper market is the largest single segment of the wafer fab equipment industry, accounting for $1,067 million, or 18 percent of the 1990
world maritet of $5,813 million. Stepper technology is a fundamental component of advanced
device processing that allows for the continuing
reduction of line geometries. Tliis means that stepper vendors regularly face huge R&D investments
to stay on die cutting edge of new lens and stepper
system development Wth the depaiHire of American Semiconductor Equipment Technologies
( A S E T ) in 1989, only seven conq>anies remain in
this market segment, and the barriers to entry are
sufficiently high that it is unlikely that new companies wiU enter the maiket. The question rernatns of
whether the smaller stepper con^uinies have suf&cient critical mass to manage the long-term pressures of global sales, service, and siqiport and
invest a sufficient level of c!q)ital in new product
development.
Clearly, Silicon Valley Group is learning firsthand the rigors of supporting an advanced lithography equipment group. In the quarter ending
December 31, 1990, SVG repotted net income of
$81,000 on sales of $60,9 million, in sharp contrast
to the average quarterly net income of $1.1 million
reported to its stockholders over the last nine years,
SVG recently held public offerings for its stock as
one means to geneiate additional coital for its
activities. Dataquest believes that future growth for
GCA and Ultratech wiU depend on General
Signal's attitude concerning its continuing participation in the semiconductor wafer fab equipment
industry. General Signal has spun out several of its
wafer fab equipment groups over (he last several
years and emertained the concept of a management
buyout of Ultratech last summer. Today, Ultratech
is stiU part of the General Signal family, meaning
that the two stepper business units compete against
each other for bolh corpotate R&D dollars and
ocdras in the mackxtplace.
Althou^ only a handful of con^Ktnies participate in the stepper market, it will contuute to
CEqpture the interest and imagination of the industry
because it represents the leading edge of technolo*
gy. SuccessAil perfocmance in thu segment brings
with it significant financial rewards.
Peggy Marie Wood (San Jose)
Kunio Achiwa (Tokyo)
0010222
DataQuest
g i n auniwMiiyaf
IJO
TticDundefKblnctCorpontKin
Research Newsletter
W E T PROCESSING EQUIPMENT: I 9 9 0 MARKET IN REVIEW
SUMMARY
The worldwide wet processing equipment
market totaled $350.3 million ia 1990, a decrease
of 1.2 percent over 1989 sales levels. Wet processing equipment comprises five categories: integrated
wet systems, manual benches, rinsers/dryers,
add/solvent processors, and megasonic cleaners.
The year-to-year decline in wet process revenue is
attributable to manual benches, rinser/dryers, and
acid/solvent processcxrs. Integrated wet processing
systems and megasonic sales bodi grew in 1990 in
spite of a weak semiconductor ciq)ital equipnent
market.
REGIONAL MARKETS
Geogr^hically, Japan is the world's largest
wet processing equipment market Sales in Japan
totaled $222.8 million in 1990 (see Figure 1). The
size of the Japanese market can be e^iplained by the
sheer size of the device production capacity of
Jj^anese fab lines and by the device manufacturers' determination to p i ^ wet processing technology to the limit
North American sales, $71.8 million in 1990,
are much smaller than J:q)anese sales for two
reasons. First, U.S. manufacturers on avenge spend
considerably less for wet processing equ^nnent per
fab line than do J^>anese manufacturers because
they have been slow to automate the wet processing area. Second, U.S. semiconductor manufacturers have focused their efforts on dry processing
rather diao wet processing.
Sales in European and Asia/Pacific-Rest of
World (ROW) are smaller, reflecting the semiconductor production c£^>acity in these regions. In
1990, sales of wet processing equ^wnent totaled
$30.5 million in Europe and $26.2 million in Asia/
Pacific-ROW. Dataquest is forecasting the Asia/
Pacific-ROW region to have the fastest-growing
demand for wet processing equipment during the
next five years. Jq)anese equipment vendors are
expected to benefit the most fiom this growtii,
given their strength in ttus equipment segment and
their proximity to die market
COMPANY RANKINGS
Japanese con^anies Dainqjpon Screen, Kaijo
Doiki, Sankyo Engineering, and Sugai lead the
worldwide ranking in sales of wet processing
equqnnent (see Table 1). Jq)an's dominance ia wet
processing equipment (see Figure 1) is attributable
to sales of integrated wet bench systems, which
make up 68 pocent of tiie total wet processing
market This leadership position is unlikely to be
relinquished any time soon.
A few U.S. con:q)anies have strong positions
in the wet processing equqnnent market. FSI International is the sixth-laigest worldwide vendor of
wet processing equqnnent, holding a particularly
strong position in the acid/solvent processor market Submicron Systrans, a fluee-year-old U.S.based company, has become the largest supplier of
integrated wet benches in the North American market Athens Corporation and Alameda Instruments
have pioneering acid reprocessing technology,
which has the potential to grow rapidly during the
next 10 years.
DATAQUEST PERSPECTIVE
The wet processing equipment market has
grown at a 21.5 percent conq>ound annual growth
rate (CAGR) from 1986 to 1990. This growth
mirrors the overall semiconductor capital equq>ment industry, which grew 21 percent in the same
five-year period.
01991 Dataqueit Incoiponted Aptil-Reproductioii Prohibiteii
SEMMS Newiletten 1991 Equipment—Etcb/Qean
0010240
Th£ content of this report represents our irjerpretation ami analysis of inprmation 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. Individual companies reported on and analy^ by Dataquest
may be clients of this and/or other Dataquest services This information is not famished in conrwction with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities This firm and its parent and/or their cfficers, stoddiolders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
WET PROCESSING EQUIPMENT: 1990 MARKET IN REVIEW
FIGURE 1
1990 Wet Processing Equipment Regional Markets and Ownership
(Revenue in Millions of Dollars)
Asla/PaclfloROW
$25.2
7%
European
Companies
$2.2 1 %
Ownership
Total = $350.3 Million
Markets
Total = $350.3 Million
Source: Dataquest (April 1991)
TABLE 1
TABLE 2
1990 Worldwide Wet Processing Company
Rankings (Millions of Dollars)
1990 Worldwide Wet Processing Equipment
Segments (Millions of Dollars)
Market
Revenue Share (%)
Sales
CAGR (%)
1986-1990
238.2
32.0
Dainippon Scieen
53.9
15.4
Integrated Wet Systems
Sankyo Engineering
38.7
11.0
Manual Wet Systems
39.6
7.5
Kaijo Denki
37.1
10.6
Rinser/Dryas
38.1
8.1
Sugai
35.7
10.2
Acid/Solvent Processors
25.6
4.5
Shimada
16.4
4.7
FSI Intematicsial
162
4.6
8.8
350.3
29.8
21.5
Verteq
12.9
3.7
Marawa
12.3
3.5
Semitool
11.5
3.3
Enya
10.8
3.1
104.8
350.3
29.9
100.0
Megasonic Qeaners
Total
Somce: Datequeit (^nl 1991)
Other
Worldwide Maiket Total
Note: Hie table ibowi cdcodar year 1990 fyitnii icvcDQcj fpnss n o
•ervice not nirhKM
Souice: Dtfaqoert (,Aftil 1991)
0010240
However, two segments within the wet
processing equipment maiket stand out in tenns of
growth (see Tkble 2). The integriUed wet piocessi&g
equipment grew at a blistering pace with a
32 percent CAGR from 1986 to 1990, and megasonic cleaners closely followed widi a 29.8 CAGR
since 1986. The other segments, manual wet
benches,rinser/dryers,and acid/solvent processors,
fell behind both liie wet processing equqnnent and
the overall semiconductor equipmoit growth rate.
ei991 DaUqueit Jhcorpoiated Ai»il-4te|nDdiieti(«i Pndubited
SEMMS Newiletten 1991 Bquifment—Eteb/aetn
WET PROCESSING EQUIPMENT: 1990 MARKET IN REVIEW
The five-year CAGR for these three segments
was 7.5 percent, 8.1 percent, and 4.5 percent,
respectively.
Although wet processing technology is well
defined, Dataquest expects this equipment to
evolve as the level of automation increases in
semiconductor fabs. Moreover, the size of the
market, $350.3 million in 1990, is expected to
attract some of the larger semiconductor equipment
vendors that currently do not participate in this
area.
Mark FitzGerald (San Jose)
Kunio Achiwa (Tokyo)
Ibe toiHcs covered by SEMMS newsletters ate selected for dieir general interest to SEMMS clients, wUch include wafer &b eqo^nnent
siqiplieis, semicondactor matwials conquniies, and semiconductor device mannfRCtmers. The topics sdected indicate die broad range of research
diat is conducted in the SEMMS groiq). Clieots, however, often have qtedfic infcmnation requiremeots Out eidier go beyond die level of detail
contained in die newsletters or beyond die scope of what is noimalfy published in die newsletters. In order to provide conplele dedsiom siqipoit
to our clients, Dataqoest has a consulting service available to bandle dkese additional infonnation needs. Please call Stan Binederle at (408)
437-8272 or Joe Oienier at (408) 437-8206 to discuss your custom requirements.
01991 Dauquett Incofpcnued Apiii-Repioduetiaii PrahiUted
SEMMS Newiletten 1991 Equipmeiil-^tch/amD
0010240
Dataquest
afonnnnvaf
t Ttic
Dun STBradstrcct Corporalion
Research Newsletter
DRY ETCH EQUIPMENT: 1990 MARKET IN REVIEW
SUMMARY
Dataquest believes that the worldwide 1990
dry etch equipment market essentially remained flat
between 1989 and 1990. The worldwide 1990 dry
etch market was $683 million. Although die worldwide market was flat, Japan's dry etch market, the
world's largest, increased 9 percoit to $360 million
in 1990. Japanese device manufacturers continued
to invest strategically in leading-edge 200mm sub-.
micron fabs in spite of a sluggish chq> demand
environment. The North American dry etch market
remained flat in 1990, while the European market
grew strongly because of the influx of North
American and Japanese capital investment into
European fabs. The Asia/Pacific-Rest of World
(ROW) market contracted severely in 1990 because
of political and economic uncertainities in that
region.
North American dry etch conq>anies, with
49 percent of the worldwide dry etch market, clung
to a slim 3 percent lead over Ji^anese dry etch
conipanies, which advanced to owning 46 percent
of the market. Ji^anese dry etch conipanies gained
market share for two main reasons: their homefield advantage of operating in the largest, most
technologically demanding market and their early
transition to single-wafer plasma-enhanced reactive
ion etch (RIE) and microwave/electron cyclotron
resonance (ECR) etch technologies. Nordi American companies are responding aggressively by
speedy globalization and rspvUy bringing innovative plasma source technology to market.
REGIONAL MARKETS
Figure 1 shows the relative proportions of
the 1990 worldwide dry etch market segmented
by region and con:q>any ownership. The North
American market made up 27 percent of the 1990
worldwide dry etch market U.S. semiconductor
companies have focused on high-value-added,
high-margin products such as microprocessors,
VLSI logic chip sets, and programmable logic
devices. They have largely retreated from highvolume, low-margin products such as DRAMs and
low-end gate arrays. Hence, key high-volume segments of the Nordi American wafer fab equqnnent
market such as steppers and dry etch equipment
have not grown as rapidly as have the corresponding Ji^anese wafer fab market segments. It should
also be noted that offshore Jq)anese fabs represent
an increasing portion of the North American dry
etch market. Leading North American device companies hope to increase their production of leadingedge high-volume microprocessors, high-end
ASICs, and chq) sets in onter to stay on the technology and volume learning curve. Tlius, the North
American dry etch market may expand more
npidly in die future.
The Japanese dry etch equipment market
grewfrom49 percent of the $669 million 1989 dry
etch market to 53 percent of tiie $683 million 1990
market The nimiber of mask levels in DRAM
processes continues to increase dramatically with
each new gener^ion. For example, the 16Mb
DRAM process may use 24 to 28 mask levels.
Japan, as the leading producer of advanced
DRAMs, is a natural technology driver for the
development of high-volume dry etch equqnnent
Dry etch equipment average selling prices have
also increased dramatically because of increasingly
complex submicton etch requirements. J{q)an has
also increased its market presence in other higjivolume device markets such as gate arrays and
other ASICs, PC chip sets, and foundry products
for fabless semiconductor compaiues. Dataquest
believes that the combination of high-volume
production and leading edge technology is a major
catalyst in accelerating the growth of core Japanese
wafer fab equipment markets such as the stepper
and dry etch markets.
01991 Dataqueit Incoiponted Apiit.ReprDdiietion Prohibited
SEMMS Newtletten 1991 Equipmeiil—Bteh/aem
0010063
The content qfthis report represents our interpretation and analysis of infiyrmation 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- Individual companies reported on and analyTed by Dataquest
may be clients of this and/or other Dataquest services This information is not fiimished in ccmnectutn with a sale or offer to sell securities or in connection with the solicitation of an
o^rlo buy securities This firm and its parera and/or their officers, stockholders, or members of ^ir families may, fiom time to time, have a Umg or short position in the securities
meraioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
DRY ETCH EQUIPMENT: 1990 MARKET IN REVIEW
FIGURE 1
1990 Regional Dry Etch Markets and Ownership
(Millions of Dollars)
European
Companies
$19.9 3%
Asla/PaclflcROV^ $44 6
Markets
Total = $683 Million
Joint
' Ventures
$14.1 2%
Ownership
Total = $683 Million
Source: Dataquest (April 1991)
The European dry etch market increased from
13 percent of the 1989 market to 14 percent of the
1990 market A major portion of European market
growth was due to the influx of offshore European
fabs built by North American and Jq)anese cotiq)anies such as Fujitsu, Hitachi, Mitsubishi, and TI.
Typically, these offshore European fabs were clones
of existing Nordi American and Japanese fabs.
Thus, a major pardon of the European dry etch
market was open only to global diy etch vendors
that had a significant presence in all three major
semiconductor manufacturing regions of the world.
Asia/Pacific-ROW dry etch market share contracted frxmi 12 perc^t of the 1989 market to
6 percent of the 1990 market. Factors such as
severely depressed DRAM prices, excessive PC
chq) set con^titiaii, stock maiket crashes, and
political/economic instability led to severe constraints on cf^ital spending by Asia/Pacific-ROW
device compames. Dataquest beUeves that the Asia/
Pacific-ROW dry etch market may inqjrovc in
1991 as capital spending is resumed for future key
production technologies such as the 16Mb DRAM
and submicron foundries.
0010063
REGIONAL OWNERSHIP
Hgure 1 shows tiie regional ownership of the
1S>90 worldwide dry etch markets. North Ameiican
companies conttniu»l to lead the worldwide dty
etch market, albeit by a small margin. North
Ameiican dry etch companies saw their market
share erode from 58 percent of the 1989 market to
49 percent of the 1990 market Factors that caused
this maiket share erosion included a major transition in product generation by Applied Materials,
the leading North America-based dry etch
conqtany; severe global competition frtnn I^ianese dry etch companies; and a rehuively flat 19SK)
North American dry etch equqiment maiket.
Japanese dry etch conq>anies' maiket share
inoeased from 39 percent of the 1989 market to
46 percent of the 1990 market. Factors ^lat caused
Ji^anese companies to gain matket share included
partic^iation in the large domestic Jf^>anese market
close strategic partnershq)s with leading Japanese
device manufacturers, cmcial technology leadei^h^
in single-wafer microwave/ECR dry etch, and
heavy emphasis on customer support and service.
ei991 Diuqueit bcoiiionted Ai)ril-4leproduetioii Prohibited
SEMMS Newdenen 1991 Equtpmeot—^Btdi/Clean
DRY ETCH EQUIPMENT: 1990 MARKET IN REVIEW
Japanese diy etch con^anies have also successfully
globalized operations to serve their customers
through joint ventures and regional subsidiaries.
Joint-venture companies such as Varian/TEL
had strong dry etch revenue growth in 1990. Dataquest believes that joint ventures repres^it a new,
viable globalization business strategy for quick
market penetration using the local joint venture
parmer's installed base of customer support and
service capabilities.
COMPANY RANKINGS
Table 1 shows the rankings of the major
players by their worldwide revenue in the 1990 dry
etch equipment market. Six North American
conq}anies, eight Jq>anese con^anies, one European conq)any, and one joint-venture conmany
made up the top dry etch 1990 company ranldngs.
Applied Materials, with 25.6 percent of the
1990 dry etch market, retained its market leadership posilioii. Applied's batch hexode 8000 Searies
systems and the PE5000 multichambcr system enabled the company to lead the dry etch market
However, Apphed's lead over the coinpedtion in
this market has narrowed in the last few years.
Apphed's dry etch revenue shrank IS p»x:ent from
$208 million in 1989 to $175 million in 1990 in an
essentially flat market environment. Applied lost
market share in 1990 because of its late transition
from the batch hexode 8000 family to the singlewafer PE5000 magnetically enhanced RIE technology. Conipanies that compete with Applied in the
TABLE 1
1990 Worldwide Dry Etch Equipment Company Rankings
(Revenue in Millions of Dollars)
Revenue
Market
Share (%)
175.0
25.6
RIE, MERIE
99.2
14.5
Plasma, RIE
Hitachi
91.5
13.4
Miciowave/ECR, RIE
Lam Research
85.7
12.5
Plasma, RIE
Sumitomo Metals
33.0
4.8
Microwave^ECR, RIE
Tegal
31.0
4.5
Plasma, RIE, triode
Anelva
29.4
4.3
Microwave/ECR, RIE
Tdsuda
24.1
3.5
RIE
Drytek
22.0
3.2
RIE, triode
Tokyo CMika Kogyo
18.9
2.8
Microwave/ECR, plasma
Flasma-Thenn
18.0
2.6
Phksma, RIE
Electiotech
17.5
2.6
RIE, triode
Varian/TEL
11.6
1.7
Plasma, RIE
Ulvac
6.6
1.0
RIE
MRC/Sony
4.7
0.7
MERIE
Bninson/IPC
3.5
0.5
Plasma
11.7
683.4
1.7
100.0
Company
Applied Materials
TQI^O
Election
Others
Worldwide Market Total
Market Segment
Note: Tbe taUe ifaDwi cdendar ynr 1990 •yilemi levenue; qnies nd ierrice ne not jnehided.
Source: Dataqoeit (April 1991)
01991 Dauquect Inecnponlsd April-Repioductioa Prahibited
SBMMS Newtletten 1991 Eqmpmmt—Etch/Cleas
0010063
DRY ETCH EQUIPMENT: 1990 MARKET IN REVIEW
dry etch market forged ahead earlier into the
single-wafer plasma-assisted RIE and microwave/
ECR dry etch technology.
The phenomenal success of the Precision
8000 technology prompted Applied to use a
strategy of incremental improvement in its proven
batch hexode technology rather than a strategy of
revolutionary single-wafer dry etch technology
using innovative plasma sources. In the domestic
North American market, Applied encountered vigorous competition from Lam Research Corporation's (LRC's) highly successfid Rainbow Series
etch systems. AppUed also encountered intense
competition in Japan from companies such as
Hitachi and TEL, which have invested heavily in
R&D, joint product development with customers,
and strategic customer support capability.
Dataquest beUeves that Applied is refocusing
its efforts in the dry etch market by emphasizing its
competitive strengths in specific segments such as
the metal etch, polycide etch, and silicon trench
etch markets. Applied's strengths in dielectric and
metal CVD and PVD will also probably result in
synergistic integration products based on the
PESOOO/Endura 5500 mainframe cluster tools.
TEL, ranked number two, captured 14.5 percent of ihe market with $99 miUion in 1990 revenue. TEL has successfully penetrated the large
Jq>anese market through its single-wafer plasmaassisted RIE technology. TEL has built up an
impressive reputation for customer support and
product reliability and is aggressively globalizing
its business through the Varian/TEL joint venture.
TEL has done well in absorbing LRC's
AutoEtch plasma etch technology from the prior
TEL/Lam joint venture and enhancing the product
line to meet the Jq)anese dry etch market's requirements. Subsequently, TEL has developed its own
proprietary plasma-enhanced RIE product line to
meet the needs of the 4Mb DRAM generation. TEL
recently introduced a new series of 200mm MERIE
systems to serve the needs of flie 16Mb/64Mb
gener^on. TEL's systems are wMely used in poly/
polycide gate, nilnde, and oxide dry etch applications. Dataquest believes that TEL's independent
corporate position regarding the major Ji^anese
business groups, together with its close strategic
partnershq>s with key global device manufacturers,
will he^ maintain its leadership position within the
market.
Hitachi sqjpears to have done phenomenally
well in 1990. Its 1990 dry etch revenue of
$91.5 million catapulted it to third place with
0010063
13.4 percent of the market. Hitachi's microwave/
ECR single-wafer systems continued to successfully penetrate the metal and poly^lycide dry etch
market Hitachi's metal etch system, widi its postetch treatment involving dry strip and in-situ wet
clean, was particularly successful in the fastgrowing metal etch market. Hitachi is at the forefront of the Japanese market movement toward
e}q)loiting the high ionization efficiencies, high
etch rates, high selectivity, and low ion bombardment of microwave/ECR-type single-wafer dry etch
technologies. Hitachi recently began emphasizing
its global equipment marketing capabilities by
e}^>anding aggressively in the international market.
LRC c^tured 12.5 percent of the 1990 dry
etch maiket. LRC would be the second-ranlrad
1990 dry etch conq)any with revenue of approximately $110 million if Sumitomo Metals' resale of
Rainbow systems were included in the Japan
market. Dataquest believes that LRC-originated
single-wafer plasma-assisted RIE technology
accounted for a substantial portion of the 1990 dry
etch market. LRC has been very successful in the
evolutionary extension of the original AutoEtch
technology to die Rainbow family.
The Rainbow system, with its patented
split-powerAevorse-phase RF source, has been very
successful in gaining market share in the oxide etch
market LRC is well positioned to e^^loit the
fast-growing metal etch market by virtue of its
joint-development projects with SEMATECH and
Du Pont LRC's prominent position in the Asia/
Pacific-ROW and European markets, together with
its successful J{q>anese partnership with Sumitomo
Metals, will enable it to continue its key global dry
etch market leadership. The LRC/Sumitomo Metals
partaership appears to be a win-win position for the
two companies because of their combined regional
presence and relative strengths in RF-based plasma
etch, microwave/ECR etch, thermal CVD, and
ECR CVD.
Sumitomo Metals, by virtue of its ECR etch
technology and successM Rainbow partnership
with LRC, ranked fifth with 4.8 percent of the
market Dataquest believes that Sumitomo Metals
wiU become an increasingly iinportant player in the
large Japanese market because of its parent
cocapany^s (Sumitomo Gioiq>'s) deep pockets and
its growing investment in customer support facilities throughout Jsqpan.
Tegal ranked sixth with 4.5 percent of the
1990 market Tegal is still in the midst of its
transition from being a Motorola Enterprises o ^
tive subsidiary to becoming a free-standing, global,
C1991 Dauqueit Incoipanted April-Reproductian Prohibited
SEMMS NewilMten 1991 Equi]niieiit-^tch/ae«n
DRY ETCH EQUIPMENT: 1990 MARKET IN REVIEW
multqjToduct equipment company. In 1990, the
company introduced its mainjGrame Series 6000
cluster tool, which will accommodate aU of its dry
etch modiile technologies in the plasma, REE, and
tnode market segments.
Anelva, with 4.3 peirent of the 1990 market,
ranked seventh. In addition to its batch and singlewafer RIE market presence, Anelva recently
entered the microwave/ECR dry etch market.
Dataquest believes that Anelva will be a key
player in future cluster tool integration markets
because of its strengths in related thin-film technologies such as dry etch, PVD, and CVD.
DATAQUEST PERSPECTIVE
range of technologies such as enhanced RF plasma
sources, MERIE sources, and microwave/ECR
sources in attempt to meet the complex
requirements of cost-effective sub-0.5-micron
device geometries. Japanese companies have
almost caught up with North American compani^
in the quest for global market share. North
American companies are actively pursuing the
large international market through joint vratures,
trading parmerships, and regional subsidiaries.
Japanese companies are leveraging off their
strengths in die large Japsaasse maiket in order
to follow their increasingly globalized customers
offshore.
Krishna Shankar (San Jose)
Kunio Achiwa (Tokyo)
Dataquest believes that conqietition has intensified in the crucial global $683 million dry etch
equipment market. European, Ji^anese, and North
American companies are aggressively pursuing a
The topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, '«4iich inchide wafer ftb equ^mient
sqypliers, semioondoctor matwriaij conqianies, and semiconductor device mannfactiueis. Hie topics selected indiratrttiebroad range of research
tbat is conducted in the SEMMS groiq). Clients, however, often have specific infomiatian requirements that dttier go beyond the level of detail
contained in die newsletters or beyond ibc scope of ^^lat is noonally published in the newsletters. In order to provide conq>lete decision sappott
to our clients, Dataquest lias a consnltiog service available to handle diese additional information needs. Please call Stan Bmederie at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
01991 Dittquest Ihcoiponted Ainit-RepraductioD PrahiUted
SEMMS Newilettan 1991 Equipiiinat • -Btrh/dem
0010063
^k^i^^''"!^^
WWM!^
ifm^ji^^^
DataQuest
acoRipanyof
Tlic Dun &^racistiul ICorpontH
^^^^Ci^
Research Newsletter
WET PROCESSING APPEARS TO HAVE NINE LIVES
Despite the widely held opinion that wet
process technology will eventually be replaced by
dry process technology, the demand for wet station
equipment continues to grow in submicron applications. For instance, the number of cleaning steps is
increasing with each new generation of DRAM
(see Table 1). Li addition, wet stations are widely
used throughout front-end processes (see the
following hst), and it is unlikely that these steps
will be replaced by dry processes in the near future.
The prevalence of cleaning steps is evident in
reviewing process flows (see Figure 1).
Wet process applications include the
following;
• Before and after CVD process
• After etching process
SURFACE CONTROL
With submicron pattern development, the role
of wet stations has shifted from convaidonal particle control to surface control, an example being
the cleaning of native oxide film prior to a chemical vapor deposition (CVD) process. As a result,
oxygen dissolved in deionized (DI) water needs to
be reduced below a lO-parts-per-billion (ppb) level.
Also, in the area of particle control, inspection
specifications of 10 particles >0.1 micron per
200mm wafer are required for 16Mb DRAMs
compared with 10 to 20 particles >0.2 micron per
150mm wafer for 4Mb DRAM. In practice,
however, because detection of 0.1-micron particles
is not possible, inspection is conducted at the
0.16-micron level, and acceptance criteria seem to
be estabUshed at about 30 particles.
• After implantation process
• Photoresist stripping process
• SUicon nitride etching process
• After cleaning of poly/PSG
In particular, silicon nitride (Si^N^) etching is
almost exclusively a wet process because dry etching shows poor selectivity. Also, wet cleaning has a
major advantage in that it does not leave a spot on
the wafer.
HIGHLY INTEGRATED SYSTEM
WITH SHUTTLE
The increasing level of clean room automation is clearly reflected in wet station equipment. The use of automatic wafer transportation
systems, which is a key element of clean room
automation, is driving the development of wet stations capable of accommodating automated guided
vehicle (AGV) systems. Of wet stations shipped ia
1989 by the largest Japanese vendor, Dainippon
TABLE 1
DRAM Process Trends
Number of mask layers
Total number of process steps
Number of cleaning process steps
1Mb DRAM
4Mb DRAM
16
20
300
35
.
350
48
Source: Dataquest (Febnuny 1991)
®1991 Dataquest Incorporated February—ReproductiQii Prohibited
SEMMS Newsletters 1991 Etch & Qean
0009357
The content of this report represents our irjerpretation and analysis of information generally available to the public or releas&i 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. Individual companies reported on and atudyzed by Dataquest
may be clients of this and/or other Dataquest services This information is not fiimished in connection with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mention&i and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
WET PROCESSING APPEARS TO HAVE NINE LIVES
FIGURE 1
Process Flow
1. Prediffusion Rinsing
NH40H/
H^O^
QDR
Hcl/
HF
QDR
QDR
FR
SD
QDR
FR
SD
FR
SD
FR
SD
QDR
H2O2
FR
SD
2 . Si3N4 Stripping
HaPO,
HaPO*
1
3. Photoresist Stripping
H2S04 /
H2O2
H2SO4 /
H2O2
4. Buffer Oxide Etching
HF
QDR
5. HF Cleaning
HF
QDR
Note: QDR = Quick Damper Rinse, FR = Final Rinse, SD = Spin Dryer or IPA vapor dryer
Source: Dataquest (February 1991)
Screen, 80 percent were systems equipped with a
shuttle.
The typical system consists of an I/O port,
clean stocker, wafer transfer section, shuttie, and
integrated wet station (see Figure 2). Carriers are
transported to the wet station I/O port by the AGV.
The carriers move into the clean stocker, which is a
buffer zone for carders at the start and finish of the
wet process step. Carriers then move to a wafer
transfer section, where the carrier is changed over
to a second carrier. The second carrier is used only
within the wet station in ordet to prevent carriers
saturated with chemicals from moving to the next
process along the AGV and contaminating the
process.
The internal carrier is transported to a loading
position of the wet station via the shuttie. The
carrier is moved through the wet process step by a
robot. On completion of the wet process, the wafers
are transferred to the original carrier and kept in the
clean stocker until it is picked up by the AGV.
THROUGHPUT
The average throughput for 150mm wafers is
12 carriers per hour, including chemical exchange
0009357
time and required maintenance time. To minimize
the process time, a weighing tank is heated up in
advance.
Equqnnent uptime varies gready, depending
on the vendor. Leading Japanese vendors typically
target an uptime specification of greater than
95 percent of production time (excluding ordinary
maintenance time). This trend suggests that reliabiHty is a key factor to market acceptance.
HIDDEN KNOW-HOW
Although the wet stations do not use leadingedge technology, equipment vendors are upgrading
systems with detailed design changes in order to
improve equipment performance and reliability.
Japanese vendors in particular have paid close
attention to many of the small design features that
hold the key to wet station reliability.
For instance, liquid-level sensors are used to
detect wafer head positions in addition to the sensing of chemical mixture. Also, some of the equipment is designed to transfer a wafer carrier in water
from the quick danger to the final rinsing process
to prevent oxidation from exposure to the air.
®1991 Dataquest IiKoipoiated Febnuiy-Repioduction Piobibited
SEMMS Newsletters 1991 Etch & Clean
WET PROCESSING APPEARS TO HAVE NINE LIVES
FIGURE 2
Highly Integrated Wet Station Structure
iVIaintenance
Zone
Stocker
Partition
Wafer
Transfer
Section
Load
Bath
Bath
Bath
QDR
FR
SD
.
J
1
i
Process
Zone
I
i
i 1
Main Robot Transportation
Robo t
Partition
Unload
Out [ In
C-to-C 1 C-to-C
Sliuttle
T
Robot
.
•«—
''
Bay
Robot
I/O Port
A G V Line
Source: Dataquest (February 1991)
Finally, the upper edge of the bath is designed
with sharp, angular edges. Suppliers have learned
that this design ensures smooth liquid ovetflow,
thereby preventing bacteria from growing in a
standing liquid. It also permits the bath to be
quickly drained during quick damper rinse.
FUTURE TRENDS
Wet station manufacturers are expected to
develop a carrierless handling system. This design
will eliminate the carrier, which is a major obstacle
in preventing cross-contamination. This design will
have the added cost advantage of eliminating
carrier wear and tear.
Ongoing trends in process equipment integration suggest that the cleaning process will eventually be incorporated into a system that can interface
wMi in-situ cluster tools. However, various problems need to be solved, including the removal of
inorganic contaminants and particles or the prevention of spots on the wafer. SiO^ etching using HF
vapor is the first step to clustering. This cleaning
process uses HF in a vapor rather than wet state.
Vapor cleaning is effective in preventing organic
and inorganic contaminants and particles ftom
dqx)siting on an activated silicon surface. Applications for HF vi^)or clean are listed as follows:
• Cleaning before silicide formation
• Removing native oxide in trench
01991 Dataquest Incorporated February-Reproduction Prohibited
SEMMS Newdetten 1991 Etch & Clean
• Qeaning before epitaxial deposition
• Cleaning after CVD oxide deposition
• Qeaning before metal deposition
As a transition to clustering for native oxide
prevention, a system is designed to carry out the
rinsing-drying process under an atmosphere of
Nj gas.
Filially, Dataquest expects wet station footprints to shrink in the future. The trend has been
toward a larger foo^nint for each new generation.
The average fab now uses about IS wet stations,
which occupy 30 to 40 percent of expensive clean
room floor space. The percentage will increase for
200mm wafer fab clean rooms because of the
forecast increase in the number of cleaning steps.
In order to counterbalance this trend, wet station
manufacturers need to design systems with smaller
footprints. The single-wafer systems have smaller
footprints, but their low throughput is not suitable
for volume production.
DATAQUEST CONCLUSIONS
Wet process equipment has historically been a
niche technology. The vendor base has been populated by a large number of small companies.
However, wet station sales are fast becoming a significant segment of total semiconductor capital
equipment sales. As the leyel of integration and the
average selling price of equipment increase,
0009357
WET PROCESSING APPEARS TO HAVE NINE LIVES
Dataquest expects only Ihe larger players that offer
highly automated stations to survive. The size of
the market may also attract the participation of
large diversified semiconductor equipment makers
that want to broaden their product line.
Although wet process technology is relatively
old in terms of semiconductor technology, Js^anese
vendors are excelling at pushing the use of this
equipment into the submicron era. U.S. equqmient
makras that have been the most vocal in sounding
the deadi knell of wet process technology should
take notice.
Kunio Achiwa (Tokyo)
Mark FitzGerald (San Jose)
The topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, which include wafer fab equipment
suppliers, semiconductor materials conmanies, and semiconductor device mauufactnrers. The topics selected indicate flie broad range of research
that is conducted in the SEMMS groiq). Clients, however, often have specific information requirements fliat eidia go beyond the level of detail
contained in the newsletters or beyond the scope of vibaX. is normally published in the newsletters. In order to provide coiiq>lete decision support
to our clients, Dataquest has a consulting service available to handle these additional information needs. Please call Stan Braederle at (408)
437-8272 or Joe Gi^iier at (408) 437-8206 to discuss your custom requirements.
00093S7
0199I DataquMt Incoiponted Febnuiy^Reproductian Piobibited
SEMMS Newiletlers 1991 Etch & Clean
'^M^mm
W
mm
Dataquest
3 company oi
TIK Dan& Bndstrect corpOfgdHm
s^l^o^
iiil
Research Newsletter
PVD EQUIPMENT: 1990 MARKET IN REVIEW
SUMMARY
The worldwide physical vapor deposition
(PVD) equipment market grew 11 percent from
$368 miUion in 1989 to $408 million in 1990. All
1990 PVD market growth occurred in the sputtering equipment market segment, which accounted
for 88 percent ($359 million) of the 1990 market.
The worldwide evaporation equipment market,
which accounted for the remaining 12 percent
($49 million) of the PVD market, remained flat in
1990. The PVD market grew in spite of an overall
1990 wafer fabrication equipment market environment that was slightly down. The onergence of
200mm double-metal 4Mb DRAM shrink production and 16Mb DRAM pilot lines, together with the
migration of microprocessors, gate arrays, and
VLSI logic devices to triple-metal processes, were
responsible for the PVD market's growth.
Japan, with 48 percent of the world's PVD
market, retained its position as the largest regional
market. The North American market, with 31 percent of the 1990 total, ronained the second-largest
regional market. Japanese companies increased
their ownership to 58 percent of the 1990 worldwide PVD market, while North American companies Ci^tured 30 percent of the total. Japanese
PVD equipment conipanies took advantage of their
home-field position within the leading-edge Ji^anese market to introduce evolutionary improved
versions of their PVD tools. North American companies, in contrast, focused on revolutionary cluster
tools that incorporated modular PVD chamber
architecture and offered future integration paths
with related chemical vapor deposition (CVD) and
dry etch processes.
REGIONAL MARKETS
Figure 1 shows the 1990 regional PVD markets and ownerships. Japan, the world's largest
PVD market, accounted for 48 percent of the total.
Jinan's emphasis on high-volume, leading-edge
DRAM production, together with its increasing
presence in gate arrays, ASICs, chip sets, and
foundry products, led to its evolution as Has largest
PVD market. The Japanese PVD market grew
rapidly in 1990 because of the move from singlemetal lMb/4Mb DRAM processes to double-metal
4Mb shrinks and 16Mb processes. The move to
IQOitaa fabs in J^an has also resulted in dramatic
increases in guttering equq>ment average selling
prices (ASPs). Japanese device manufacturers
focused heavily on double-metal DRAM processes
and triple-metal gate airay processes in thenattempts to stay ahead of comnuxlity DRAM and
gate-arr^ competition from Asia/Pacific-Rest of
World (ROW) device manufacturers. Small, fast
DRAM and ASIC chips inq)lemented in multilevel
interconnect technologies and small outline packages will offer Jiq)anese device manufacturers a
cost4)eTformance edge in the highly competitive
DRAM and gate array markets.
The North American PVD market, with
31 percent of the world's total, remained the
second-largest maricet. North America has traditionally been the leading producer of highperformance microprocessors and other advanced
VLSI logic chips. Such random logic chips rely
heavily on double-metal interconnect technology.
In the mid-1980s, the North American production
region led the world in the migration to multilevel
intercoimect processes. However, microprocessor
and advanced VLSI logic unit volumes are much
smaller conq>ared with DRAM and gate array unit
volumes. North American device manufacturers
such as Intel and Motorola have achieved higher
revenue streams per dollar of wafer fab capital
investment due to the higher ASPs oftiieirproprietary microprocessor devices. Typically, North
American device manufacturers have bought less
core wafer fab equqnnent such as PVD, CVD,
lithognphy, dry etch, ion in:q)lantion, and diffusion
0 1 9 9 1 Dataquest Incoxpoiated April-Reproductioii Prohibited
0010193
SEMMS Newtletten 1991 Equipment—Deporitian
The content of this report represents our irjerpretation and analysis c^ inforrmuion generally available to the public or releaseti by responsible individtmls 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 Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services- This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
<^r to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, jkmi time to time, have a long or short position in the securities
mentioned and may sell or buy such securities,
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
PVD EQUIPMENT: 1990 MARKET IN REVIEW
FIGURE l
1990 PVD Equipment Regional Markets and Ownership
(Revenue in Millions of Dollars)
Asla/PaolflcROW $28 7%
Markets
Total = $408.4 Million
Ownership
Total = $408.4 Million
Source: Dataquest (April 1991)
equipmoit in order to ramp up to a given revenue
level than have their DRAM-focused Japanese
counterparts. Dataquest believes, however, that die
North American PVD market will exhibit increased
growth during the next few years because of the
rapid migration of microprocessors, gate arrays,
and VLSI logic devices to three- and even fourlevel metal processes. The increasing number of
offshore J^anese fabs in North America will also
contribute to the growth of the North American
PVD market
The European market grew a healdiy 25 percent to occupy 14 percent of the 1990 PVD market.
.A substantial portion of Europe's market growth
was due to the r a n ^ i ^ of offshore Ji^anese and
North American semiconductor operations in
Europe. Dataquest e;^>ects continued PVD market
growth in Europe, fueled mainly by the globalization of the semiconductor industry. However, die
offshore European fabs toid to be clones of domestic Js^anese and North Americanfiibs.Hence, PVD
equipment con^anies have to be global in customer siqjport and service in order to get a slice of
die market in all major worldwide semiconductor
manufacturing regions.
The Asia/Pacific-ROW market shrank fix>m
10 percent of the 1989 PVD market to 7 percrait of
die 1990 PVD market. Factors such as excess
DRAM c{9>acity and low DRAM prices, excessive
PC chip set coiiq)etition, inability to raise ciqntal
0010193
due to depressed stock market valuations, and
political instability were responsible for the weak
1990 Asia/Pacific-ROW market. Dataquest expects
growth to pick up in die Asia/PadfLc-ROW market
in 1991 and beyond because of a new round of
capital investment in 200mm 16Mb DRAM fabs in
South Korea, new subtnicron ASIC and foundry
fabs in Taiwan, and new fabs in developing countri^ such as Malaysia, Indonesia, Thailand, and
India.
REGIONAL OWNERSHIP
Figure 1 shows regional ownership of the
worldwide 1990 PVD equ^iment market J^ianese
PVD companies dramatically increased their
owTkearship fiom 44 percent of the $368 million
1989 market to 58 percent of die $408 million
1990 market. Sony's acquisition of Materials
Research Corporatioii (MRC) of Orangeburg, New
York, toward die end of 1989 was a significant
factor in the rqnd PVD maiket share gain by
Japaaesc equipment companies. Jf^anese compa-nies have also been successful in jointly developing
higb-vohime fluttering systems with their large,
advanced DRAM-focused domestic Japanese
semiconductor manufacturers. Japanese PVD
companies snch as Anelva and Ulvac are large,
diversified companies that cater to a wide range of
01991 Dttaquett bieoiponUsd April^tepradoctiaii PioUbited
SEMMS Newdetten 1991 Equipment—^DepontiaD
PVD EQUIPMENT: 1990 MARKET IN REVIEW
\
industrial vacuum coating sqpplications. Diversified
target markets and large size enable Japanese PVD
companies to successfully weather the cylicality
and capital-intensive nature of the wafer fabrication
equipment industry.
The sheer size of die domestic Japanese market, together with the growing number of offshore
Jqxmese fabs, has provided Jjpmese PVD companies with economies of scale in sputtering system
production and an^le R&D funds for new product
development. Dataquest expects Japanese PVD
conipanies to leverage off their domestic strengths
and increase their international business activities
as they follow their increasingly globalized Japanese semiconductor manufacturers offshore. Ji^anese PVD companies will face numerous
challenges relating to global customer supfxnt and
service capabilities, new business culture and customer/vendor relation^ps, and increased competition from North American and European PVD
equipment con^anies that will attenq)t to protect
their home base.
North American coii^)anies owned 30 percent
of the 1990 PVD market North American PVD
conipanies are in die vanguard of the worldwide
PVD market movement toward flexible, modular,
cluster-tool sputtering systems that can integrate
future related processes such as CVD, dry etch, dry
strip, and rapid thermal processing. North American companies are confronted by a relatively
slower-growing domestic market con^ared with
the international market. Hence, they are r^idly
redirecting their efforts toward penetration of the
international PVD markets. North American companies also face the challenges of global customer
support and service, new business cultures, and the
high costs associated with penetrating international
markets through regional subsidiaries and distributors.
European coiiq)anies ci^tured 12 percent of
the 1990 PVD market. Companies such as Balzers
and Electrotech were active in the clust^-toolbased sputtering equtpment market. European companies expect to capture an increasing portion of
the European PVD business provided by domestic
and offshore fabs. Europe's stance on domestically
mamifactured chips, together with global semiconductor con^anies' attempts to stimulate the
domestic infrastructure by procuring equj|)ment
locally, is expected to support the growdi plans of
European PVD conipanies. European PVD companies such as Balzors and Leybold-Heraeus will
benefit from the deep pockets of their diversified
industrial parents in a maimer similar to Japanese
PVD companies.
COMPANY RANKINGS
Tible 1 shows the worldwide 1990 PVD company rankings and the revenue split between
TABLE 1
1990 Worldwide PVD Equipment Company Rankings
(Revenue in Millions of Dollars)
Revenue Split by
Market Segment (%)
Revenue
Market
Share (%)
Anelva
94.9
23.2
Sputtering: 96, Evaporation: 4
Vaiian
84.0
20.6
Sputtering: 100
Ulvac
72.0
17.6
Sputtoing: 81, Evaporation: 19
MRC
62.6
15.3
Sputtering: 100
AppUed Materials
15.0
3.7
Sputtering: 100
E.T. Electrotech
13.5
3.3
Sputtering: 100
Balzeis
13.5
3.3
Sputtering: 52, Evaporation: 48
Temescal
13.1
3.2
Evaporation: 100
39.8
408.4
100.0
Company
Others
Worldwide Market Total
9.7
Sputtering: 88, EvapcHation: 12
Note: HK table diowt calendar year 1990 •yitema levonie; tpaet and aecHcet aie not included Some oatannu do not add to totali diown became of
lOUQfllQB.
Soune: Dataqueat (^ail 1991)
ei991 Dataqueat iKoiporated Apiil-Reprodiictiati Pioliibitad
SEMMS Newilett«n 1991 Equipment—Depotitioa
0010193
PVD EQUIPMENT: 1990 MARKET IN REVIEW
sputteiing and evaporation. Tbree European companies, three J^anese companies, and two North
American conq>anies make iq> the top eight PVD
rankings.
Anelva, with 23.2 percent of the 1990 market,
was the top-ranked PVD company. Anelva's modular Series 1051 system continued to be successful
in the Japanese madcet. Anelva is a key player in
the integrated titanium-titanium nitride barrieraluminimi alloy sputtering market for submicron
supplications and is also beginning to ship a growing number of its sputtering systems to offshore
J^anese fabs. Anelva introduced a 200mm version
of its Series 1051 product family in 1990. The
cornpany is actively developing integrated-process
cluster tools incoiporating PVD, CVD, and dry
etch/ECR etch modules. Anelva's Series 1551 is
targeted at R&D and pilot-line 16Mb/64Mb
integrated process applications. Processes developed on the Series 1551 can be transferred to the
high-volume Series 1051 cluster tool systems.
Anelva's market presence in the disk coatings,
LCD flat panel coatings, and other industrial coatings segments synergistically complements its
semiconductor PVD equipment market activities.
Varian, with 20.6 percent of the 1990 market,
retained its position as the second-ranked PVD
coinpany. Vaiian successfully raniped \sp production of its flagship M2000 cluster-tool-based sputtering system in 1990. The company focused its
efforts in production enhanconents, new sputtering
source technology such as the Quantum source, and
penetration of key global semiconductor fabs. Varian's M2000 Series is the company's platform for
future integrated process s^lications involving dry
etch, CVD, and PVD. In addition, Varian's older
Series 3000 family of multichamber sputtering systems continues to be a successful cash cow product
that targets capacity expansion fabs and more
mature device fabs in the Asia/Pacific-ROW
regions.
Varian's recent divestiture of all its noncore
businesses leaves it with a highly focused semiconductor equipment market strategy in the core ion
implant and sputtering systems business. Varian's
partnexship with Tokyo Eiectron Limited (TEL) has
positioned the company well in achieving a global
market presence. La addition, there is great syn^gy
between Varian's PVD technology and TEL's dry
etch technology. TEL has also retained the license
to Varian's tungsten CVD technology in Jq)an.
Dataquest believes that Varian and TEL may collaborate in the future to market an integrated tungsten CVD/etchback/alumimim PVD system based
on the M2000 cluster tool platform.
0010193
Ulvac, the third-ranked PVD conq>any, owns
17.6 pecceat of the 1990 market. XJlvac's sputtering
systems accounted for $58 million (81 percent) of
its 1990 PVD revenue; evaporation equipment
made up the remaining 19 percent. Ulvac, with its
new MLX-3000 series flexible cluster tool, is
actively involved in the modular sputtering systems
market. The MLX-3000 system is capable of
200mm processes and integration with other related
Ulvac thin films technology such as plasma clean,
selective tungst^ plugs, dry etch, dry strip, and
RTP processes. Dataquest believes that Ulvac is in
the forefront of the Jiqpanese market movement
toward integrated process capabilities. Ulvac's
future C-2111 Stellar mega cluster tool is designed
to link two MLX-3000 cluster tools together with a
transfer module. Ulvac is aggressively expanding
its presence in North America and Europe dux>ugh
wholly owned regional subsidiaries.
MRC, the fourth-ranked PVD company,
owned 15.3 percent of the 1990 market MRC's
Echpse sputtering system continues to successfully
penetrate key worldwide fabs. MRC recently
introduced die Eclq}se-Star system^ which features
several significant enhancements in process, particle control, and reliability. The company has the
unique advantage of offering one-stop capability in
sputtering target materials development and sputtering systems development As device geometries
head into the submicron regime, the coupling of
target design, source materials, and sputtering system design becomes more crucial. MRC is also a
leading merchant supplier of targets to other PVD
equipment vendors and semiconductor manufacturers. MRC appears to have successfiiUy made the
transition from being an independent publicly
owtted conQ>any to a wholly oiraed subsidiary of
Sony USA, Inc. Dataquest believes that Sony's
global business image, deep pockets, and leading
market position in optical and magnetic thin film
technology for consumer applications will
influence MRC to broaden its semiconductor
equipment focus to include related areas such as
mi^eto-optical disks, coraps^t disc coatings, and
LCD flat panel coatiiigs. Dataquest believes that
MRC wUl continue aggressive development of
advanced flexible cluster tools that combine its
sputtering and dry etch capabilities.
Applied Materials, which entered the PVD
marixt in 1990, c^tured 3.7 percent of the 1990
market Applied has built up great momentum in
penetrating the sputtering process at several major
worldwide submicron fabs. The Endura 5500 PVD
syston's unique staged ultrahie^ vacuum enviranmeat, advanced source design, and integration
ei5)91 Dataqueit Ineoiponted Apiil-RapRMbictiaii Pndnbited
SEMMS Newdetten 1991 EquipmBat—Deposition
PVD EQUIPMENT: 1990 MARKET IN REVIEW
capabilities with related processes such as tungsten
CVD and etchback puts Applied in a strong position to offer a global, one-stop ULSI interconnect
solution. Motorola's recent acceptance of the
Endura system for its advanced MOS 11 submicron
20Qmm fab in Austin, Texas, represents a significant design win for Applied Materials. Dataquest
believes that AppUed Mat^als will face vigorous
competition from other leading PVD con^Mmies
that are racing to introduce their own cluster tools
wMi integrated PVD, nmgsten CVD, and etchback
capabilities to market.
Balzers and Electrotech, each with 3.3 percent
of the 1990 PVD market, were the sixth- and
seventh-ranked companies. Both companies are
actively marketing PVD cluster tool appUcations
for barrier metal and aluminum interconnect
processes. Balzers' recent acquisition of Spectrum
CVD fix)m Motorola New Enterprises (the new
entity is called BCT-Spectrum) is aimed at exploiting the syneigy between Balzers' PVD technology
and Spectrum CVD's tungsten plug technology.
The acquisition also provides Balzers with an
inofKntant manufacturing and customer support
base in the inqxjrtant North American market.
ElectDtech, with its Sigma Series PVD cluster tool,
is engaged in developing comprehensive dun film
deposition and dry etch c^abilittes. The Sigma
PVD pl^oim nicely conqilements Hectrotech's
Delta CVD platform and Omega dry etdi platforoL
Dataquest believes that Electrotech has positioned
itself as a key domestic European thin films company that will play an inqxntant role in the fastgrowing European semiconductor manufacturing
arena.
Tbmescal, the eight-ranked PVD company, is
entirely focused on the niche evsqxxration equq)ment market. Evaporation processes continue to be
used in lift-off metallization and GaAs conq>ound
device q?plications. Evaporation technology also
plays an inqxntant role in backside gold dqtosition
processes to improve chip conductivity and adhesion to packages. Temescal is integrating more
closely with HOC, its British parent, in order to
apply its expertise to BOC's industrial Ihin film
market activities.
DATAQUEST PERSPECTIVE
Dataquest believes that the $408 million PVD
market will continue its technology-driven growth
through the 1990s. The sputtering equipment market is characterized by intense coiiq>etition as
several new entrants challenge the entr^iched market leaders. Integration technology involving PVD,
CVD, and dry etch will play an important role in
the cluster tool market The PVD equ^nnent market is dividing into a standalone, high-throughput
rigid multichamber sputtering market and a cluster
tool market with modular products that combine
dry/plasma clean, sputtering, CVD, and dry etch
processes.
The migration of 4Mb/16Mb DRAM devices
to double-level metal processes and the migration
of microproce^ors, ASICs, and VLSI logic to
triple-metal processes promise a healthy future for
the PVD equipment market. However, parallel
developments in CVD copper, alviminum, tungsten,
and titanium nitride tedmologies could put a
damper on fiitme growth of the PVD niaiicet.
Dataquest predicts that an intetestiDg mg-of-wai
between the CVD and PVD conductor film markets
will occur in the next few years. Maricet application
forces characterized by PVD and metal CVD
convergence wiU lend an interesting twist to this
conflict.
Krishna Shankar (San Jose)
Kunio Achiwa (Tokyo)
The topics covered by SEMMS newsletters are selected for dieir general interest to SEMMS cheats, vAadi inchide wafer fiab equqiment
suppliers, semionidactar matmBls conqianies, and semiconductor device manufbctniers. t h e topics selected indicate die broad range of research
diat is conducted in the SEMMS groiqi. Clients, however, often have specific infotmattcni requirements that eidter go beyond the level of detail
contained in die newsletters or beyond die scope of what is nonnally published in the newsletters. In order to provide conqilete decision sqiport
to our clients, Dataquest has a ctmsulting service available to handle these additional infotmatton needs. Please call Stan Bruederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
ei991 Dataijueft Incoipanled Apiil-Reiiroductiim Pioliibited
SEMMS Newiletten 1991 Equpment^-Depoiitiai
ooiom
Dataquest
atompsnyof
The DOnii Bndstreet corpofation
Research Newsletter
C V D EQUIPMENT: 1990 MARKET IN REVIEW
SUMMARY
The worldwide chemical vapor deposition
(CVD) equipment market grew 13 percent from
$609 million m 1989 to $689 million in 1990. The
CVD market exhibited growth in spite of an overall
1990 wafer fabrication equipment market environment that was sUghtly down. Technology drivers
such as the emergence of 200mm double-metal
4Mb shrink production and 16Mb pilot DRAM
processes and the migration of miaoprocessor/
IILSI logic devices toward triple-metal processes
were responsible for the persistent CVD raaifcet
growth. The vertical low-pressure CVD tube market segment and the LPCVD/PECVD-dedicated
reactor market segment were the highest-growth
segments of the 1990 worldwide CVD market.
North American con^anies ci^tured a dominant 60 percent share of the 1990 worldwide CVD
equipment market, even though their biggest market is in Japan, which now constitutes 46 percent of
the worldwide CVD market. North American companies dominate the LPCVD/PECVD reactor
market segment, while Japanese companies dominate the fast-growing vertical LPCVD tube market
segment.
The North American CVD market maintained
its 32 percent share of the 1990 worldwide CVD
market. North America has traditionally been the
leading producer of high-performance microprocessors, microcontrollers, and VLSI logic devices. All
of these random logic devices rely heavily on
advanced CVD films for multilevel interconnect
technology. Capacity e?q)ansion and new fab additions by offshore Je^anese fabs in North America
also contributed significantiy to the North American CVD equipment market.
The European CVD market exhibited healthy
growth in 1990, growing to represent 13 percent of
the 1990 CVD market A substantial portion of this
growth was due to the expansion of North American and Jiq>anese device nuinufacturers into Europe
in order to meet the potential requirements for
domestically diffused European semiconductors.
The Asia/Pacific-Rest of World (ROW) CVD
equipment market declined 33 percent between
1989 and 1990. Falling DRAM prices, economic
and political uncertainties, and the collapse of the
Taiwanese stock market had a significant dampening effect on flie Asia/Pacific-ROW market.
REGIONAL MARKETS
REGIONAL OWNERSHIP
Figure 1 illustrates the 1990 regional CVD
equipment markets. As [neviously stated, Japan,
with 46 percent of the 1990 marlret, is the latest
CVD equj^nnent market. This share is in line with
J e a n ' s stature as the world's largest semiconductor
production, consumption, and wafer fabrication
equipment market. Many Japanese device manufacturers continued to modestly increase fheir coital
spCTding over 1989 levels in spite of a relatively
flat chip demand envirorraient Japan-based device
manufacturers invested strategically in 200ram fabs
geared toward double-metal 4Mb DRAM shrinks
and 16Mb DRAM pilot lines.
Figure 1 shows tiie 1990 world CVD maritet
by regional ownership of equipment companies. As
stated. North American companies continued to
dominate with 60 percent of the 1990 market,
North American compani^ continued their market
focus on dedicated LPCVD/PECVD reactors for
qiphcations such as low-temperature plasma oxide,
plasma nitride, and tungstenAungsten sUicide fHios.
North American CVD companies have a significant
lead in the development of cluster-tool-based
integrated process sequences involving LPCVD/
PECVD and dry etch applications.
0 1 9 9 1 Dataquest Incoiponted Apiil-4leproduction Piobilnted
00100S7
SEMMS Newtletten 1991 Equipment—Depotiticin
The content of this report represents our interpretation and analysis cf 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 protided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
nuxy be clients of this and/or other Dataquest services. This information is notfiimished in connection with a sale or offer to sell securities or in connection with the solicitation of an
i^r to buy securities This firm and Us parent and/or their cfficers, stockholders, or members of their fomilies may,fromtime to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
CVD EQUIPMENT: 1990 MARKET IN REVIEW
FIGURE 1
1990 CVD Equipment Regional Markets and Ownership
(Revenue in Millions of Dollars)
Joint
Ventures
$21 3%
Markets
Total = $689 Million
Ownership
Total = $689 Million
Source: Dataquest (April 1991)
Japanese CVD equipment companies, which
captured 25 percent of the 1990 CVD market,
focused on the vertical LPCVD tube segment for
high-volume thermal nitride, polysilicon, and
BPSG/thermal oxide applications. Dataquest
believes that J^anese CVD equipment companies
will address the low-temperature plasma oxide/
plasma nitride film markets in the near term.
Ji^anese equipment companies are developing
ECR CVD and other low-temperature CVD
processes for sub-0.5-micron device q>plications.
European conq)anies owned 12 percent of ttie
1990 CVD market. Conq)anies such as ASM International and Electrotech successfully addressed
high-growth market segments such as vertical
LPCVD tobes and dedicated PECVD reactors.
European semiconductor R&D programs such as
JESSI and ESPRIT are actively involved in
developing advanced CVD capabiUty for Europebased 16Mb DRAMs and microprocessorAJLSI
logic implications.
COMPANY RANKINGS
Table 1 ranksfliemajor worldwide players by
their 1990 revenue in the CVD marketplace. Five
of the top ten spots are occiq)ied by North American coiiq)anies. Three Japanese conq>anies and two
0010057
European companies are included in the top ten
CVD conipany rankings. It is apparent that U.S.
companies have maintained their dominance of the
CVD equipment market.
Applied Materials ci^tured 29 percent of the
1990 CVD market, emerging with $201 million in
revenue. The PE5000 dielectric CVD system continued to proliferate new ^rplic^ons for a range of
low-temperature plasma oxide/plasma nitride
processes. La addition, Applied also captured a
significant portion of flie 1990 tungsten LPCVD
reactor market with its new PE5000 WCVD system. The PE5000 family's process versatility,
together with Applied's global presence and customer support, has successfully maintained
Applied's leadersh^ in the CVD market.
Novellus climbed with meteoric success to
the number two CVD spot in 1990. Novellus leveraged off its successful Concept One dielectric CVD
architecture to make further inroads into the worldwide low-temperature plasma oxide^lasma nitride
market. Novellus is poised for significant future
growth as it enlarges its product portfolio in 1991
and 1992 to address die burgeoning metal CVD
and cluster tool market.
Kokusai Electric and TEL exhibited spectacular growth in flie vertical UXTVD tube madcet.
They benefited from Japaxnese device manufacturers' r^id transition to vertical diffusion and
01991 Diuqoest Incoiponted April-Reproduction Prohibited
SEMMS Newiletten 1991 Equipmoit—^Depoiitioa
CVD EQUIPMENT: 1990 MARKET IN REVIEW
TABLE 1
1990 Worldwide CVD Equipment Company Rankings
(Revenue in Millions of Dollars)
Market Share
Company
Market Segment
Revenue
(%)
201.0
29.2
Novellas
63.0
9.2
PECVD reactors
Kokusai Electric
62.7
9.1
LPCVD reactors, LPCVD tubes
ASM International
56.0
8.1
LPCVD/PECVD reactors,
LPCVD/PECVD tubes
Tokyo Electron Limited
45.8
6.7
LPCVD reactors, LPCVD tubes
Genus
44.0
6.4
LPCVD reactors
Watkins-Johnson
41.0
6.0
APCVD reactors
Silicon Valley Group
22.5
3.3
LPCVD reactors, LPCVD tubes
Electrotech
17.5
2.5
PECVD reactors
Amaya
16.0
2.3
APCVD reactors
BTU International
12.2
1.8
LPCVD tubes
106.8
15.5
688.5
100.0
Applied Materials
Others
Woridwide Market
Total
LPCVD/PECVD reactors
Note: The taUe shows cahwirtar year 1990 systems revenue; spares and service are sot included.
Source: Dataquest (Apiil 1991)
LPCVD equipment for leading-edge 200mm 4Mb
DRAM and ASIC fabs. Vertical LPCVD tube average selling prices soared dramatically because of
increased automation, process control, and defect
reduction features. Dataquest believes that Kokusai
Electric and TEL will expand their vertical LPCVD
process applications to include a variety of thermal
nitride, polysilicon, and thermal BPSG/undoped
oxide applications.
ASM International and Electrotech were the
European entrants in the top 10 CVD conq}any
rankings. ASM Intonational introduced a 200mm
version of its successful PECVD tube product for
intermetal dielectric and passivation applications.
ASM International is also pursuing the LPCVD/
PECVD reactor market with its Advance 600 platform for applications such as low-temperature
oxides, nitrides, and aluminum-based metal CVD
films. Electrotech continued to expand the range of
CVD sgiplications for its Delta Series cluster tool
family.
0 1 9 9 1 Dataquest Incorporated Apiil-RepiDductian Probibited
SEMKfS Newsletters 1991 Eqmpment—DepotitioD
Genus, with $44 million m 1990 CVD systems revenue, maintained its leadership in the
metal CVD mariKt. Genus introduced the 8720ST
blanket tungsten system and the Series 6000 deposition/etchback cluster tool during 1990. The company also continued to rule the tungsten silicide
market, unchallenged. Genus has enhanced its
product porfotmance by successfully leveraging off
its SEMATECH Equipment Improvement Project
and its close relationship with IBM. Dataquest
e}q)ects Genus to broaden its product offerings as it
responds to increased competition in the metal
CVD marketplace.
Watkins-Johnson and Amaya continued to
occupy significant positions within the top 10 CVD
rankings because of their domination of die mature
high-volume BPSG film market. Competition,
however, is heating i^ in the BPSG film market as
a variety of atmospheric-pressure cluster tools and
vertical LPCVD tubes are e:q>ected to challenge the
belt-fiimace dominance of the BPSG film market.
0010037
CVD EQUIPMENT: 1990 MARKET IN REVIEW
BTU International and Silicon Valley Group
carved out spots in the CVD company rankings
because of their horizontal and vertical LPCVD
tube products. Both companies significantly repositioned tiiemselves in 1990 to address the fastgrowing vertical LPCVD tube market. Dataquest
expects BTU International and SVG to vigorously
compete in the quest for worldwide dominance in
the high-growth vertical LPCVD tube market.
DATAQUEST PERSPECTIVE
The CVD equipment market, which grew to
$689 million in 1990, is expected to continue its
healthy growth pattern during the next few years.
The market continues to be technology driven
because of the need for a variety of dielectric and
conductor films in increasingly complex submicron
interconnect processes. The CVD equipment market displayed growth in 1990 in spite of otho- key
segments such as lithogrq)hy and dry etch markets
being flat or slightly down. This growth illustrates
the technology-driven expansion of the CVD
equipment market.
North American companies continued to
dominate the CVD equqmient market with 60 percent of the 1990 market even though their biggest
market was Japan, which constituted 46 percent of
the world CVD market. North American companies
have maintained their CVD market lead^ship by
sustained innovation in reactor architecture and
new films coupled with successful globalization
business strategies. J^ianese CVD equ^nnent companies have captured 25 percent of die 1990 CVD
market by focusing on the large batch, automated
LPCVD tube market for thermal oxide, nitride, and
polysUicon {plications. European CVD companies
continue to pursue the global CVD market with the
support of European R&D consortia such as JESSI
and ESPRTT. Dataquest believes that conqietition
wUl intensify in the high-stake, high-growtii CVD
market as global competitors square off in a global
marketplace.
Krishna Shankar (San Jose)
Kunio Achiwa (Tokyo)
The topics covered by SEMMS newsletters are selected for flidr general interest to SEMMS clients, Miich include wafer fab eqajjunent
suppliers, semicondnctor wmtmaU companies, and semiconductor device mannfoctiirets. The topics sdected indicate the broad range of research
that is conducted in the SEMMS groiq>. Clients, however, often have specific information requirements that either go beyond the level of detail
contained in flie newsletters or beyond die scope of \rtiat is nonnally published in the newsletters. In order to provide coiiq>lete decision support
to our clients, Dataquest has a consultiog service available to handle tfiese additional information needs. Please call Stan Bmederle at (408)
437-8272 or Joe Gienier at (408) 437-8206 to discuss your custom requirements.
0010057
01991 Ditaqueit Incorporated April-Reptodiictiaii PiohiUted
SEMMS Newdetten 1991 Equipment—Depotitian
WW- PM^-^
Dataquest
accmipanyof
Tfic Dun & Bradstieet COTporatnn
• - .
" • -
• • • • " • # ;
Ij^^^r^
^pfjiV'j^K
^m-'i'L'-'-m
^rr^'^
Research Newsletter
SILICON EPITAXY EQUIPMENT MARKET--I990 MARKET IN REVIEW
Worldwide silicon epi reactor sales totaled
$68.2 million in 1990. Sales decreased 9.1 percent
from 1989 levels, reflecting very weak European
and Asia/Paci&c markets. Dataquest forecasts 1991
sales to total $55 million, down 14.7 percent on a
year-to-year basis. Sales in 1991 are e}q>ected to
decrease further because of the recession in North
America and flat demand for epi wafers outside
North America.
REGIONAL MARKETS
North American epi reactor sales totaled
$35.7 million, accounting for 52.3 percent of the
worldwide market (see Figure 1). Sales increased
12.6 percoit in 1990, fueled by e3q)losive growth in
CMOS epi wafer demand. Dataquest estimates that
North American merchant epi wafer demand grew
20 to 25 percent in 1990. Microprocessor manufacturers Intel and Motorola and IBM's DRAM lines
accounted for most of the growth in the North
American epi wafer market.
The J^anese ntiarket was the second largest,
with sales totaling $18.2 million or 26.7 percent of
the worldwide market. Reactor sales were down
12.1 percent on a year-to-year basis. Bipolar and
discrete jqjplications account for 73 percent of the
epi wafer demand in J^an, yet these markets are
FIGURE 1
1990 Silicon Epitaxy Reactor Market
by Region and Ownership
(Millions of Dollars)
/
/
/
\
\
Japan
$16.2
/
/
/
/
_
Europe
$11.9
North America
$35.7
/
1
\
\ Asla/Paclfic
1
1 t? 4
1
y
By Region
$68.2
European
JK
Companies
y^ \
$24.7
X
\\
/Japanese
—--.,_
y^ Companies \
""---^
V
\
\
/
North American
Companies
$36.8
*S-7
I
/
/
/
By Ownership
$68.2
Source: Dataquest (April 1991)
®1991 Dataquest Incorporated Apiil-Repioductioii Prohibited
SEMMS Newsletters Equipment—Deposition
0010041
77i« content of this report represertis our interpretation and analysis of information generally available to the public or released by responsible individuals in the subject companies, bM
is not guaranteed as to accuracy or completeness. It does not coruain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
affir to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
SILICON EPITAXY EQUIPMENT MARKET—1990 MARKET IN REVIEW
growing slowly and require only incremental
increases in reactor capacity.
The 1990 European and Asia/Pacific markets
decreased 38.7 and 60.7 percent, respectively.
European sales totaled $11.9 million, and Asia/
Pacific sales totaled $2.4 million. Epi reactor sales
in both markets suffered because of weak cs^ital
spending and flat growth in epi wafer demand.
However, in the past two years, single-wafer
reactors pioneered by ASM Epitaxy have made
significant inroads, especially in the larger-diameter
CMOS applications. Sales have grown from
$6 million in 1989 to $19.2 million in 1990.
The other vendors in the epi reactor market
are marginal players. Little development work is
going on in these con^anies, and their sales are
tied more closely to the bipolar and discrete epi
markets, which are growing very slowly.
COMPANY MARKET SHARE
The 1990 epi equipment market was dominated by one North American vendor. Applied
Materials, and one European vendor, ASM Epitaxy
(see Figure 1). AppUed Materials' worldwide sales
totaled $25 million; ASM Epitaxy's sales totaled
$19.2 million (see Table 1). Both con^anies have a
geographically diversified customer base in contrast to other players, whose sales were concentrated in a single region.
Applied Materials traditionally has been the
market leader in the silicon epi market since the
introduction of its radiant-heated barrel reactor in
the late 1970s. Its reactor design virtually eliminated thermally induced slip. As a result, these
reactors have been the system of choice for the
production of CMOS epi wafers.
TABLE 1
1990 Worldwide Silicon Epitaxy Reactor Market
by Company Share
(Millions of Dollars)
Company
Market
Share
Percent
Applied Material
25.0
36.7
ASM Epitaxy
19.2
28.2
Lam Research
7.9
11.6
LPE
5.5
8.1
Toshiba Machine
4.5
6.6
Moore Technology
3.0
4.4
Kokusai Electric
2.2
3.2
0.9
68.2
1.3
100.0
Rapro
Total—Worldwide
DATAQUEST CONCLUSIONS
The silicon epi market is being pulled in two
directions. Future applications, especially DRAM
devices, will require tighter thicloiess, resistivity,
pardcle, and oxygen specifications. These specifications cannot be met with current reactors but wUl
require new reactor designs. Trends in semiconductor capital equipment prices suggests that new reactors win be very expensive. Yet, the price of epi
films will have to decline dramatically before semiconductor manufacturers widely incorporate the
films into their device designs.
The diverging trends are limiting the growth
of the sihcon epi equipment market Dataquest
does not anticipate that the epi equipment market
will evOT grow to be much more than $100 million
per year worldwide without achieving a IOWCT film
cost.
Consequendy, many equipment maniifacturers
are loath to spend their development dollars on a
new reactor for such a small market. One company,
Moore Technology, is successfully addressing the
cost issue by redesigning existing systems to
improve the throughput of the reactor, but Dataquest does not ciqpect these systems to meet the epi
film specifications required for the 64Mb DRAM.
Future growth will depend on equqnnent vendors
resolving both performance and cost issues.
Kunio AcMwa (Tokyo)
Mark FitzGerald (San Jose)
Note: Odumu may not add to totals shonn because of rounding.
Souice: Dataquest (hpA 1991)
0010041
01991 Dataquest Incorporated April-Reproduction Prohibited
SEMMS Newsletten Equipment—^Depositian
Kspp^
DataQuest
fefi'-'v^'^'-'v''.^'-^
^:!fil:Ji^{t/^
aoHTipanyof
The Dun & Bradstreet Corporation
•'*'••!
!^''.^'-?(" ' • " ' • • • * i i ,
Wmw^^'^^r'
- V ^ & j ;
•••
^i/Sttk: .
Research Newsletter
DIFFUSION EQUIPMENT MARKET: 1990 IN REVIEW
SUMMARY
Dataquest estimates the 1990 worldwide
diffusion equipment market to be $322 million,
down 2 percent from $330 million in 1989. The
vertical thermal reactor (VTR)-based diffusion tube
segment grew explosively—67 percent from
$96 million in 1989 to $160 million in 1990.
However, the horizontal diffusion tube segment
declined dramatically by 31 percent from
$234 million in 1989 to $162 million in 1990. The
fast-growing VTR diffusion tube segment now
accounts for 50 percent of the worldwide diffusion
market Factors such as automation compatibility
with 200mm fabs, better process control, and small
equipment clean room footprint were responsible
for the high growth in the VTR diffusion market
segment
Japan, the largest regional diffusion equipment market accounted for 54 percent of the
1990 worldwide market Japan's thrust into 200nMn
0.5-micron 4Mb shrink/16Mb pilot line fabs
continues to sustain high growth for the Japanese
VTR diffusion equipment market. North America,
as a region, accounted for 24 percent of the 1990
worldwide diffusion market Relatively flat capital
spending conditions in North America, together
with the lack of activity in large-volume new fab
construction, led to a weak 1990 North American
market The European diffusion equipment market
represented 12 percent of the 1990 total, while the
Asia/Pacific-Rest of World (ROW) region
accounted for the remaining 10 percent of the
market
Japanese equipment companies dominated the
diffusion market with 52 percent ownership of the
1990 world market Japanese companies, with their
early entry into the VTR diffusion and VTR lowpressure chemical vapor deposition (LPCVD) market, controlled a major portion of the crucial VTR
market. North American diffusion companies,
which are currently aggressively entering the VTR
market owned 31 percent of the 1990 worldwide
diffusion market.
REGIONAL MARKETS
Figure 1 shows the 1990 diffusion equipment
market by region and ownership. Dataquest
includes diffusion, wet/dry oxidation, armeal,
implant drive-in, and BPSG glass reflow processes
within ttie diffusion equipment market The lowpressure tube CVD and plasma-enhanced tube
CVD processes are included within the overall
CVD equipnent market rattier than the, diffusion
tube market (see SEMMS newsletter entitled
"CVD Equipment: 1990 in Review").
Japan, the largest regional wafer fabrication
equipment market, was also the largest diffusion
equipment market with 54 percent of the
$322 million 1990 market. Japan is the leading
production region for high-volume advanced
devices such as 4Mb DRAMs, 1Mb SRAMs, gate
arrays, embedded microcontrollers for consumer
electronics, high-integration VLSI chip sets, and
systems logic for Japan's burgeoning computer
industry. Hence, the Japanese wafer fab equipment
market drives tiie requirements of high-throughput
and leading-edge process technology. It is no
surprise then that Japan constitutes the largest
regional market for all core segments of the wafer
fabrication equipment market such as steppers,
etchers, CVD, PVD, diffusion, and ion implant.
Japan's eariy shift to VTR diffusion tube
technology led to rapid increases in tube average
selling prices (ASPs) because of increased automation and process-control requirements in 150 and
2(X)mm submicron fabs. The Japanese diffusion
equipment market size increased in 1990 because
of an increase in unit shipments as well as tube
ASPs. VTR tube ASPs are higher in Japan because
01991 Dtuquest Incorponted April-Reproduction Prohibited
SEMMS Newsletters Equipment-DiSasion/Implant
0010241
The content of this report represents our irJerpretation and analysis frf information generally available to the public or released by responsible individuals in the subject companies, but
is not guaranteed as u> accuracy or completeness It does not contain material provided to us in corifidence by our clients Individual companies reported on and analyzed by Dataquest
Truly be clients of this and/or other Dataquest services This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of theirfomiUesmay, from time lo time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
DIFFUSION EQUIPMENT MARKET: 1900 IN REVIEW
FIGURE 1
1990 Diffusion Equipment Regional Markets and Ownership
(Revenue in Millions of Dollars)
European
Companies
$33
10%
Joint Ventures
$22.8
7%
Markets
Total = $322.4 Million
Ownership
Total = $322.4 Million
Source: Dataquest (April 1991)
of extensive fab automation requirements. Many
Japanese semiconductor manufacturers have
adopted VTRs for almost all diffusion operations
within the new 150 and 2(X)mm fabs buUt in 1990.
Spillover capital spending effects ftom 1989 into
1990, together with Japan's strategic manufacturing
focus, sustained Japan's 1990 difiusion equipment
market growth.
The North American diffusion equipment
market accounted for 24 percent of the 1990 worldwide market. North America, as a region, did
not add many large production fabs in 1990.
Fab capacity expansion, upgrades, and offshore
Japanese fabs in NorOi America accoimted for the
bulk of the 1990 maricet. North America is also
joining the move toward VTR diffusion and
LPC!VD systems, especially for new 200mm submicron applications. Captive semiconductor
manufacturers such as Digital Equipment (Corporation (DEQ, Hewlett-Packard (HP), and IBM were
more consistent in new fab addition and capacity
expansion purchases than were their merchant
counterparts.
North American semiconductor companies
have largely retreated from memory, low-er^ gate
arrays, and other high-volume, low-margin commodity devices. High device ASPs and lower
unit volumes for leading-edge devices such as
32-bit microprocessors and VLSI systems logic
have enabled North American semiconductor
manufacturers to achieve targeted revenue and
0010241
profitability levels with lower capital spending
expenditures. The new wave of fabless North
American semiconductor companies appears to
stretch this business model to the limit by completely eliminating wafer-fab-related capital spending and relying instead on foundry capacity in
Japan and the Asia/Pacific-ROW regioa Growth in
the diffusion tube market depends on multiple
orders from high-volume, leading-edge production
fabs. The North American regional diffusion equipment has not provided such an environment for
sustained growth in tiie last few years. However,
the growing presence of offshore Japanese fabs
may again stimulate growth in the North American
difiusion equipment market
The European diffusion equipment market
accounted for 12 percent of the 1990 market. Most
of the European market activities revolved around
offshore Japanese and North American fabs in
Europe. Companies such as DEC, Fujitsu, Hitachi,
IBM, Mitsubishi, and Texas Instruments (Tl) continued to expand their European operations in order
to comply with possible future requirements for
domestically difhised European semiconductors.
These offshore European fabs typically were clones
of parent North American and Japanese fabs. The
choice of a global diffusion equifxnent vendor
ensures a consistent process implementation and
automation strategy for global semiconductor
manufacturers with many regional fabs. Only
global diffusion equipment companies with a presence in all major semiconductor manufacturing
61991 Dttaqnett Incoqionted Apiil-4lepnMluction Prohibited
SEMMS Newfletten Equipment-DiifuiianAnipbnt
DIFFUSION EQUIPMENT MARKET: 1990 IN REVIEW
regions could hope to get a significant slice of the
1990 European diffusion market
The Asia/Pacific-ROW diffusion tnaiket contracted from 21 percent of the $330 million 1989
market to 10 percent of the $322 million 1990
maricet. Factors such as depressed DRAM prices
and excessive PC chip set competition together
with political and economic uncertainities were
responsible for the decline in the 1990 market
Dataquest believes that the 1991 Asia/Pacific-ROW
diffusion market should stage a moderate recovery
as new 16Mb pilot lines and submicron ASIC/
foimdry fabs are built.
REGIONAL OWNERSHIP
Figure 1 shows tiie worldwide 1990 diffusion
market byregionalownership. Japanese companies,
with 52 percent of the market, dominated tte 1990
diffusion market because of their lead in advanced
VTR production technology. Japanese diffusion
companies such as Tokyo Electron Limited (TEL)
and Kokusai Electric virtually owned the large
domestic Japanese VTR market The wave of new
200rrmi Japanese DRAM fabs ensured a flow of
large orders for technology-driven diffusion VTR
products. Japanese diffusion equipnent companies
also are increasingly penetrating international markets as they hasten to set up global business organizations in order to serve their global customers.
North American company ownership of
the worldwide diffusion maricet contracted from
35 percent in 1989 to 31 percent in 1990. North
American diffusion equipment companies made a
late transition from horizontal diffusion to vertical
diffusion technology. They ptirsued a strategy of
incremental improvements in their horizontal diffusion products in areas such as automation, particle
reduction, and process enhancement Meanwhile,
they underestimated the market momentum toward
vertical diffusion tubes, especially for submicron
new 200mm fabs. North American diffusion companies such as BTU International and Silicon
Valley Group (SVG) significanflyrepositionedtheir
product development efforts in 1990 as they
addressed the fast-growing vertical diffusion segment Dataquest expects North American diffusion
companies to vigorously challenge Japanese companies for dominance of the worldwide VTR
maricet
COMPANY RANKINGS
Table 1 shows tiie woridwide 1990 diffusion
equipment company rankings. Tlie rankings do not
include LPCVD or PECVD tube shipments. The
1990 diffusion market total was made up of
approximately 50 percent horizontal tubes and
50 percent vertical tubes.
TABLE 1
Worldwide 1990 Diffusion Equipment Company Rankings
(Millions of Dollars)
Company
Tokyo Electron Ltd.
Silicon Valley Group
Kokusai Electric
ASM International
BTU International
Ulvac/BTU
Koyo Lindberg
Centrothenn
General Signal ThinFilm
Gasonics
Others
Worldwide Market Total
Revenue
92.2
54.2
50.9
24.4
24.0
16.5
12.1
8.4
8.0
8.0
23.7
322.4
Market
Share (%)
28.6
16.8
15.8
7.6
7.4
5.1
3.8
2.6
2.5
2.5
7.4
100.0
Revenue Split by
Market Segment (%)
52 vertical, 48 horizontal
30 vertical, 70 horizontal
91 vertical. 9 horizontal
36 volical, 64 horizontal
25 votical, 75 horizontal
20 vertical, 80 horizontal
63 vertical, 37 horizontal
0 votical, 100 horizontal
63 votical, 37 horizontal
0 vmical, 100 horizontal
50 vertical, 50 horizontal
Note: The u U e shows calendar year 1990 system levenne; spares and service are not incbided.
Samce: Dataquest CApril 1991)
#
01991 Dauquest bicorparated April-Rqnoduction Piohilnted
SEMMS Newsletters Equipment-Difiiision/Implant
00m41
DIFFUSION EQUIPMENT MARKET: 1S90 IN REVIEW
TEL, the top-ranked company, captured
28.6 percent of the 1990 maikeL TEL's diffusion
unit shipments were balanced almost equally
between horizontal and vertical units. Dataquest
also separately accounts for TEL's North American
and European diffusion tube shiiments under the
Vaiian^'EL joint venture, which had diffusion revenue of $6.3 million in 1990. TEL's success in the
diffusion and LPCVD tube market prompted the
company to construa another laige-capacity VTR
tube manufacturing facility in Japan in 1990.
Silicon Valley Group (SVG), the secondranked company, owned 16.8 percent of the 1990
diffusion maiket. SVG continues to actively maiket
its Thermco horizontal diffusion systems for capacity expansion,replacement,and noncritical applications in new fabs. SVG's recent design win at the
Motorola MOS 11 200mm fab in Austin, Texas, is
an indication of the company's growing strength in
the North American diffusion maiket SVG's future
challenge is to penetrate the laiger international
market, especially in Japan and the Asia/PacificROW regioa
Kokusai Electric, with 15.8 percent of
the 1990 market, is the third-ranked diffusion
equipment company. The company was one of the
early pioneers in VTR diffusion and LPCVD technology. Since then, the company has successfully
launched several generations of VTRs into the
marketplace. Kokusai Electric is aggressively
addressing the international matket by establishing
regional customer support and applications facilities in the North American and European regional
maikets.
ASM International, with 7.6 percent of the
1990 market, was the fourth-ranked diffusion
equipment company. The company, which
dominates the horizontal PECVD tube maiket, has
recently focused its efforts on the fast-growing
diffusion and LPCVD VTR maiket. ASM Japan, a
wholly owned subsidiary of ASM International,
hopes to capture a slice of the competitive Japanese
VTR maiket by designing and building its VTR
systems through joint development with its
Japanese customers.
BTU International, the fifth-ranked 1990
diffusion company, owned 7.4 percent of the
market The company is executing a major product
transition from its older horizontal systems to vertical systems. The company recently ramped up
shipments of its VTR products. Sematech has
recently awarded equipment improvement contracts
to BTU International and SVG for their VTR
development projects.
DATAQUEST PERSPECTIVE
The $322 million 1990 woridwide diffusion
equipment maiket remained essentially flat compared with its $330 million 1989 level Although
the overall diffusion maiket remained flat between
1989 and 1990, the horizontal tube segment of the
market declined dramatically by 31 percent while
the vertical tube segment grew explosively by
67 percent The horizontal and vertical diffusion
market segments were approximately equal in
1990.
Japan, which is the largest maiket, has almost
completely switched over to vertical diffiision tubes
for its new 200inm sutxnicron fabs. Vertical tubes
have the advantages of smaller footprint, modular
automation, and better process control on 200mm
submicron difiiisioii/oxidation processes. Applications for vertical diffusion tubes include thin
gate oxidations, low-temperature anneals, and
BPSG reflow.
The North American, European, and Asia/
Pacific-ROW maikets are also switching over to
VTR diffusion processes for critical applications.
Dataquest believes that horizontal diffusion
processes will continue to be used in certain noncritical applications such as long, hi^-temperature
drive-ins and wet field oxidation. A distinct cost/
performance tradeoff exists betweoi vertical and
horizontal diffusion tubes. Semiconductor manufacturers appear to be leaning more toward the higherperformance characteristics of vertical diffusion
tubes for critical 200inm submicron fabs in spite of
significantly hi^er tube ASPs.
Krishna Shankar (San Jose)
Kunio AcMwa (Tokyo)
The topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, whidi include wafer fab equipment
sqipliers, semiconductor materials companies, and semiconductor device manufacturers. The topics sdected indicate die broad range of research
that is conducted in the SEMMS group. Oients, however. cSvea have specific infonnation requiremenu that either go beyond the levd of detail
contained in the newsletters or beyond the scope d[ what is nonnally published in the newsletters. In order to provide complete dedsian support
to our clients, Dataquest has a consulting service available to handle these additional infomuttion needs. Please call Stan Bruedede at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
0010241
61991 Dataquest Ineapoiated Apnl-KepiDdoetiaa PraUlnied
SEMMS Newilsttea Equipment-Diffuiiao/ImpUm
DataQuest
acoinpanyDf
ThcDartb BrKkuutcovpontion
Research Newsletter
C D SEM EQUIPMENT—SMALLER GEOMETRIES
LEAD TO LARGER MARKET
SUMMARY
W H Y CD SEM?
The critical dimension scanning electron
microscope (CD SEM) equipment market has
experienced robust growth during the past five
years, and, as shown in Figure 1, now exceeds the
optical CD equipment market by a sizable margin.
Dataquest's preliminary estimate of the 1990 CD
SEM equipment market is $100.6 million, up
almost 25 percent from its 1989 level of
$80.6 million. This healthy growth is all the more
impressive when measured against a total wafer
fabrication equipment market in 1990 that is
e^)ected to be down a few percentage points from
its 1989 level. This newsletter examines several of
the factors that are behind the emergence and
acceptance of CD SEM measurement equipment in
today's semiconductor manufacturing environment.
In the latter half of the 1980s, the field of CD
measurement diversified into a multitude of technologies. Historically, conventional CD tools have
been white-light microscopy systems. These systems are considered adequate for measurements
down to about 1.0-micron geometries. Several of
the white-light microscopy systems have been
enhanced with sophisticated image processing
c^abUities to extend their performance. In addition, laser-based measurement systems, confocal
scanning laser microscopy (CSLM), and
coherence-probe imaging (CPI) technologies have
been developed to perform CD measurements in
the submicron regime. CSLM and CPI systems
have received only modest maiket acceptance to
date because of the significant effort required to
FIGURE 1
Optical CD and CD SEM Equipment Markets
Millions of Dollars
120-,
r X I Optica! CD
100
^M
CD SEM
80
604020-1
0
1986
1987
1986
1989
1990
Note: 1990 l8 preliminary.
SoiiTce: DaUquest (December 1990)
O1990 Dataqiieft Inccnponted December-Repioductian PnduUted
SEMMS Newsletten Pioceis Coonol
0009130
The content qflhis report represents our interpretation and analysis (^information generally available to the public or released by responsible individuals in the subject companies, but
is na guaranteed as uj accuracy or completeness. It does not contain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataguest
may be clients of this and/or other Dataquest services. This iriformation is not jumished in connection with a sale or q^r to sell securities or in connection with the solicitation cfan
qf^r to buy securities. This firm and its parent and/or their t^ficers, stockholders, or members of theirfamiliesmay, from time to time, have a long or short position in the securities
mentioned attd may sell or buy stich securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
CD SEM EQUIPMENT—SMALLER GEOMETRIES LEAD TO LARGER MARKET
characterize the equipment for the measurement of
CDs in a production environment
At the same time, SEM tools, traditionally
relegated to the analytical lab, have been
redesigned to meet the needs of submicron
manufacturing for the production environment. The
SEM tools designed for IC metrology are lowvoltage systems because of the concem regarding
damage to the wafer at higher levels of electron
irradiation. In addition, equipment manufactures
have designed the tools to be more user friendly
than their analytical counterparts by simplifying the
operator control panel.
C D SEM ADVANTAGES
The advantages of CD SEM equipment
include better measurement resolution and depth of
focus than optical tools. Some manufacturers report
better-than-0.2-micron measurement capabiUty;
however, most agree that today's CD SEM tools
are fuUy characterized for production only down to
about 0.5-micron geometries. CD SEM equipment,
like the advanced optical techniques of CSLM and
CPI, also has the ability to capture threedimensional information of the line profile. The
slope of the sidewaUs becomes increasingly iiaportant in linewidth measurement as manufacturers
move to submicron geometries and features with
higher-aspect ratios necessitating tighter CD
control
Throughput
Throughput of CD SEM tools remains a
major issue. Conq)ared with optical tools, most CD
S ^ I tools still have relatively low throughput,
typically on the order of 8 to 12 wafers per hour at
five measurement sites per wafer. This is because
in most systems, wafers are processed serially
between the load lock and measurement chamber.
Several conq>anies have specifically incorporated
throughput enhancement features in their design to
overccnne this factor. Opal (a subsidiary of printed
circuit board inspection manufacturer, Optrotech)
increases system throughput by measuring one
wafer while a second ws^ex is being punq>ed down
in the load lock. An internal exchange unit allows
the first wafer to move aside upon completion of
its CD measurements so diat the second wafer can
be moved from the load lock to the measurement
chamber. Opal claims a throughput of iqTproximately 20 wafers per hour and expects it to be even
0009130
higher in the future. Nanoquest (the former Nickers
Instruments operation acquired by BioRad in 1989)
takes a different approach; its system accepts a fuU
cassette of wafers into the load lock rather than a
single wafer. Thefrillcassette is pumped down, and
a wafer transport mechanism is used to move
individual wafers into the measurement chamber.
REGIONAL MANUFACTURING PRACTICES
One of the interesting aspects of the CD
measurement equipment market is that some semiconductor manufacturers tend to perform less
measurement and inspection than other companies.
These conq>anies have adopted a manufacturing
philosophy to conq}letely characterize and understand their process in the R&D environment prior
to moving die device into high-volume production.
Once the device is fully characterized for production, only mioimal measurement and inspection is
performeid to monitor the fabrication process. Thus,
fewer measurement tools are needed. This is in
contrast to the practice of characterizing the process in a production mode, which requires more
measurements and adjustments on the fly more
frequently.
Figure 2 illustrates that ia the last several
years, regional variations have emeged in the use
of CD measurement equipment In 1986, manufacturers in the United States and Jj^an accounted for
almost equal share of the worldwide wafer fab
equqnnent market as well as almost equal share of
the combined optical and CD SEM equipment market By 1989, however, this situation had changed
substantially. Although the United States as a
region accounted for 29 percent of the wafer fab
equqnnent market it represented 38 percent of aU
spending on CD measurement systems. In contrast,
Japsax accounted for 45 percent of the wafer fab
equipment market, but only 40 percent of the CD
equipment market.
Even fruther regional manufacturing distinctions exist within the category of CD measurement
equ^mient Manufacturers in J^an use significantly more CD SEM systems than do their counterparts in the United States. Figure 3 shows that in
1986, semiconductor manufacturers in Jiq)an and
the United States accounted for approximately
equal share of both the optical CD and CD SEM
equqnnent markets. In 1989, the United States
spent more on optical CD measuremoit equipment
with 44 percent share of the world market while
Js^an had only 27 percent share. Js^an, however,
01990 Dataqueit Incoiporated December-Reproduction Profaibited
SEMMS Newtletters Proceu Conbol
CD SEM EQUIPMENT—SMALLER GEOMETRIES LEAD TO LARGER MARKET
FiGUKE 2
Regional Equipment Market Trends
United States and Japan, 1986 and 1989
Percentage of Dollars
60Unitad States
50
Japan
40-1
50
20
10-1
0
Wafer Fab
Equipment
Optical and CD SEM
Wafer Fab
Equipment
Optical and CD SEM
1989
1986
Source: Dataquest (December 1990)
FIGURE 3
Regional TVends—CD Equipment Market
United States and Japan, 1986 and 1989
Percentage of Dollars
60United States
50
Japan
40
30
20
10
0
^^ ^ ^
Optical CD
CD SEM
1986
Optical CD
CD SEM
1989
Source: Dataquest (December 1990)
accounted for over one-half of the CD SEM equq>ment market in 1989 with 51 percent shate, in
contrast wifli U.S. share of 33 percent
This move by Japanese manufacturers to CD
SEM equipment is due, in part, to die prevalence of
DRAM manufiacturing in Japan, which is the technology driver for processing submicron geometries.
01990 Dataqueit Incoiponted December-Reproductian Prohibited
SEMMS Newdettera Praceu Contiol
Dataquest believes, however, that Jq>anese semiconductor manufactums have chosen to leapfirog
the advanced optical CD measurement technologies
and move directly to SEMs because diey are not
convinced yet that advanced optical technologies
can be pushed to or beyond the 0.5-micron processing regime. There are also concerns that the
0009130
CD SEM EQUIPMENT—SMALLER GEOMETRIES LEAD TO LARGER MARKET
advanced optical tools have not been characterized
fiiUy for the semiconductor production environment. Finally, semiconductor manufacturers in
Japan historically have been supported by a strong
domestic vendor base in CD SEM.
C D SEM COMPANIES
Many companies currently are pursuing the
CD SEM equipment market, including the following U.S. companies: Amray, Angstrom Measurements, Metrologix, Nanometrics, Nanoquest, and
Opal. Japanese CD SEM conq)anies include Akashi
Beam Technology (recendy acquired by Toshiba),
Hitachi, Holon, and JEOL. This market includes
well-established equipment conq>ames as well as
start-ups. Hitachi, however, held and maintained
dominant market share throughout much of die
1980s. As shown in Table 1, Dataquest estimates
that Hitachi commanded 75 percent of world market share in 1990. In 1986, when the world madcet
for CD SEM equipment was only $15.4 million,
Hitachi still accounted for more than one-half of
the market with 56 percent share. Dataquest
believes that Hitachi's success is due, in part, to the
company's extensive e:q>erience in e-beam technology for electronics as well as other q>pUcations, in
addition to its early product focus on developing a
user-Mendly CD-SEM tool for semiconductor
production {q>phcations.
TABLE 1
1990 CD SEM Equipment Company
Preliminary Market Share
(Millions of Dollars)
Revenue
Percent
Share
Hitachi (Japan)
75.5
75.0
HoloD (Japan)
9.4
9.3
Nanoquest (U.S.)
6.9
6.9
Nanometrics (U.S.)
3.1
3.1
Opal (U.S.)
2.4
2.4
Others
3.3
3.3
100.6
100.0
Company
Total
Somoe: Dataqaest (Deceniber 1990)
0009130
N E W oppoFiTUNmES—CLUSTER T O O L
PROCESSING
As hnewidth geometries continue to shrink,
the overall market opportunities for CD SEMs are
growing. One of the interesting opportunities in the
CD SEM equipment market comes from the
developing market for cluster tools. CD SEM
measurement equipment is particularly well suited
for a cluster tool vacuum environment designed for
etch, strip, and deposition processes. Linewidth
measurement would be performed on the wafer
after etch/strip processing. The wafer then would
be moved directly to a deposition module, thus
eliminating the need to remove the wafer from the
cluster tool for CD measurement prior to deposition. Metrologix, acquired by venture capital firm
Nazem and Company in July 1990, is well suited to
pursue such a strategy because of its association
with Tegal. Tegal, anodier company funded by
Nazem, is a well-established player in the plasma
etch and strq) equipment markets.
DATAQUEST FORECAST
Dataquest anticipates CD SEM tools will continue to experience healthy growth in the years to
come as a larger percentage of the semiconductor
device product mix moves into the submicron
processing regime. CD SEM measurement technology has already gained widespread accqptance in
Japan, the largest semiconductor manufacturing
region in the world. CD SEM equqnnent is establishing a presence infront-endmanufacturing in the
other manufacturing regions of the world as well.
Dataquest e:q)ects the CD SEM equipmrait maiket
to be iq>proximately $245 million by 1995, reflecting a 19.5 percent conqiound annual growth rate
(CAGR) between 1990 and 1995.
Dataquest notes, however, that optical CD
measurement tools are not likely to disiqppear
entirely. Several of the advanced optical tools have
been designed specifically for submicron measurement performance and thus wiU comptto directly
with CD SEM. Conventional optical CD tools still
provide a cost-effective, high-throughput option for
the measurement of l.O-nadcron geometries and
larger. Finally, optical CD tools still will be
required to perform overlay measurements in most
applications. CD SEM equqnnent is not particularly well suited for overlay measurement because
of the physics of the measurranent procedure. An
optical tool can "see"ttirougha transparent film to
the alignment marks on an underlying layer. CD
O1990 Dataquest Incoiparateil Deceniber-ReinDduetian Piohibited
SEMMS Newiletten Proceii Cootiol
CO SEM EQUIPMENT—SMALLER GEOMETRIES LEAD TO LARGER MARKET
SEM measurement technology primaiily relies on
secondary electrons scattered off the wafer surface
to determine its measurement signal, and thus, in
most applications, is restricted to the measurement
of surface features. Therefore, optical CD tools
with joint linewidth and overlay measurement
capabilities or dedicated overlay tools will continue
to be purchased in the 1990s. Dataquest forecasts
the optical CD equipment market to be $105 million in 1995, reflecting a five-year CAGR of
10.1 percent.
DATAQUEST CONCLUSIONS
A single conc^any has dominated the CD
SEM equipment market to date. The other nine
con^anies in this market segment face significant
challenges. They must overcome the "play it safe"
attitude of semiconductor manufacturers diat chose
to buy from the market leader. These companies
must establish a market presence strong enough to
allow them to generate a sufficient income stream
to invest in future technology development. At the
same time, they must expand their international
operations. Partnershq)s and alliances with larger
equipment companies can provide the support that
some of these companies will need to nurture longtemi growth. For the U.S. equipment vendors of
CD SEM tools, the Ji^anese market wiU be particularly diflScult to penetrate because of the overwhelming strength of the domestic vendor base and
the significant cost of doing business in Japan.
Opportunities, however, always exist for a company able to sell, service, and support its equipment
in an increasingly demanding customer base.
Kunio Achiwa
Peggy Marie Wood
The topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, vtiiich include wafer fab equmment
suppliers, semicondnctor mnterials coiiq>anies, and semiconductor device mannfactiiTers. The topics selected indicale (he broad range of research
that is conducted in the SEMMS group. Clients, however, often have specific infonnatian requirements that either go beyond the level of detail
contained in the newsletters or beyond the scope of what is nomally pubUshed in tbe newsletters. In order to provide conq>lete decision siqqxnt
to our clients, Dataquest has a consulting service available to handle these additional infonnatian needs. Please call Stan Bmederle at (408)
437-8272 or Joe Gienier at (408) 437-8206 to discnss your cnstom requirements.
O1990 Dataquest Incacixiiated December-Jleproducttoa Prohibited
SEMMS Newiletten Pioceii Control
0009J30
^PMf
Dataoyest
m;^
a company of
T I K Dun & Bradsticct Corporation
Research Newsletter
MERCHANT DRAM SUPPLIERS: ANOTHER SHAKEOUT COMING?
The prices of 1Mb and 4Mb DRAMs have
been falling steadily. In the fourth quarter of 1989,
the worldwide average selling price (ASP) of a
1Mb DRAM was $9.45. Today, the price of a 1Mb
DRAM ranges from $4.30 to $5.00. In the fourth
quarter of 1989, the ASP of a 4Mb DRAM was
$87.78. Today, the price of a 4Mb DRAM ranges
from $18.00 to $23.00.
Part of this decline is normal. Learning-curve
price declines are a part of each generation of
DRAMs. These declines stimulate demand and
allow buyers to economically cross over to the next
generation of DRAMs. However, in the fourth
quarter of 1990, 1Mb DRAM price declines
seemed much sharper than would be expected
from learning-curve experience only. Prices were
squeezing profits. Japanese DRAM manufacturers responded with production cutbacks of
1Mb DRAMs.
These recent and larger-than-expected
1Mb DRAM price declines are due to a simple
economic fact: oversupply. Based on our analysis
of existing merchant capacity and planned, publicly
announced capacity additions, Dataquest believes
that there is an oversupply of 1Mb and
4Mb DRAMs. We believe that this condition is
likely to continue.
D R A M SUPPLY AND DEMAND
Dataquest maintains a worldwide fab database
that contaiiis wafer start capacity for individual
semiconductor companies. From our fab database
we determine current DRAM capacity in wafer
starts and add to it all the announced plans for
bmlding future DRAM Ciqpacity. This DRAM wafer
start cq>acity is converted to DRAM unit capacity
by applying the set of assumptions for die size and
yield shown in Table 1. Unit DRAM capacity is
then cotr^ared with Dataquest's forecast of DRAM
demand to determine if there is or will be an
imbalance of DRAM demand and supply.
We believe that our capacity assumptions are
on the conservative side. For example, all fab lines
that produce non-DRAM devices in addition to
DRAMs were excluded from the analysis. If these
fabs and some appropriate fraction of tiieir capacity
were included in the analysis, the resulting DRAM
cq>acity numbers would be much higher.
Many fabs in our database are listed as e n a ble of producing either 1Mb or 4Mb DRAMs.
Thus, this study is based on an aggregate analysis
of 1Mb and 4Mb DRAM capacity. In this newsletter, we provide three scenarios regarding the mix of
fabs capable of either 1Mb or 4Mb DRAM production. Tliese scenarios reflect assumptions for low
TABLE 1
Assumptions for DRAM Die Size, Probe Yields, and Good Die Per Wafer
Die
Size
(mm^
Probe
Yield (%)
Good Die
per 6-Inch
Wafer
1Mb DRAM
40
79
293
4Mb DRAM
96
48
66
4Mb DRAM (Shrink)
71
65
128
Source: DaUujuest (March 1991)
0 1 9 9 1 Dataquest Incorporated March-Reproduction Prohibited
SBMMS Newiletters Manufacturing: Semiconductor Mamifacturing/Industiy Trends
0009642
The content of this report represents our interpretation and analysis t^ 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 Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dtitaquest services This information is notfitmished in connection with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy sitch securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
MERCHANT DRAM SUPPLIERS: ANOTHER SHAKEOUT COMING?
producing both 1Mb and 4Mb DRAMs will
actually shift production to 4Mb DRAMs.
DRAM demand and our estimated DRAM
c£^acity for the three scenarios are shown in
Figure 1. What is striking about this figure is that
the combined capacity of 1Mb and 4Mb DRAMs—
under all three of our cq>acity scenarios—exceeds
the projected combined demand for these devices
by a substantial margin. For example, in 1994, the
low-capacity scenario projects a capacity of slightly
over 2,500 million units and a demand of
just under 2,000 million units, an excess of
over 25 percent.
Table 2 evaluates the supply-and-demand estimates in Figure 1 in terms of percentages. For
example, in 1994, demand will be only 66 percent
of the potential supply under the high-capacity
scenario and 74 percent of supply imder the lowcq)acity scenario.
capacity, high capacity, and intermediate capacity
for 1Mb and 4Mb DRAMs.
The low-capacity scenario assumes that all
1Mb DRAM fabs that are also capable of
4Mb DRAM production will indeed shift their
production to 4Mb DRAMs. This scenario is low
capacity because fewer 4Mb DRAM die can be
fabricated from a wafer than 1Mb DRAMs. (The
die for 4Mb DRAMs are larger than the die for
1Mb DRAMs.)
The high-c^acity scenario assumes that none
of the fabs listed as capable of either 1Mb
or 4Mb DRAMs will shift production to
4Mb DRAMs. This scenario is high capacity
because more 1Mb DRAM die than 4Mb DRAM
die can be fabricated from a wafer.
The intermediate-capacity scenario assumes
that 50 percent of the fabs listed as capable of
FIGURE 1
Estimated Aggregate of 1Mb and 4Mb DRAM
Supply and Demand
Millions of Units
3200T^
1991
1990
-t«B4
Source: Dataquest (Maich 1991)
TABLE 2
Estimated Merchant Demand as a Percentage of Worldwide Merchant Capacity
for 1Mb and 4Mb DRAMs Combined
Low
Capacity
Intermediate
Capacity
High
Capacity
1990
63
52
40
1991
77
63
53
1994
74
70
66
Source: Dataipiest (March 1991)
0009642
01991 Dataquest Incorporated Maicb-Reproduction Prohibited
SEMMS Newsletters Manufacturing: Semiconductor Manufacturing/Industry Trends
I
MERCHANT DRAM SUPPLIERS: ANOTHER SHAKEOUT COMING?
DATAQUEST CONCLUSIONS
It should be emphasized that this large
DRAM capacity has to be utilized for oversupply
to occur. In the short term, DRAM producers could
very weU choose not to use their full-production
capacity. This happened in fall 1990 when Ji^anese
producers announced cutbacks in production of
1Mb DRAMs.
A longer-term strategy to reduce overc^acity
is to switch some DRAM capacity to other
products, such as SRAMs or ASICs. An example
of this strategy would be Motorola's recent
announcement that its MOS 11 fab in Oak Hill,
Texas, wUl be used to produce SRAMs rather than
4Mb DRAMs as originally planned. A third alternative strategy is that a DRAM fab facing overcapacity could aggressively pursue foundry relationships to fill unused capacity.
In the early and mid-1980s the industry faced
a similar, although not identical, situation. There
were too many DRAM manufacturers, and oversupply resulted. Many DRAM producers decided to
leave the DRAM business for what they hoped
were more profitable product lines. Capital spending fell more sharply than ever before. Equ^nnent
con:q)anies folded, merged, restructured, laid off
staff, and some even cut back on R&D.
Since the mid-1980s, the industry has been
much more circumspect about adding capacity. The
growth rate of capital spending was much less in
the second half of the 1980s than in the first half.
Inventories, because of just-in-time deliveries and
closer suppUerA^endor relationships, are much better managed than in 1985. Since the shakeout in the
mid-1980s, the industry has matured.
However, new DRAM players have emerged
since the mid-1980s. These companies (Motorola,
Asia/Pacific companies, and second-tier Japanese
companies) have added capacity in order to gain
market share. The capacity from the new, plus the
capacity from the established players, today add up
to overcapacity. Clearly, the industry faces a challenge: how to manage overcapacity without the
corrective of another shakeout.
01991 Dataquest Incorporated Marcb-Reproduction Probibited
SEMNfS Newsletters Manufacturing: Semiconductor Manufactuiing/Industiy Trends
George Burns
0009642
mm
'^'^a^^Tv^'f-
r!^'^.:A''<:
DataQuest
acom(»nyof
The Dun & Bradstiect Cwporation
Research Newsletter
WAFER SIZE AND MANUFACTURING COSTS:
THE PUSH TO LARGER WAFERS
INTRODUCTION
The ratio of initial fab cost to monthly square-inch
capacity decreased from $422 per square inch to
$398 per square inch—a decrease of 6 percent. In
other words, although the capital cost of a new fab
increased dramatically from 1970 to 1987, capital
cost per unit of capacity actually decreased, albeit
slightly.
This decrease in itiitial capital cost per unit of
capacity, in spite of increasing process complexity,
was due to doubling the number of wafers a fab
could produce and doubling the size of the wafer.
Increasing wafer size dramatically increases the
area of a wafer (see Table 2). Thus, from 1970 to
1987, wafer size doubled from 3 to 6 inches, and
wafer capacity also doubled from 10,000 to 20,000
wafers per month. This doubling of both wafer size
and cq>acity offsets the increasing costs of c£q}ital
because of increasing process complexity.
However, since 1987, initial fab costs for
6-inch wafer fabs have continued to rise while
capacity has typically remained constant. As a
result, tiie ratio of initial fab costs to square-inch
capacity has increased since 1987. (For more
information on the causes of rising fab costs, see
Dataquest's Dataquest's SEMMS newsletter
entitled "Technology Trends and Fab Costs,"
November 1990.)
Dataquest believes that large manufacturing
cost gains can be made by using larger wafers in
high-voliune production. We believe that this is
true today, as manufacturers begin to move from
using 6-inch wafers to using 8-inch wafers, and it
will be true in the year 2000, when the industry
moves from 8- to 12-inch wafers.
This newsletter highhghts the manufacturing
gains associated with using larger wafer sizes in
volume production (20,000 wafer starts per month
in today's state-of-the-art facility). Our analysis
shows that the increased output outweighs any
increases in capital costs associated with using
larger wafers.
HISTORICAL FAB COSTS AND CAPACITY
In 1970, the initial coital cost (facility and
equipment) of a high-volume, state-of-the-art fab
was $30 million. By 1987, this cost had risen to
$225 million (see Table 1). During this same period, typical fab ci^acity as measured in square
inches of silicon per month rose from $70 million
to $565 million—a rate of increase slightly
exceeding the growth of initial fab costs.
TABLE 1
Fab Cost per Monthly Square Inch of Capacity
Monthly Capacity
Fab Cost per
Square Inch ($)
Year
Fab Cost
(Millions of Dollars)
1970
30
10,000 (3-inch)
71,000
422
1987
225
20,000 (6-inch)
565,000
398
1990
295
20,000 (6-inch)
565,000
522
Wafers
Square Inches
Souice: Dataquest <^ril 1991)
01991 Dataquest Incoiporated Apiil-4(eproduction Prohibited
SEMMS Manufacturing—Semiconductor Manufacturing/Industry Trends
0010002
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 amipanies, but
is not guaranteed as to accuracy or completeness It does not contain material provided to us in confidence by our clients Individual cmqmnies reported on arul analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
q^r to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
WAFER SIZE AND MANUFACTURING COSTS: THE PUSH TO LARGER WAFERS
TABLE 2
Wafer Size and Area
Wafer Size (Inches)
Area (In )
Ratio of Area to
6-Inch Wafer Area
6
28
1.0
8
50
1.8
10
79
2.8
12
113
4.0
Souice: Dataquest {Ajail 1991)
COSTS AND WAFER SIZE
However, semiconductor manufacturers need
not be helpless in the face of increasing ci^ital
costs. Our analysis indicates that if semiconductor
manufacturers increase the size of their wafers,
they will slow the rise in initial fab cost per square
inch of silicon capacity. And, perhaps more important, using larger wafers wiU lower manufacturing
cost per die.
Initial Capital Cost and
Larger Wafer Sizes
Our estimates of 1Mb and 4Mb DRAM fab
costs are shown in Figure 1. We have provided
initial fab costs for 6- and 8-inch fabs for both
generations of DRAMs. We have normalized fab
capacity for both 6- and 8-inch fabs to 20,(XX)
wafer starts per month. As would be expected, fab
costs rise with each generation of DRAMs. Fab
costs also rise within a DRAM generation as larger
wafers are used.
As Figure 2 shows, however, within a DRAM
generation, fab costs per square inch of capacity
decline as larger wafers are used. For example,
initial facility and equipment cost per square inch
of capacity for a 20,0()()-wafer-start-per-month fab
making 4Mb DRAMs is over $500 per square inch
if 6-inch wafers are used, but falls to under
$400 per square inch if 8-inch wafers are used.
Because the area of the wafer increases much faster
than does cq)ital cost, initial capital cost per square
inch of silicon capacity decreases when larger
wafers are used.
FIGURE l
DRAM Fab Costs by Wafer Size
Equipment and Facilities (Millions of Dollars)
450400
350-
[ys^ 6-lnch
e-lnch
300
250
200-1
150
100
500
4Mb DRAM
1Mb DRAM
DRAM Generation
Soiwce: Dataquest (April 1991)
0010002
01991 Dataqueit Incoiponted Apiil-Reptoduction PiobiUted
SEMMS Manuficturuig—Semicoiiductor Manufactuiing/lBdiutry Tiend*
i
WAFER SIZE AND MANUFACTURING COSTS: THE PUSH TO LARGER WAFERS
FIGURE 2
Capital Cost/In* of a 20,000 Wafers/Mo. 4Mb DRAM Fab by Wafer Size
Dollar per Square Inch of Start Capacity per Month
600540
480
420
360
300
240180120
60
0
6-lnch Wafers
Wafer Size
Source: Dataquest (April 1991)
Manufacturing Cost per Die and
Wafer Size
Dataquest has a DRAM manufacturing cost
model capable of analyzing DRAM die cost
manufactured on 6-, 8-, or even 10-inch wafers.
Manufacturing cost (for exanqile, labor, materials,
and overhead) per good die at a given yield is
shown in Figure 3 for 6-, 8-, and 10-inch wafers.
Although no 10-inch 4Mb DRAM liires exist (and
it is urJikely that they ever will), we ran the model
using 10-inch wafers for the sake of iUustration.
Manufacturing cost per good die declines as wafer
size increases, because the number of die at a
constant jrield increases with wafer size much more
rapidly than materials cost, depreciation, or any
other manufacturing cost variable. The semiconductor mantifacturer that can maintain yields and
move to larger wafer sizes will have a tremendous
cost advantage over those that stay with the smaller
wafers.
INCREASING WAFER SIZE INCREASES
PROCESS COMPLEXITY
The main focus of this newsletter has been on
the cost advantages of moving to larger wafer
sizes. However, such a move is not easy; some
major processing hurdles have to be overcome. For
exmaple, stepper depth of focus is sensitive to
variations in flatness across the greater areas
01991 Dstaquest Incorporated April—Reproductian Prohibited
SEMMS Manufacturing—Semiconductor Manufacturing/Industry Trends
of larger wafers. The surface area of a larger wafer
also presents uniformity problems for deposition,
diffusion, and etch equipment. Robotics and waferhandhng capabilities for all equipment have to be
reconfigured and upgraded in order to handle
wafers that are both larger and substantially
heavier.
Solving these and other problems wiU cost
equipment and materials vendors time, effort, and
money and require much effort. Additionally, after
the equipment and materials are available, semiconductor manufacturers will have to transfer their
working processes to the larger wafers, which also
requires time, effort, and money. It is an unfortunate but true fact that semiconductor manufacturers that are flie leaders in moving to a larger
wafer size pay a price for their leadership: an initial
loss of yield.
DATAQUEST CONCLUSIONS
This time, effort, and money are necessary to
transfer a successful process to a larger wafer size;
yet, because substantial savings are involved, we
beUeve that the move to larger wafer sizes is
inevitable. Semiconductor manufacturers can
achieve substantial savings in initial fab cost per
square inch of capacity and in the manufactured
cost of a yielded die by using larger wafers.
Although for years IBM was the only major
manufacturer using 8-inch wafers, today other
0010002
WAFER SIZE AND MANUFACTURING COSTS: THE PUSH TO LARGER WAFERS
FIGURE 3
DRAM Cost per Die by Wafer Size
Cost per Die (Dollars)
66-Inch
543
8-lncti
^^n
ID-Inch
2
1
0
m^m
C^KN^SKK-
4Mb DRAM
Source: Dataquest (April 1991)
companies are announcing plans to build 8-inch
facilities. The industry is continuiag its historic
move to larger wafers. Indeed, by the mid-1990s,
Dataquest expects to see 10-inch or, more likely,
12-inch pilot fab Unes announced. Because there
was a time lag of three to five years between the
&st appearance of 8-inch wafers and widespread
use of 8-inch wafers, we do not expect to see
widespread use of 12-inch wafers until about the
year 2000.
Because it takes six to seven years to develop
a new process, those conq>ames (for example,
materials suppliers, equipment vendors, and
leading-edge semiconductor manufacturers) that
wish to remain ahead of dieir competition should
be looking to the next wafer size today.
In addition to pushing so noany other limits
(for example, lithogriq)hy, interconnect, deposition,
contamination) by the end of this decade, the
industry will be fabricating devices on a 12-inch
wafer—a surface that is 225 percent larger than an
8-inch wafer surface. It is a much larger plajdng
field, and its size will also be reflected in the size
of R&D budgets in the conmng years.
George Bums
The topics covered by SEMMS newsletters are selected for their gooeral interest to SEMMS clients, which include wafer fab equipment
suppliers, semiconductor materials coiiq)anies, and semiconductor device maimfactnTas. The topics selected indicate the broad range of research
that is conducted in the SEMMS groq). Clients, however, often have specific mformation requiremoits that either go beyond the level of detail
contained in the newsletters or beyond the scope of wbat is normally published in flie newsletters. In order to provide coiiq>lete decision support
to our clients. Dataquest has a consnltiiig service available to handle these additicmal information needs. Please call Stan Broederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
0010002
01991 Diuiiueft Incoiporated April-Reproduction Prohibited
SEMMS Mmn&wtuiiiig—Semicondnctor Manufactuiing/Ihduftiy Tteait
i
•ifji-^-a
M^
mm
Iv.:;;
iL.''}-.'---i,'.-''
lass
fe^^E:
Dataquest
a company of
ihc Dun & Brad^icet Coqioration
Research Newsletter
THE ROLE OF AUTOMATION IN SEMICONDUCTOR MANUFACTURING
SUMMARY
The purpose of this research newsletter is to
highlight some of the major development activity
that is currently being undertaken in the area of
semiconductor factory automation. Increasingly
large automation investments are being made in all
regions of the world, especially Japan. These
investments are being directed toward robotics, cell
controllers, and automated data collection syst^ns
in an effort to help manufacturing managers better
manage their fabs. Semiconductor manufacturers
plan to realize positive returns on their automation
investments through increased equipment utilization, reduced cycle times, improved yields,
increased labor productivity, elimination of
misprocessing, and inventory reduction.
These systems contribute to increased fab productivity by performing wafer transfer ntuichine load/
unload functions and through effective work in
process (WIP) managonent. Robotics designed for
Class 10 clean room environments are being used
extensively in Japan for both inter-bay and intrabay material transport. At least one U.S. robotics
manufacturer recendy developed a cell automation
system that will provide cassette-to-cassette automation for an entire cell of process equipment or
throughout the complete fab. The cell may contain
equipment such as a preprocess wafer-cleaning
tool, a stepper, a post-process metrology station,
and a WIP management station to disposition the
wafers after cell processing.
Cell Controllers
DEVELOPMENT ACTIVITY
The Robotics Market
Robotics are playing an increasingly important role in factory automation because submicron
geometries mandate that manufacturing managers
maintain tighter control over their processes.
Tights control will be achieved by using equipment that requires minimum operator intervention,
which, in turn, will decrease wafer misprocessing,
reduce contamination, and reduce cycle time.
Reduced cycle time is important for three reasons:
the probability of contamination increases the
longer the processing time, a reduction in cycle
time translates into lower manufacturing costs, and
yield learning rates improve.
The increasing demand for robots in the
manufacturing process is pushing robotic technology forward. For example, robotics are a key part
of automated material control/transfer systems.
A cell is a set of process equipment, materialhandling equipment, and metrology equipment
designed to perform related functions under central
control. The ceU controller is a combination of
software and hardware tools that manage and coordinate cell resources by acquiring data and adjusting operations based on the results of that data.
Cell-control software developers are challenged to
develop software that is compatible across computers with different architecture. Conq)atibility is
necessaiy to accommodate the diversified computer
product line that exists today.
Semiconductor manufacturers are looking for
equipment that has the capability to operate in an
"independent" manner. Independent operation
refra°s to having equipment that is citable of shutting down and restarting through the use of remote
control. The restart capability should include autoniatic initialization of both the equipnent status
and fault-detection systems. WiSi these systems
initialized at start-up, the equipment will automatically return to a steady state that is ready for
processing.
101991 Dataquest Incoipotated March-Reproduction Prohibited
SEMMS Newsletters 1991 Sennconductor Manufacturing/Industry Trends
0009575
The content of this report represents our interpretation and analysis of information generally available to the ptibllc 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 conjidence by our clients. Individual companies reported on and analyzed by Datctquest
may be clients of this and/or other Dataquest services. This information is not furnished in conttection with a sale or q^r to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of their fomilies may, from lime to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE ROLE OF AUTOMATION IN SEMICONDUCTOR MANUFACTURING
Automated Data Collection
The demand for automated data collection
systems will increase as process complexity
increases (over 500 process steps may be required
to manufacture a l ^ I b DRAM) and line geometries get smaller. Automated data collection systems are designed to help manage process complexity, improve yield through statistical process
control, and manage inventories more tightly
through WIP management programs.
Bar code systems represent one of the largest
growth areas in automated data collection. The
demand for bar code systems is derived from a
low-cost product that is easy to install and maintain
in a manufacturing environment.
THE INTELLIGENT MANUFACTURING
SYSTEMS PROJECT
The J^anese Ministry of International Trade
and Industry (Mm) detramined that there was a
demand for factory automation systems that have
the capability to integrate across companies. As a
result, MlTl formed an international cooperative
effort called the Intelligent Manufacturing Systems
(IMS) project. The purpose of the project is to
formulate an integration strategy. The effort is
financed equally by the particq)ants, and its mission is to develop intelligent, advanced con^uteraided manufacturing systems. Projects currently
being considered include zero-defect quality programs and concurrent engineering. The key to IMS
success will be the equitable exchange of technology between the particq>ants.
CiM DEVELOPMENT AT SEMATECH
SEMATECH is also woridng on an advanced
conq>uter-integrated manufacturing (CIM) architecture program that is scheduled for demonstration
and transfer of ci^ability to its member companies
during the diiid quarter of 1991. The purpose of
the project is to develop and adopt a common
0009575
architecture for the integration of CIM system
elements and provide a framework for technology transfer to its monber con^anies.
AUTOMATION BARRIERS
Common barriers to autonoation include lack
of investment cqntal, lack of commercially available hardware, nonstandard equipment intofaces,
the inability to quantify benefits of automation to
senior management to overcome investment hurdles, and the lack of qualified automation hardware
and software engineers. The demand for engineers
is being driven by manufacturers that want factory
automation suppliers to train their own engineers as
well as provide on-site support.
DATAQUEST CONCLUSIONS
Factory automation is feasible and practical
for the semiconductor industry. The majority of
Japanese semiconductor manufacturers view factory automation as a strategic necessity. For
instance, the majority of new O.Sum Japanese
DRAM fabs will have both inter-bay and intra-bay
automated material-handling systems. The same
material-handling systems are also being installed
in their ASIC fabs. As wafer sizes increase, WIP
inventories will be extremely expensive, so lengthy
throughput times and large WIP inventories will tie
up equally large amounts of working capital. Factory automation can help control these costs. Dataquest believes that factory automation has the
potential to give semiconductor manufacturers who
exploit its c{q)abilities a competitive advantage in
the 1990s by allowing those manufacturers to
maintain tighter control of their manufacturing
processes. The s^niconductor industry's capitalintensive nature demands competitive yields, high
throughput rates, and maximum utilization of
resources to achieve a competitive cost per wafer.
Jeff Seerley
01991 Dataqueit Incoiponted Mucb-Reproductiom Prohibited
SEMMS Newsletten 1991 Semiconductor Mamifactuiing/Induitiy Tiendi
DataQuest
'Smi
•9^' "V? 1E8I
TS^&vklnctaiiponUton
^
MM
Research Newsktter
SEMATECH'S CONGRESSIONAL REVIEW: IS THE BEST IT CAN
BE GOOD ENOUGH?
"SEMATECH could meet all of its R&D
objectives on schedule and still notrestoreU.S.
manufacturing leadership." SEMATECH 1990: A
Report to Congress
SUMMARY
The National Defense Authorizatioo Act for
fiscal years 1988 and 1989 established the
Advisoiy Council on federal participation in
SEMATECH and chaiged it with, in the council's
words, "reviewing SEMATECHf's operations each
year and assessing continued federal participation."
The Advisory Council, chaired by John A. Betti,
Under Secretary of Defense for Acquisitions,
has once again recommended to Congress that
the federal government continue its funding
of America's most visible high-technology
consortium.
The Advisoiy Council report, however, also
raises issues about SEMATECH'S long-term effectiveness. In particular, the report expresses
concems about the following:
• SEMATECH has not been a sufficient antidote
to continuing erosion in the U.S. semiconductor
manufacturing equipment and materials base.
• SEMATECH'S focus on external R&D activities
has "exposed a division of interest among the
consortium's partidpants."
• SEMATECH'S Phase 2 and 3 objectives rely too
much on current-generation lithogr^hy.
Although these concerns may seem to reflect
a gloomy assessment of the consortium's future,
the report as a whole shows a healthy sense of
pragmatism. Dataquest has been concerned that
SEMATECH m i ^ t be held accountable for goals
that are simply unrealistic, given the consortium's
structure and resources. Overall, the Advisory
Council report views SEMATECH'S main benefits
to Americans as "indirect" in the sense that they
are "likely to come from the continued operation
of commercially vigorous U.S.-based manufacturing Snns ready and able to exploit emerging technologies." The value of such indirect contribution
to U.S. competitive strength, in the Advisory
Council's opinion, is sufficient justification for
continued federal support.
This newsletter reviews the conclusions of
SEMATECH 1990: A Report to Congress and
focuses on the Advisoiy Council's assessment of
SEMATECH'S progress in 1989 and the coticems it
raises about the program's future.
REORGANIZATION IN I989
Tie Advisory Council report attributes much
of SEMATECH'S progress in 1989 to its reorganization, which signaled a shift in balance between
the two models that have guided the consortium
since its founding. The report said that the first of
tihese models concerned "the development and
demonstration of world-class manufacturing
processes on-site, and the transfer of resulting technology directiy to members in large, integrated,
connectable chunks." Yet another operative model
for the consortium was fliat of a facifitator aiKi
testing ground for leading-edge semiconductor
equipment and materials.
Although SEMATECH theoretically adheres
to both of these models, the consortium's limited
resomces have made it impossible to develop an
ambitious in-house production strategy as well as
invest in the preservation of domestic sources of
first-class tools and materials. According to the
Advisoiy Council report, "The consortium cannot
&199I Dataquest Incoiponcal ranuaiy-RepKiductioa Pmbitnted
SEMMS Stmicoaductor Mamtfamixiasfladuj try Trends
0009170
77K content t^this report represents our interpretation and atmlysis of inprmation genemtly 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. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Datatptesl services. This infarmtaion is notjumished in connection with a sale or offer to sell securities or in cormection with the solicitation of an
offer to buy securities. This firm and its parent attd/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
^P^A
SEMATECH'S CONGRESSIONAL REVIEW: IS THE BEST IT CAN BE GOOD hi/
afford to address strategic interests of the industry
at large and install Mly integrated high-voluine
production lines at the same time."
Under the late Bob Noyce, SEMATECH
increased its emphasis on the second of its two
underlying models. The June 1989 reorganization
called for an increased volume of off-site R&D
projects to meet specific equipment, materials, and
manufacturing process reqiiirements for 0.50- and
0.35-micron production. The consortimn amended
its mission statement to read (SEMATECH's mission is) "to provide the U.S. semiconductor industry ttie dome^c capability for world leadership
in manufacturing." To SEMATECH, domestic
capability transktes into U.S. equipment and
materials suppliers. In preserving this capaibUity,
SEMATECH has crystallized its mission into the
following two highly strategic objectives:
• Protea U.S. semiconductor manufacturers from
dependence on foreign sources of supply
m Ensure that new and improved equipment
and materials are developed "in place with
chipmakers; purchasing cycles for the next
two generations of semiconductor device
technology"
Resource Reallocation
Given its objectives, SEMATECH underwent
a major reallocation of its resources in 19S9. In
rationalizing its form with its function, the consortium canceled plans for a second fab, scaled back
hiring projections, and doubled its original budget
for off-site R&D. Of SEMATECH's $260 million
1990 budget, 53 percent ($137 million) was set
aside for R&D contracts. In contrast, extemal R&D
accounted for only 20 percent of the consortium's
1988 budget. Whereas plant and equipment
accounted for $75 million of SEMATECH's 1989
expenditures, the current operating plan calls for
plant and equipment spending to fall to $19 million
in 1991—9 percent of the consortium's projected
budget.
Beyond the issue of resources, the report
expresses other doubts concerning SEMATECH's
success in the in-house d^nonstration of a highvolume production capability. The report states dxat
"to establish and operate a fuUy integrated fab
line...SEMATECH would have been obliged to produce some version of a saleable device and to rely
on its members to supply the necessary device and
process designs. Whether tnembers would have
provided such support is uncertain."
0009170
Impact on Programs
The Advisory Council credits SEMATECH's
reorganization witti progress in the following areas:
• Technology development
• Technology transfer
• Improving supplier relations
• Strengthening the U.S. technology base
Technology Development
At the heart of SEMATECH's project-based
operating system is its Master DeUverables List
(MDL). This list is the result of SEMATECH's
conq)etitive analysis and comparison of U.S, and
foreign manufacturing capabilities. Out of this
analysis has come the current targeting of the following major thrust areas: lithography, metrology,
multilevel metaUlzatioD, furnace and implantation
technology, and manufacturing methods and
processes. At the time of the Advisory Council
report, SEMATECH's MDL mcluded 56 projects in
various stages, more than one-half of which weare
being generatedfliroughjoint-development projects
(JDPs).
The Advisory Council report reveals that the
increased momentum of JDP activity observed during fee latter half of 1989 was due not only to
changes in the consortium's stmcture and budget,
but, more importantly^ to changes in the development contract process. Quoting from an SEMI/
SEMATECH annual report, the Advisory Council
notes that "intellectual property proved an insurmoimtable barrier to starting up the development
contract process." The Advisory Council report
cites progress in making the participation
agreements more flexible and observes that
"SEMATECH now negotiates the rights to jointly
developed technology (e.g., prefesential purchasing
and licensing rights) on a case-by-case basis, with
final arrangements largely dependent on how
project costs are shared and the market strcngtti
of tiie contractor."
With the removal of intellectual property barriers, contracting activity accelerated at
SEMATECH during the second half of 1989. As a
result, the Advisory Council report states that
"senior officials at SEMATECH and DARPA
report that the consortium's R&D program is on
track and on time." Table 1 Usts SEMATECH
contract activity, as observed by Dataquest,
organized on the basis of major technology thrust
areas.
®1991 Dataquest Incorporated Januaiy-Repioduction Prohibited
SEMMS Semicoiiductor Manufacturing/Industiy Trends
g e
TABLE 1
11
SEMATECH External Development Contracts
SEMATECH Thrust Areas
Lithography
Submicion Reticle and Mask Exposure System
Optical Wafer Stepper
Advanced Photoresist Processing
X-Ray
1-Line Steppers
Laser Mask Writer
Advanced Reticle and Mask Exposure System
Ion Implant
High-Energy Implantation Technology
PECVD
Dielectric CVD
Global Planarization Process
Dry-Etch Technology
Metal-Etch Systems
Plasma-Etch Technology
Electron Cyclotron Resonance (ECR)
Low-Temperature Etch
Sputter Cluster Tbol
Metrology
Wafer Defect Detection
Critical Dimension Measurement ^f0eBiS
Metrology Standards
Process Architecture/Integration
Test Chips
Advanced Isolation
Manufacturing Methods
Ultrapure Gas Management Systems
Establishment of SETEC
Manufacturing Specialist Training Proi^iun
Furnaces
Vertical Furnace
Contract Partner
Program 1 ^
ATEQ
GCA
SUicon VaUey Group
Hampshire
GCA
ATEQ
ATEQ
JDP
JDP
JDP
JDP
EIP
Other
JDP
Ion Implant Services
TAA
Applied Materials
Westech Systems
EIP
JDP
Lam Research
Oak Ridge Nat'l. tal».
Lam Research
Drytek
Eaton
EIP
TAA
JDP
EIP
JDP
KLA
Angstrom M64$W^pli^
NIST
JDP
EIP
JDP
HP
NCR
JDP
JDP
SemlGas Systems
Sandia Nat'l. Lab.
Texas State Tech. Inst.
JDP
TAA
TAA
SVG
EIP
JDP = JcM-DenJopniBit Program, EIP •• Equipment Inqnovement Program, TAA = Technical Assistance Agreement
N A = Not avulaUc
Souice: Dataquest (Fmiuiy 1991)
C
SEMATECH'S CONGRESSIONAL REVIEW: IS THE BEST IT CAN BE GOOD ENOUGH?
Technology Transfer
Prior to its reoi;ganization, SEMATECH had
stressed a horizontal transfer of technology to its
member companies. This strategy was largely
predicated on on-site technology development. The
Advisory Council report observes that SEMATECH
now relies more heavily on two-way vertical technology transfers "mediated by SEMATECH but
occurring with increased frequency in direct
exchanges between members and suppUCTS."
AUhough some of the tool development and
prototype testing that originally was planned for the
consortium's canceled tool applications process
facility (TAPF) wiU be performed in its main fab,
most of this work will be assigned to member
conqjanies. An exanople of such an arrangement
iavolves GCA. At an estimated cost of $24 million
to $32 million, SEMATECH is buying between
15 and 20 GCA i-line steppers to distribute to
five or more member companies. With technical
support from GCA, the member companies wiU use
the steppers in their production lines, compare
them witti foreign alternatives, improve upon them,
and share the resulting information with GCA.
W^th its eniphasis on vertical relationships,
much of SEMATECH'S horizontal ti-ansfer of
technology will take place through its member
company assignees and through SEMATECH
technology transfer teams that regularly visit
member companies.
Improving Supplier Relations
The Advisory Council concludes that, in a
broad sense, SEMATECH has improved communication among equipment suppliers and users
through its program of woikshops, symposia, and
joint sessions of the SEMATECH and SEMV
SEMATECH boards. SEMATECH also has created
a suppUer relations action council consisting of
senior purchasing and materials managers from its
membCT corapanies. The purpose of this groiip,
known in SEMATECH circles as "the partnering
posse," is to "promote strategic relatioiis with U.S.
supphers at their home companies." The report
credits these efiforts with the structuring of a joint
effort by U.S. semiconductor manufacturers and
supphers to acquire Perkin-Elmer's e-beam and
optical lithography divisions rather than risk its
acquisition by a non-U.S. con^any.
Strengthening the Technology Base
In addition to its R&D contract efforts, the
Advisory Council report notes that SEMATECH'S
$10 million investment in 11 universities (the
0009170
SEMATECH Centers of Excellence, or SCOEs) has
"generated some early unanticipated retums." The
report cites four cases involving improved
scientific understanding, six cases involving new
experimental capability, and seven cases involving
new fn-oduct concepts. As of December 1989,
SEMATECH had graduated more than
75 employees from its manufacturing specialist
program. In August 1989, SEMATECH established
a Semiconductor Equipment Technology Center
(SETEC) with Sandia National Laboratory. The
center, which will receive $10 million in
SEMATECH funds during the next three years, is
charged with the development of reliability technology for semiconductor manufacturing equipment. In another National Lab program,
SEMATECH and the Oak Ridge National Laboratory joined forces in December 1989 to develop
electron cyclotron resonance etch reactors for
0.5-micron wafer processing.
THE EROSION CONTINUES
Ironically, although the Advisory Coxmcrl
report on SEMATECH provides Congress with a
highly favorable evaluation of the consortiimi's
activities, it also points out that during this period
of positive accomplishments, "erosion in the market position of U.S.-owned semiconductor
manufacturing equipment and materials companies
seemed to accelerate." According to Sam HarreU,
president of SEMI/SEMATECH, 65 U.S.
equipment and materials companies were acquired
during the year of SEMATECH'S incorporation.
Out of these acquisitions, 37 were made by U.S.
companies, 12 were made by European companies,
and 16 were made by Js^janese companies.
In addition to the issue of consoUdation, the
report runs through a now-familiar htany of industry woes such as declining U.S. market share in
equ^Muent and materials, greater size and business
diversity of foreign competitors, lower hurdle rates
on prospective investments for foreign conipetitors,
and lower rates of capital spending by U.S. semiconductor cotnpanies coinpared with their Japanese
rivals.
In the face of such broad and pernicious
environmental factors, the report concludes that
"even at their most successful, SEMATECH and
similar measures are palliatives—selective and
temporary efforts to compensate for general conditions in the U.S. economy that have contributed
to coir^titive weakness in a range of domestic
industries." Moving from palliative to corc^titive
®1991 Dataquest Incoiporated Jamiary-Repioductioa Probitnted
SEMMS Semiconductor Manufacturing/Industiy Trends
i
SEMATECH'S CONGRESSIONAL REVIEW: IS THE BEST PT CAN BE GOOD ENOUGH?
antidote would, however, make SEMATECH a
much more expensive prescription for the
industry's ills. The Advisory Council states the
recommendation of the National Advisory
Committee on Semiconductors (NACS) that
SEMATECH be used "to channel increased R&D
support to the U.S. SME (semiconductor manufacturing equipment) and materials industry." However, the NACS rqport estimates that "a full-scale
effort to meet the needs of U.S. equqnnent and
materials firms would require an additional
$800 million over the next three years."
Areas of Concern
Beyond the question of financial resources,
the Advisory Council report raises interesting
queries about SEMATECH'S future based on some
implications of its 1989 restracture. First, the report
notes that "SEMATECH intends to sustain or
create one world-class U.S. producer in each major
category of chipmaking equipment, secondsourcing only in special cases where the back-up
company uses an entirely different tool architecture
or represents a particular high-risk/high-retum
investment opportunity." This objective seems to
confirm fears voiced by equipment makers during
SEMATECH'S creation that the consortium would
be a catalyst to further attrition in the U.S. semiconductor equipment base rather than a force for
heahhy diversity.
The report also says that SEMATECH's
project-based approach and external R&D activities
has "exposed a division of interest among the
consortium's participants." The nature of this
division as characterized by the rqport has to do
with the desire of smaller members for major infusions of leading-edge process technology.
SEMATECH's focus on the preservation of domestic sources offirst-classtools and materials is more
in keeping with the priorities of its larger members,
which already have advanced processing capability.
Under SEMATECH's 1987 Partnership Agreement,
December 1989 marked the first time that consortium members could give the required two-year
notice activating their option to leave the alliance.
As the report observes, "some of the consortium's
smaller &ms may have reassessed their ability to
support the considerable cost of membership"
(which is 1 percent of the previous year's semiconductor sales with a $1 million minimum and a
$15 million cap).
®1991 Dataquest locorporated Jamiaiy-Repioduction Piobibited
SEMMS Semicaodactor Maraifactuiing/Industiy Trends
The Advisory Council report also e^qiresses
concern over the iiiq>lications of SEMATECH's
einphasis on short-loop rather than fuU-flow wafer
processing. The report concludes that this approach
"will be insufficient for conclusive demonstrations
of equipm«at and wiU impose some limitation on
the development of important process technologies." In addition, the generic process architectures
on vAMx SEMATECH will base its Phase 2 and
3 objectives "could omit iinportant steps or tools
that member firms would need to make their own
0.50- or 0.35-micron products." TTie Advisory
Council, however, recognizes that the establishment
of an in-house, high-volume capability for
SEMATECH is more than just a matter of fiimcial
resources. "To establish and operate a fuUy
integrated fab line...SEMATECH would have been
obliged to produce some version of a saleable
device and to rely on its members to supply the
necessary device and process designs. Whether
members would have provided such support in
uncertain."
The report's concern over SEMATECH's
long-term effectiveness extends also to its R&D
contracts, which its notes "focus mainly on wafer
processing rather than inqxHtant antecedent steps
(e.g., product design, materials development) or
final chip assembly and packaging." The Advisory
Council also worries that SEMATECH's Phase 2
and 3 objectives rely too much on currentgeneration lithognq}hy rather than "technologies
that may be the basis of conq)etitive high-volume
production at the end of die 1990s."
RECONCILING PUBLIC AND PRIVATE
INTERESTS
In coming to grips widi the adequacy of
SEMATECH's Phase 2 and 3 goals, the Advisory
Council correctly concludes that increasing
SEMATECH's funding would not necessarily provide a solution. Hie problem is more in the very
nature of SEMATECH's public/private identity.
The report points out fliat SEMATECH's "tendency to sharten planning horizons ^ipears to be a
recurrent pattern in consortia exposed to market
pressures." Althou^ a key objective of publicly
supported cooperative R&D would seem to be the
extension of private investment horizons, the aim
of industry leadership is to keep such programs
responsive to market requirements—in other
words, keep R&D in phase with chipmakers'
purchasing cycles.
0009170
SEMATECH'S CONGRESSIONAL REVIEW: IS THE BEST IT CAN BE GOOD ENOUGH?
DATAQUEST CONCLUSIONS
The Advisory Council views the NACS
recommendation of increasing the SEMATECH
budget by $100 miUion in the near term as a way
to "reduce the risks inherent in the consortium's
R&D enterprise" through die development of a
full-flow demonstration environment and increased
on-site testing of unsolicited equipment and materials. Increasing federal funding by a larger amount
would allow the consortium to address high-cost,
long-term projects such as X-ray and excimer laser
lithography technology, advanced device concepts,
and new materials. What concerns the Advisory
Council about such a funding increase is the strain
it could place on the SEMATECH alliance. An
increase in federal funding in support of long-term
projects (and the consequent increase in membership fees to preserve the public^jrivate balance)
would "appeal mainly to SEMATECH'S largest
members (and) would conflict with the consortilun's evolving corporate culture, which is inclusive,
cooperative, and responsive to near-term madcet
conditions."
0009170
Based on this evaluation, the Advisory Council argues for the status quo in terms of federal
support to the consortium while concluding that the
SEMATECH program may not be sufficient on its
own to revitalize the global competitiveness of the
U.S. semiconductor industry. However,
SEMATECH clearly is a means to that end and has
played a major role in facilitating the £^Ucation of
technology from existing sources within industry,
academia, and government for the improvement of
existing manufacturing tools and the creation of
new ones. The Advisory Council concludes that to
hold SEMATECH accountable for the ultimate
salvation of the U.S. semiconductor industry is to
ignore the responsibilities of its member companies
to incorporate the technologies it fosters and "the
role of national poUcy in geaeraL"
Note: This newsletter was produced as a
cooperative effort between the Semiconductor
Industry Service (SIS) and the Semiconductor
Equipment, Manufacturing, and Materials Service
(SEMMS).
Jeff Seerley
Michael J. Boss
01991 Dataqaest Ihcoipoiated Januaiy-Reprodiictiaii Piofaibited
SEMMS Senncomiuctor Manufactuiing/bidastiy Trends
i
Worldwide Fabs
1^
ift
lis
Dataquest
acompanyof
The Dun& Bradstiect Corporation
Research Newsletter
THE FAB DATABASE SERIES: THE SHIFT TO SUBMICRON
GEOMETRIES
SUMMARY
This research newsletter analyzes production
wafer fab capacity trends by minimum drawn
linewidth. The Semiconductor Equipment,
Manufacturing, and Materials Service (SEMMS)
worldwide fab database is the foundation of this
analysis. Dataquest defines a production fab as one
that is capable of front-end processing more than
1,250 wafers per week. Minimum drawn linewidth
is defined as the minimum linewidth at the critical
mask layers as drawn.
The crossover point at which submicron
processing accounts for the highest percentage of
worldwide production will occur in 1993; this point
is illustrated graphically in Figure 1. DRAMs,
SRAMs, and MPUs wiU account for the majority
of submicron production when crossover occurs.
IMPLICATIONS FOR SEMICONDUCTOR
EQUIPMENT SUPPLIERS
Figure 2 illustrates 1991 worldwide theoretical submicron production capacity measured in
millions of square inches, ia. 1991, submicron
production represents 32 percent of the total worldwide semiconductor production edacity. In 1992,
submicron production wiU represent 37 percent of
the total worldwide semiconductor production
FIGURE 1
Worldwide Production Fab Capacity by Minimum Linewidth
(Millions of Square Inches)
Percentage
ES3 <1.0um
E l
>1.5um
>1.0 and <.1.5um
100
80
60
40.
20
1986
1989
1990
1991
1992
1993
1994
1995
Source: Dataquest (April 1991)
ei991 DaUquest Incoiponted April-Reprodnctioii Prohibited
SEMMS Newiletten 1991 Mamifactuiing—Woridwide Fabs
0010404
The content qflUs report represents our interpretmion 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 materialpmvided to us in corfidence by our clients. Individual companies reported on and analyml by Dataquest
may be clients of this and/or other Dataquest services. This it^formation is not furnished in connection with a sale or qfir to sell securities or in connection with the solicitation of an
offer to buy securities. This firm and its parent and/or their officers, stockholders, or members of their families may. from time to time, have a hr^ or short position in the securities
mentumed and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE FAB DATABASE SERIES: THE SHIFT TO SUBMICRON GEOMETRIES
FIGURE 2
Worldwide Submicron Production Capacity-1991
(Millions of Square Inches)
/
/
1
North
America
29%
/ Europe
\
,' 10% ^ - A
/
^
\
\
,/ROW
B%
\
\
Japan
53%
Source: Dataquest (April 1991)
capacity; inl993, submicron capacity will represent
40 percent of the total worldwide semiconductor
pFoducdoD cspaaiy. It is no surpdse that Japan
is the land of opportunity for submicron semiconductor equipment manufacturers, commanding
53 percent of the worldwide submicron cq>acity in
1991. Because S-inch wafers will play an increased
role in submicron processing, many equ^iment suppliers are offering or will soon offer 8-inch tools.
Submicron leading-edge DRAM fabs have a
substantial influaice on the tool development process. However, DRAM fabs are not the only ones
converting to submicron 8-inch tools. For exanq>le,
Motorola's MOS II fab in Austin, Texas, will start
microprocessor and fast static RAM production on
8-inch tools. J^anese, U.S., and European equipment suppliers are developing 8-inch tools with
0.5-micron process capability. Motorola claims feat
87 percent of aU MOS 11 equipment dollars were
spent with U.S. equipment suppliers rather
than foreign equipment supphers. If this is the case,
then U.S. equipment suppliers have already
completed a substantial amount of 8-inch tool
development.
IMPLICATIONS FOR SEMICONDUCTOR
MANUFACTURERS
The submicron era will faring a multitude of
changes to manufacturing. Semiconductor
manufactures have e:q>erienced a 30 percent
0010404
increase per generation in tool costs. To offset such
increases, manufacturers need to set competitive
goals such as achieving at least 75 percent tool
utilization, maintaining fab yields of 98 percent or
^realer, and mainlining at least 75 percent wafer
sort yields.
Automation will play a stronger role in
managing the manufacturing piocess. Single wafer
processing will become more common as wafer
sizes increase. Ouster tools will be utilized, but not
throughout the Mi. Equipment selection will consist of a mix-and-match strategy employing both
standalone and cluster tools. Semiconductor
manufacturers will feorougbly characterize these
tools and processes in an effort to eliminate
manufacturing variability. Concurrent engineering
will become a way of life to ensure that new
processes and products are transfened into production at competitive yields.
IMPLICATIONS FOR SEMICONDUCTOR
MATERIAL SUPPLIERS
Semiconductor material suppliers will sec an
increased demand for gases and chemicals that
meet the ultrapure specifications required for submicron processing. J^xan dominates the semiconductor materials madcet If the United States continues to become dependent on foreign sources for
semiconductor nuUerials, it may not be granted
access to the roo^ advanced materials as quickly as
will its foreign competitors, putting the United
States at a severe disadvantage.
ei991 DitaqoMt hwatpoatod AfriMUrrodnctiga PrahiUtod
SEMMS Newdelten 1991 Mwrnftrtniiin—Woridwide Fate
V
THE FAB DATABASE SERIES: THE SHIFT TO SUBMICRON GEOMETRIES
THE RISK ADVERSE STRATEGY
To avoid or at least reduce the risks involved
with new overseas subtnicron fab investments,
companies are adopting innovative financing strategies such as advance paym^its from customers,
joint ventures, or foreign government incentives.
Texas Instruments (TI) is one of die most innovative companies when it comes to diversifying
the risk associated with a new submicron fab.
For instance, in exchange for technology, TI
received $125 million toward the initial investment
of $250 million from the Italian government for its
fab in Avezzano, Italy. Also, one year ago TI
entered into a joint venture widi Kobe Steel to
manufacture VLSI logic and ASICs in Japan. Hie
company has also entered into a joint venture
with ACER to manufacture 4Mb DRAMs in
Taiwan. TI's strategy of diversifying both financially and geographically is an example of true
global thinking.
ei991 DiUqiiMt IncwpotlBd i^dMtopndncticMi Praliibitod
SEIAIS NawdMim 1991 Mrnifwimriiig—Waddiride Fite
DATAQUEST CONCLUSIONS
Submicron processing will pose a multitude
of challenges for the semiconductor industry
throughout this decade. E>uring this time, manufacturers will go from 6-inch wafers to 8- or 12-inch
wafers. Factory utilization and yield itiqnrovement
win become die driving forces behind efforts to
remain competitive.
In 10 years, leading-edge DRAM production
will go from 0.50-micron 16Mb DRAMs wifli an
excess of 400 process steps to 0.15-micron 1Gb
DRAMs with about 700 or more process steps.
Equipment and materials suppliers will play a key
role in semiconductor fab design and operation
throughout tiie 1990s.
J^ Seerley
0010404
I
i
(O
wmS:
.*is
DataQuest
aamiranyof
The Dun & Bradsticet Corporation
Research Newsletter
U.S. MERCHANT SEMICONDUCTOR CAPITAL SPENDING:
THREE LONG-TERM TRENDS
Dataquest has just completed an analysis of
U.S. merchant semiconductor manufacturers' C£^ital spending plans. The analysis reveals evidence of
long-term structural change in the industry, and this
newsletter highlights those structural changes. (For
results and detailed analysis of individual U.S.
merchant manufacturers' spending plans, please see
the accompanying SEMMS newsletter entitled
"1990-1991 U.S. Merchant Semiconductor
Manufacturers' Capital Spending: Moving Beyond
Tactical.")
We have identified three types of structural
change that have taken place:
• A tendency for semiconductor ci^ital spending
as a percentage of semiconductor revenue to
increase
• A tendency for the top three U.S. merchant
semiconductor conq>anies to account for a larger
portion of total U.S. merchant semiconductor
company csqpital spending (concentration of buying power)
• A tendency for total U.S. merchant semiconductor company capital spending to account for an
ever smaller portion of total worldwide merchant
semiconductor company capital spending
LONG-TERM TRENDS
Capital Spending as a Percentage
of Revenue
Figure 1 charts the change in the ratio of
ci^ital spending to revenue since 1983. Cs^ital
FIGURE 1
U.S. Merchant Capital Spending as a Percentage of Revenue
Source: Dataquest (April 1991)
<D1991 Dataquest Incoipoiated April-Repfoduction Prohibited
SEMMS Newfletten 1991 Mamifactuiing—Capital Spending
0010361
The content of this report represents our interpretation and analysis q^ infi)ntmtion 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 pruvided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this andlar other Dataquest services. This ir^brmation is not fitmished in connection with a sale or offir to sell securities or in connection with the solicitation of an
i^er to buy securities This firm and its parent and/or their officers, stockholders, or members of theirfiimiliesmay, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Parle Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
U.S. MERCHANT SEMICONDUCTOR CAPITAL SPENDING:THREE LONG-TERM TRENDS
spending as a percentage of revenue was at an
all-time higji of 23 percent in the boom of 1984.
With the industry downtum of 1985 and 1986,
however, the ratio of semiconductor capital spending to semiconductor revenue feU to 11 percent.
After 1986, it rose steadily to 16 percent in 1989;
in 1990, the ratio declined ever so slightly to just
over 15 percent.
The question of whether the ratio of capital
spending to revenue will resume its climb is difficult to answer. There are very strong forces pushing it up: The continuous increase of device and
manufacturing complexity will increase the cost of
both equipment and facilities. Offsetting these
forces will be the efforts of manufacturers to cut
capital costs through automation, use of microenvironments, simplifying both process flow and
recipes, and increasing equipment reUability and
throughput.
Concentration of Buying Power in the
Top Three
Motorola, Texas Instruments, and Intel are the
top three in U.S. merchant semiconductor device
market share, with 44 percent of the total U.S.
merchant semiconductor company revenue. Not
surprisingly, these conq)anies are also the top three
in capital spending for the U.S. merchants. In 1990,
these top three con^anies accounted for 56 perc^it
of all U.S. merchant ci^ital spending.
This share of 1990 U.S. merchant conpany
capital spending is much larger than the top three's
share of U.S. merchant cotapaay capital spending
just a few years back. In 1984, flie top three U.S.
merchants (Texas Instruments, Motorola, and
National Semiconductor) accounted for 45 percent
of aU U.S. merchant semiconductor Ciqrital spending.
A Smaller Piece of the Capital
Spending Pie
U.S. merchant saniconductor conq)anies as a
whole represent a declining share of total worldwide cs^ital spending. In 1984, U.S. merchant
semiconductor capital spending accoimted for
40 percent of total worldwide merchant cq)ital
spending. By 1990, the U.S. merchant share of total
capital spending had declined to 29 percent This
0010361
decline is due to two factors. First, the U.S.
merchant semiconductor companies have lost market share in the worldwide semiconductor
marketplace—^from 48 percent of the total worldwide merchant market in 1984 to 36 percent in
1990. Second, U.S. merchants' share of total worldwide capital spending has also declined because
U.S. merchant semiconductor companies have not
spent as high a proportion of their revenue on
ci^ital spending as have the Japanese and Asia/
Pacific companies. For example, in 1990, U.S.
merchants spent 15 percent of their revenue on
coital s{>en(ling, Japanese companies spent 20 percent, and Korean companies spent 63 percent.
DATAQUEST CONCLUSIONS
The semiconductor industry is already so capital intensive that many new companies have chosen to go fabless rather than undergo the expense
of constructing and aperadng their own manufacturing facilities. Even companies with fabs, such as
Intel, have chosen to use foundries for their lowermargined product lines. If the ratio of capital
spending to revenue continues to increase, DsUaquest expects to see more fabless con^anies, especially start-ups, and an increased use of foundries
by even the largest semiconductor companies for
their low-margined products.
WhethOT the ratio of capital spending to revenue will rise again or the efforts of semiconductor
manufacturers to hold down die growth of capital
intensity wiU bear fruit, one thing is clear—
semiconductor manufacturing will remain cs^ital
intensive. As such, the cost of capital, which is
higlher in the United States than in other regions,
will remain a competitive disadvantage for U.S.
merchants.
Much attention has been paid recently to the
loss of worldwide market share by U.S. merchant
semiconductor conq>anies. One effect is that U.S.
merchants as a whole buy a smaller piece of the
worldwide capital spending "pie." Those equipment vendors that focus on U.S. merchants as their
prunaiy target customers, therefore, are targeting a
piece of the pie that, relatively speaking, has
become smaller, even as the size of the pie has
grown larger.
Because the top three U.S. merchants now
rejHesent over 50 percent of the U.S. merchants'
coital spending, those vendors that target only
U.S. merchants and are unable to sell successfiilly
01991 Dataquest Incorporated April-Reptoduction Prohibited
SEMMS Newdetten 1991 Manufacturing—Capital Spending
U.S. MERCHAm- SEMICONDUCTOR CAPITAL SPENDING:THREE LONG-TERM TRENDS
FIGURE 2
U.S. Merchant Semiconductor Company Share of Worldwide Merchant Capital Spending
1984
1990
Top 3 1
U.S.
Merchants |
17%
\,,^
Other
\
U.S.
\
1 Merchants
\
23%
\
Non-U. S.
Com^ianles
60%
Worldwide Merchant
Capital Spending
Worldwide Merchant
Capital Spending
Source: Dataquest (April 1991)
to the top three are relegated to operating within a
very thin slice (13 percent) of the whole worldwide
capital spraiding pie (see Figure 2). That slice does
not represent a very strong growth strategy.
Dataquest believes that a more robust strategy (but
one difficult for start-ups to follow) is to target
worldwide semiconductor conq)anies, regardless of
the country where their headquarters are located.
George Burns
#
The topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, ^ u c h include wafer fab equipment
suppliers, semioonductor inini»riai« coiiq>aiiies, and semiconductor device mannfactuters. The topics selected indicate the Ixoad range of research
diat is conducted in the SEMMS group. Clients, however, often have specific infonnatian requirements that eittter go beyond the level of detail
contained in fbe newsletters or beyond the scope of what is normally published in the newsletters. In order to provide complete decision siq>port
to our clients, Dataquest has a consulting service available to handle these additional information needs. Please call Stan Bruederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
ei991 Dataquait Inccnporated Apiil-Reproduetiai Prohibited
SEMMS Newdetter> 1991 Manufactoring—Ci^ntal Spending
0010361
DataQuest
aomipanyof
The Dun & Bradstrect Corporation
Research Newsletter
U.S. MERCHANT SEMICONDUCTOR CAPITAL SPENDING
HIGHLIGHTS: 1990-1991
Dataquest has just completed its survey of
U.S. merchant semiconductor manufacturers' c o i tal e3q)enditures and plans for 1990 and 1991. This
newsletter highlights the major features of those
plans and e^)enditures.
The U.S. merchant semiconductor companies
responding to this survey represent 78 percent of
U.S. semiconductor merchant revenue. Tlieir plans
indicate that U.S. merchant, company-funded,
semiconductor cs^ital spending will increase 3 percent in 1991 over 1990 (see Table 1). However,
when Texas Instruments' (TI's) joint-venture
expenditures (TI/Acer, TI/Kobe, and TJ/Singj^Koe)
are added, total (con^any funded and TI's joint
ventures) U.S. merchant semiconductor capital
spending will increase by 15 percent in 1991 over
1990.
U.S. MERCHANT SEMICONDUCTOR
CAPITAL SPENDING: 1990-1991
Advanced Micro Devices (AMD) is planning
to spend approximately $150 million in 1991,
down from $275 million in 1990. The company's
Subnoicron Development Center (SDC) fab was
completed in Santa Clara, California, in 1990. The
SDC is both a production fab and development fab
and is one of die most advanced fabs in the world.
TABLE 1
U.S. Merchant Semiconductor Capital Spending
(Millions of Dollars)
Company
AMD
Analog Devices
AT&T
Cypress
General Instrument
Harris
IDT
Intel
LSI Logic
Micron Technology
Motorola
National Semiconductor
TI
Vitelic
1989
159
44
150
40
20
68
57
380
115
230
540
200
'641
36
1990
275
38
100
36
NA
74
48
550
71
107
550
146
707
35
1991
150
53
170
60
NA
70
34
900
100
100
640
140
460
20
Percent Change
1990-1991
-45
41
70
67
-5
-29
64
41
-7
16
-4
-35
-43
(Cortinwd)
0 1 9 9 1 Dataqueit Incoiponted April-Reproduction Prohibited
SEMMS Newsletters 1991 Manufacturing—Capital Spending
0010338
The content of this report represents our interpretation and analysis of information generally available to the public or released by responsible irtdividuals in the subject corrqjanies, but
is not guaranteed as to accuracy or ccmipleteness. It does not contain material provided to us in cor^idence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services. This information is notjumished in connection with a sale or offer to sell securities or in amnection with the solicitation cfan
o^r to buy securities. This firm and its parent and/or their cheers, stodckolders, or members of their families may, firm time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
U.S. MERCHANT SEMICONDUCTOR CAPITAL SPENDING HIGHLIGHTS: 1990-1991
1 (Continued)
U.S. Merchant Semiconductor Capital Spending
(Millions of Dollars)
TABLE
Company
VLSI Technology
Western Digital
Others
Company-Funded Total
Percent Change
TI Joint Ventures
Percoit Change
Total
Percent Change
Percent Change
1990-1991
0
-62
-1
416
1990
55
106
330
1991
55
40
326
3,145
3,228
3,318
15
3
3
150
NA
550
267
267
3.145
3.378
3,868
15
15
7
15
1989
50
•
3
NA = Not avaUaUe
Souice: Dataquest (Apnl 1991)
AMD's 1991 spending will be down because of the
completion of this major project. The conq)any's
efiforts in 1991 will be directed toward ramping up
capacity for its 80386 microprocessor in Austin,
Texas. It is possible that AMD will build a new fab
in two years.
Analog Devices recently acquired Precison
Monolithics and is in the process of consolidating
its manufacturing operations. As part of its consolidation process, the company recently closed its
Greensboro, North Carolina, fab.
AT&T's capital budget has been low in recent
years as the company struggled with reorganization. Its Orlando, Florida, fab originally was
underutilized, but in the last two years it has been
the site of significant foundry activity for AT&T. It
has been rumored that the Orlando facility has been
short of ciq)acity and late on some semiconductor
deliveries recendy. We e:q)ect AT&T to continue to
Tamp production at its fab in Spain and add new
capacity at its Allentown, Peimsylvania, and
Orlando fabs.
Cj^iress Semiconductor recently acquired
Control Data Cotporation's fab in Bloomington,
Minnesota, for $13.7 million. This fab is one of fbs
few fabs in die world that is fiilly equi^>ped with
standard mechanical interface (SMIF) technology.
In addition to its Minnesota fab and its fab in
San Jose, California, Cypress also has a fab in
Roundrock, Texas, that it shares with Altera Semiconductor. Under this agreement. Altera paid
Cypress $7.4 million and manufacturing rights for
Altera's Multiple Array matrix (MAX); in retum,
Altera gains access to a certain portion of the fab's
capacity.
0010338
Harris is reorganizing its semiconductor
operations. It has announced tiiat it is delaying
plans to construct a new fab and assembly and te^
facility in PlymouA, England.
IDT is ramping up its new Technology
Development Center fab in San Jose. Tim fab is a
26,000-square-foot Class 1 facility that is both a
development and production fab. It currently has a
cf^acity of 5,000 6-inch wafers per mondi.
Intel has the most aggressive capital spending
plan ($900 million) of any company in the semiconductor industry. Its increase alone of $350 million in 1991 capital spending exceeds the 1991
capital budgets of all U.S. merchant companies
except Motorola and Texas Instruments.
Intel is upgrading its D2 development line in
Santa Qara from 6- to 8-iQch wafers. The coiiq)any
is building a new 8-inch fab MPU development
line in Aloha, Oregon, and is also building its first
European fab (also an 8-inch fab) in Leixlip,
Ireland. The coo^pany plans to begin production at
fab 9.2 in Albuquerque, New Mexico, by mid-1991
and is continuing to add equipment to fab 9.1 in
Albuquerque and to its fab in Israel.
LSI Logic has transferred high-volume ASIC
production £com its facility near London to its fabs
in Tsukuba, Js^ian, and San Jose. LSI plans to
manufacture low-volume analog/digital devices at
its fab in the United Kingdom near London. LSI is
currently spending $150 million to build an additional fab in Tsukuba. This fab is planned to be
ready for equ^iment by this October.
In the first quarter of 1991, Micron Technology completed the conversion of Fab I ^ fit>m
5- to 6-indi wafers. Micron has no major plans
61991 Dataqueit Inccnpoiated Apiil-Repioduction Prohibited
SEMMS Newfletters 1991 Manufacturing—Capital Spending
U.S. MERCHANT SEMICONDUCTOR CAPITAL SPENDING HIGHLIGHTS: 1990-1991
for expansion this year. It continues to add
capacity through shrinking its die size. For
example, by shrinking the die for its 1Mb
DRAM, Micron claims to have increased its
1Mb capacity by 100 percent.
Motorola completed MOS 11 at its new facility in Oak Hill, Texas, in 1990 and will run first
silicon in May. This 8-inch fab is the industry's
first for non-DRAM products. Motorola is also
adding capacity for 1Mb DRAMs at its East
Kilbride, Scotland, facility and building a new
bipolar developmental and production line in
Chandler, Arizona.
National Semiconductor recently sold
Matsushita its Puyallup, Washington, BiCMOS fab
that it acquired when it bought Fairchild Semiconductor. National is transferring processes that were
formerly run at its Puyallup fab to its Santa Clara
fab. The conq)any plans no significant edacity
additions in 1991.
TI's company-funded capital spending plans
are down significantly in 1991 because the company recently completed adding significant new
capacity to its Avezzano, Italy, facility and its
joint-venture facility (TI/Acer) in Hsin-chu,
Taiwan. However, die company continues to
vigorously upgrade its existing faciUties and add
capacity through noncompany-fimded joint ventures such as KTI, the joint venture between Kobe
Steel and TI in Japan. TI recently announced that it
will build a $330 million joint-venture DRAM fab
in Singapore with the Singapore Development
Agency (SDA), Canon, and Hewlett-Packard (HP).
TI and SDA wdU each own 26 percent of the new
company, and Canon and HP will each own
24 percent. Constraction of the new fab will begin
by this summer, and production is expected by
mid-1993.
Vitelic broke ground for a new fab in
Hsin-chu in late 1989, but stopped construction in
late 1990 because of weak demand for its products.
It is currently discussing a merger with Hualon
Semiconductor, a Taiwanese company. Our
estimate for "^teUc's 1991 spending level is based
on the assuiiq)tion that these discussions will result
in a merger. >^telic also bought a fab in Hong
Kong from El Cap in 1989 to use as a pilot line
and as a means to penetrate the People's Republic
of China.
VLSI expanded and upgraded its San Jose fab
in 1990 and plans to continue to facilitize its new
San Antonio, Texas, fab in 1991.
Western Digital produced first silicon at its
new submicron fab in Irvine, California, in March
1991, Manufacturing at this facility will be primarily for the production of engineering prototypes and
production volumes aimed at facUitizing early market penetration. Tbtal cost of the facility and equipment was just under $120 million.
DATAQUEST CONCLUSIONS
In a year of a general and nagging economic
recession and uncertainty, the expansion of U.S.
merchant capital spending and cq)acity is significant. AT&T, Cypress, Intel, LSI Logic, and Motorola are increasiag their spending plans and adding
significant ci^acity now. Of the eight U.S. merchant semiconductor manufacturers that reported a
decline in capital spending, five have recentiy completed significant state-of-the-art edacity additions.
One of these companies, TI, is still adding significant new capacity through its noncompany-fiinded
joint ventures.
With all the new capacity currentiy being
added and recentiy having beeai added by U.S.
merchant soniconductor manufacturers, the claim
can no longer be made that U.S. merchant capital
spending is cut back during a downturn or a period
of business uncertainty. Widi the inclusion of TI's
noncompany-fiinded joint ventures, total U.S. merchant semiconductor cs^ital spending wiU increase
15 percent in 1991.
By adding cq)acity now, U.S. merchant semiconductor manufacturers are moving beyond the
tactical and ate looking forward to ibe next expansion. U.S. merchant csq)ital spending is strategic in
nature, at least for this business cycle.
George Burns
Tlie topics covered by SEMMS newsletters are selected for their general interest to SEMMS clients, viiiich include wafer fab eqoqnnent
suppliers, semiconductor materials conq>anies, and semiconductor device nuumfiBctuiers. The topics selected indicate the broad range of research
that is conducted in tbe SEMMS group. Clients, however, often have specific information requirements ttiat either go beyond the level of detail
contained in the newsletters or beyond the scope of what is normally published in the newsletters. In order to provide cooqtlete decisicHi support
to our clients, Dataquest has a consulting service available to handle these additional infonnation needs. Hease call Stan Bruederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom leqniiements.
01991 Ditaqueit Incoipormted April-^eproductian Piohitrited
SEMMS Newtletten 1991 Mmiifaetuiing—Capital Spending
0010938
Dataoy^est
suxHTipanyfrf
The Dun &lBrad5tieet CcMporation
Research Newsletter
FIVE-YEAR CAPITAL SPENDING FORECAST: ITS A GOOD WORLD
IN SPITE OF 1991'S UNCERTAINTY
SUMMARY
The year 1991 will be a year of uncertainty.
Dataquest's forecast for worldwide spending on
semiconductor property, plant, and equipment
(PPE) calls for 11 percent growth (see Table 1),
assuming that there wiU not be a war in the Persian
Gulf. If there is a war, we predict that it will be
short and that its economic effects will be limited.
We are also assuming, independently of the crisis
in the Persian Gulf, that the major economies of the
developed world will not slide or tumble into a
recession. If any of these events occur, investment
plans currently under way or about to be
implemented wiU almost certainly be pushed back.
Although the passage through the first year of
our five-year forecast may contain hazards such as
war and recession, we are optimistic about the
five-year period as a whole. Demand for devices
will continue to grow as clever engineers from
around the world find new applications for ever
more complex and sophisticated devices. Historically, semiconductor capital spending has doubled
every five years. We beUeve that this doubling will
continue. Capital spending will grow at a compound annual growth rate (CAGR) of 15 percent
ovCT the period from 1990 to 1995. We expect
worldwide capital spending for PPE to reach
$25 billion by 1995.
TABLE 1
Regional Capital Spending Forecast
(Millions of Dollars)
CAGR (%)
1990-1995
1989
1990
1991
1992
1993
1994
1995
5,348
16
5381
1
6,045
8,651
9,169
19
6
10,178
11
14
12
7,270
20
3,846
3,949
4,265
5,223
6,267
6,831
7,515
14
12
3
8
22
20
9
10
1,961
1,736
1,875
2,581
4,214
4,425
16
(11)
8
38
3,665
42
15
5
1,141
1,313
1,575
2,466
2,811
3,233
22
15
20
1,989
26
24
14
15
12,296
12379
1
13,760
11
17,062
24
21,049
23,026
23
9
25,352
10
Capital Spending
Japan
Percent Change
United States
Percent Qiange
Asia/Pacific
Percent Change
Europe
Percent Change
Worldwide
Percent Change
22
21
20
15
Source: Dataquest (January 1991)
0 1 9 9 1 Dataquest Incoipoiated Januaiy-Reproduction Prohibited
0009219
SEMMS Newsletters 1991 Q ^ t a l Spending
I companies, but
yzcrf by Dataquest
.M,^ f„ t« „ ,.„ ;• -n.- c
J "
J, , • '^
'
r o^ to sell securities or in connection with the soUcitatiim of an
c^rlo buy securities This firm and ,1s parent and/or their cfficers. stockholders, or members cf their families may, from lime ta lime, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
FIVE-YEAR CAPITAL SPENDING FORECAST: IT'S A GOOD WORLD IN SPITE OF 1991's UNCERTAINTY
The regions of Europe and Asia/Pacific wiU
continue to be the fastest-growing markets for C£5)ital spending, although Ji^an and the United States
will continue to be the two largest markets for
capital expenditure.
Although we are optimistic about the upcomingfive-yearperiod, we note that there is currendy
an excess of DRAM capacity and if recently
announced DRAM c{q)acity additions take place as
plaimed, then an excess capacity situation wiU
extend throughout our forecast period. This excess
is a negative factor in the forecast and could both
change individual companies' plans and slow down
the growth rate of capital spendiag.
will take place outside of Japan, in the United
States and Europe. As evidence of this, we note
that in 1985 there was only one Japanese fab line
operating outside of Japan. However, in 1990, we
count 13 such fab lines, and we expect that there
will be 27 fab lines operating outside of J^an
by 1992.
This trend of J^anese companies building
fabs outside of the Japanese region wUl slow down
the long-term growth rate of spending in Ji^an. We
expect capital spending in Japan to almost double
by 1995 to $10.1 billion. The CAGR for the 1990
through 1995 period is expected to be 14 percent,
which is less than the growth rate expected for the
European and Asia/Pacific regions.
REGIONAL MARKETS
The discussion that follows is about regional
markets. Spending in these regions is defined to
include spending by both domestic and foreignbased companies taking place within diat region.
Spending by domestic companies that takes place
offshore is included within the region where it
takes place. For example, spending by a U.S. company in Japan would be counted as spending within
the Japanese regional market and not as spending
within the U.S. regional market.
Japan
Japan is the largest market for capital spending in the world. Total spending in Jiq)an for PPE
in 1990 was $5.4 billion, up 1 percent fiom 1989's
level of $5.3 bilUon. Measured in dollars, these are
record levels of spending. However, it should be
pointed out that spending in the peak years of 1984
and 1985, measured in yen (¥924 bilUon and
¥794 billion, respectively), was higher than in
1989 and 1990 (¥754 billion and ¥767 bilHon,
respectively).
Growth in spending in Jiq)an will be fueled
mainly by J^anese company spending, in both the
near term and long term. This spending will be to
maintain market share, as well as technological
parity and technological leadership in device technology and manufacturing. In addition to defending
their leadership position in commodity DRAMs,
Japanese conq)anies will continue to invest in
ASIC and noncommodity DRAM markets such as
application-specific DRAMs.
In evaluating the strength of the Js^anese
regional market, it is ixoportsaat to note that a
growing percentage of Ji^anese coiiq>any spraiding
0009219
United States
The United States is the second largest
regional market for coital spending in the world.
Total spending in the United States for PPE in 1989
was $3.8 biUion. la 1990, the market in the United
States increased 3 percent to $3.9 billion. This
spending increase was led by capital spending
increases by major U.S. merchants and by the
increasing presence of Japanese manufacturers in
the United States. We estimate that spending by
Japanese companies in the United States increased
firom more than $4(X) million in 1989 to more than
$600 miUion in 1990.
Overall spending in the United States as a
region was influenced negatively by the fact that a
substantial portion of spending by con^anies such
as IBM, LSI Logic, Motorola, National Semiconductor, and Texas Instruments was done outside of
the United States. Cutbacks in spending by many
of the second- and third-tier U.S. conqjanies also
brought down the 1990 growth rate for die U.S.
region.
We e:q)ect capital spending in the United
States to almost double by 1995 to $13 billion.
The CAGR for the 1990 to 1995 period is expected
to be 14 percent. One of the major reasons for this
low growth rate is that U.S. conq>anies will continue to spend a substantial portion of their cs^tal
e:q>enditure offshore.
Asia/Pacific
Asia/Pacific includes the Pacific Rim countries, excluding Japan. The size of capital spending
in Asia/Pacific exceeds that of cq)ital spending in
Europe. Capital spending in the Asia/Pacific region
01991 Dataquett Incoipoiated Januaiy-Reproductiai Prohibited
SEMMS Newiletten 1991 Capitil Spending
FIVE-YEAR CAPITAL SPENDING FORECAST: IT'S A GOOD WORLD IN SPITE OF 1991's UNCERTAINTY
declined 11 percent from $1.9 billion in 1989 to
$1.7 billion in 1990. This decline was due to
cutbacks in spending by South Korean con^anies.
Part of this decline was made up by an increase in
spending in Taiwan. However, because South
Korea is the largest market for cs^ital investment
in the Asia/Pacific region, the overall trend in 1990
wiU be down.
Dataquest expects fairly robust growth in
spending in South Korea in 1991, spurred by additions of DRAM cj^acity and also by the recent
cocunitment of South Korean companies to
become major players in the ASIC markets.
Spending in Taiwan could be flat ia 1991.
This could be due to a rapid decline in the
Taiwanese stock market (about NT$11,000 in the
first quarter of 1990 to about NT$3,000 in the third
quarter of 1990). Partly as a result of this crash,
and also because of a general uncertainty about
business conditions, Taiwanese semiconductor
manufacturers are now delaying orders by a quarter
or two.
We expect capital spending in the Asia/Pacific
region to grow over thefive-yearforecast period at
a CAGR of 21 percent and to reach $4.4 \Mm
by 1995.
Europe
Spending in Europe by all companies
increased by qjproximately 15 percent in 1990.
This spending increase in Europe was fueled
primarily by offshore companies from the United
States and Japan building new fabs in Europe. We
expect that capital spending will increase 20 percent in Europe in 1991, again primarily because of
ibs increased activity of off^ore manufactureis.
J^anese companies will approximately double
their spending in Europe in 1991 to about
$g00 noilUon. Fujitsu, Hitachi, NEC. Mitsubishi,
01991 Dataqueit lacotpantod Jamiaij^Reproductian Piobilrited
SEMMS Newdetten 1991 Ca|rital Spendiiig
and peiiifq)s Toshiba will have fabs or be building
idibs in Europe by the end of 1991.
We expect capital spending to reach
$32 billion by 19S5 and to grow at a CAGR of
20 percent from 1990 to 1995. This growth will be
fueled by the general economic growth that will be
the result of the integration of (he EC and the
desire of major non-European manufacturers to
have a fab line close to their customers and also
adjacent to Eastern Europe.
DATAQUEST CONCLUSIONS
In a year in which the possibility of war and
recession are the major topics of conversation,
foxecasters and businesses necessarily concentrate
on the short run. The year 1991 could be one of
moderate growth in cs^ital spending, or it could be
a year in which, because of war and recession,
there is no growth. In the shoat run, semiconductor
companies are platming to go ahead with their
expamion plans as if there will be no major
adverse effects from the crisis in the Persian Gulf
or from a possible recession. Howevex, ordea^ can
be canceled and inq^lementation of plans can be
delayed. Our forecast for 1991 is for capital spending to increase by a modest U percent. This forecast is fringed with uncertainty.
to die long temi, however, we do not see a
future clouded with any more uncotainty than is
usual. We expect semiconductor ci^ital spending to
continue to increase at healthy rales for the ^veyear period to 1995. This increase will be fiieled by
semiconductor companies buying equipment and
building fab lines for half-nucron fabs; by new
entrants into the industry, such as the Taiwanese
and South Koreans; and by the need for all large
manufacturers to have a manufacturing presence in
every region of the world.
George Burns
0009219
Semiconductor R&D
' • • _ *
^
- - •
"•
"
II
I
- i ^ - r ^ r n ^T
^.^'
P • • .Tn*^jj- = ' I I F J
W^. U "-rm^^i >-i '>^-
DataQUest
iCHitpatlVof
The Dunft Bracktirct Corpocabon
iff
Research Bulletin
S E M I G A S : T H E BALANCE BETWEEN TRADE POLICY AND
OPEN MARKETS
The Antitrust Division of the Department of
Justice announced on December 28, 1990, that it
will sue to prevent the sale of SemiGas Systems to
Nippon Sanso. SemiGas is a manufacturer of highpurity gas distribution and control systems sold to
semiconductor manufacturers. Nippon Sanso is a
Japan-based supplier of industrial and specialty
gases and one of the five largest industrial gas
companies in the world, with 1990 estimated revenue of ¥177,000 million. The Justice Department
plans to file a civil antitrust suit in U.S. District
Court within the next few weeks.
The Justice Department pointed out that its
suit is an antitrust action based strictly on market
share concentration. It emphasized that the suit is
not politically motivated although its investigation
into the sale was initiated by a con[q>laint from
Sematech. Sematech is an R&D consortium of U.S.
semiconductor companies that receives $100 million annually in funding from the Federal government. The sale would work in direct opposition to
Sematech's stated purpose of preserving a U.S.
suppUer base.
ANALYSIS
According to Dataquest's 1989 market share
estimates (see Table 1), the combined SemiGas
and Matheson/Nippon Sanso sales would equal
44.4 percent of the U.S. market for gas cabinets.
Nippon Sanso's concentration in the industry
following the sale would exceed the threshold in
the Justice Department's Merger Guidelines,
namely 1,800 points on the Herfindahl-Hirschman
Index and a change in that index of 50 points.
However, if the the Justice Department's
Merger Guidelines were i^Iied to the other bidders, all would be excluded except for the management lead buyout. Purchasers with even the
smallest existing North American gas cabinet sales
would exceed the threshold because SemiGas has
such a dominant market share of a very small
market. Dataquest expects Hercules Corp., the
U.S .-based parent of SemiGas, to fight the Justice
Department's smt, arguing that the Justice Department's Merger Guidelines should not ^jply for
these very reasons.
Indeed, Hercules has a very strong incentive
to fight the suit, because a successful challenge by
the government is widely expected to result in
a much lower seUing price. Nippon Sanso's bid,
estimated to be $23 ntillion, was well above the
second place bid, estimated to be $18 million. In
addition, if Hercules must go out for bids again,
Dataquest estimates that the current market will not
support the price multiples originally bid in April
1990 because of overall market conditions. In
TABLE 1
North American Gas Cabinet Market—1989
(Millions of Dollars)
North American Production
SemiGas Systems
16.9
Air products
8.8
ASGT/Air Liquide
5.2
sa
4.0
HoPure/Union Caibide
2.0
Scott Specialty Gases
1.2
Airco/BOC
1.0
Matheson/Mippon Sanso
Total North America
0.8
39.9
Souice: Dataquest (Januaxy 1991)
®1991 Dataquest Xncoiporated January—Reproduction Prohibited
SEMMS Newsletters 1991 Materials/Industry Trends
0009209
77K cmtent cfthis report represents <mr interpretation and analysis ofir^rmaJion generally available to the public or released by responsible individuab in the subject comparues. but
is not guaranteed as to accuracy or completeness. It does not contain material provided to us in confidence by our clients. Individual companies report^ on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is notfiimished in conneaion with a sale or offer to sell securities or in connection with the solicitation of an
offer to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may. from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
SEMIGAS: THE BALANCE BETWEEN TRADE POLICY AND OPEN MARKETS
today's market, Dataquest estimates that SemiGas
would sell for $12 million to $15 miUion.
The only potential buyer is the SemiGas
management team. The management team faces the
problem of putting together a leveraged buyout
at a time when banks are avoiding such deals.
Dataquest believes that a likely resolution is for an
industrial gas company, i.e., Airco, Nq>pon Sanso,
or another semiconductor equipment manufacturer,
to take a minority stake in a management lead
buyout. It is also possible that Hercules wiU decide
to pull SemiGas off the market.
DATAQUEST CONCLUSIONS
Although the SemiGas sale is relatively small
in terms of doUars, Dataquest believes that there
are significant ramifications to the Justice Department's suit. First, it is widely recognized diat the
Committee on Foreign Investment in the U.S.
0009209
(CFIUS), which is under the Treasury Department
and is vested with protecting technology critical to
national interest, has lost much of its credibility.
Consequently, Dataquest e:q)ects high-technology
trade policy to be pursued more actively through
the Justice Departmoit's antitrust arm and new
legislation currently beiag proposed in Congress.
Second, the government's decision to challenge the sale of high-technology companies will
have the downside effect of lowering the competition for these assets. In the case of SemiGas,
Dataquest expects this effect to translate into a
lower seUing price for the company. Hercules is
expected to receive $8 million to $11 million less
in today's market for SemiGas because of the
Justice Department's action. Government
policymakers must carefully balance their public
policy objectives and the impact of these policies
on the market.
Mark FitzGerald
01991 Dataquest Incoip<HBted Januaiy^Reproduction Prohibited
SEMMS Newdetten 1991 Materiali/Industiy Trends
w
o
o
3
d
Dataquest
a company of
The Dun & Biadsticct Corporation
'^m
Research Newsletter
FLAT FZ WAFER MARKET
The worldwide demand for float zone (FZ)
silicon wafers is estimated to grow at less than
a 5 percent annual compound growth rate
(CAGR) during the next five years. High-power
discrete devices, which currently account for the
bulk of the FZ wafer demand, are expected to to
grow more slowly than the total power discrete
market. In addition, the development of improvedquality thick films is allowing many of the higherpower transistors to move away from FZ to epitaxial films, eroding the growth opportunities for
FZ wafers in this segment. Low- to medium-power
discretes, which make up the bulk of the discrete
device unit sales, generally use epitaxial films,
which are more cost-effective.
MARKET TRENDS
Power discrete devices fall into three general
categories—^thyristors, power diodes, and power
transistors. Dataquest estimates that these product
segments had worldwide revenue of $617 million,
$1,340 million, and $2,360 mUlion, respectively, in
1990 (see Table 1). Power diodes make up the
largest unit volume with an estimated 12.6 billion
sold in 1990; power transistors accounted for
4.9 billion units and thyristors for 1.2 billion units
(see Table 2). About 80 percent of the power
transistors currently produced are bipolar, although
the MOS market is starting to develop.
TABLE 1
Estimated Worldwide Consumption of Power Discrete Devices
(Millions of Dollars)
Device
CAGR (%)
1990-1995
1990
1991
1992
1993
1994
1995
617
671
725
774
811
821
Power Diodes
1,340
1.508
1,700
1,900
2,022
2,097
5.9
9.4
Power Transistors
2,360
2,700
3,140
3,625
4,000
4,185
12.1
4,317
4,879
5.565
6,299
6,833
7.103
10.5
Thyristors
Total
Source: Dataquest (March 1991)
TABLE 2
1990 Worldwide Unit Consumption of Power Discrete Devices
(Billions of Units)
Device
Thyristors
Power Diodes
Power Transistors
Total
Total
Low to
Medium
High
Percentage
High (%)
1.20
1.04
0.16
13.3
12.6
11.26
4.44
1.34
0.46
10.6
9.4
16.74
1.96
10.5
4.9
18.7
Source: Dataquest (March 1991)
<&>1991 Dataquest lacorporated Marcb-Repioduction Prohibited
SEMMS Newsletters 1991 Materials—Silicon
0009581
TV 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. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation of an
q^r to buy securities This firm and its parent and/or their officers, stockholders, or members of their families may, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
FLAT FZ WAFER MARKET
Although total power discrete revenue is
expected to have a 10.5 percent CAGR during the
next five years, the revenue trend masks the true
growth potential for FZ wafers. FZ wafers are sold
into the high-power discrete applications. The highpower segment is a small subset of the power discrete market, accounting for only 1.96 billion units
or 10.5 percent of the total power discrete unit
consumption in 1990 (see Table 2).
Moreover, the growth opportunity in each of
the three device categories is in the low- to
medium-power device segment. Low- to mediumpower devices are typically used in a broad range
of consumer electronic equipment such as hair
dryers and televisions. High-power devices are
typically used in more durable products such as
automobiles, motor controls, and elevators. Dataquest estimates that the high-power device revenue
will grow at only a 5 to 7 percent CAGR during
the next five years versus the low- to mediumpower segment, which is estimated to grow at
a 10 to 12 percent CAGR. The difference in
growth rates reflects the end-use appUcations into
which these devices are being sold.
PRODUCTION TRENDS
High-power thyristors and diodes are fabricated almost exclusively from high-resistivity
FZ wafers using a double-sided diffusion process to
provide the semiconductor junction. With these
devices, the voltage is applied between the front
and back sides of the wafer, and the current flows
through the wafer using metal electrodes placed on
both sides. Relatively thin doped FZ wafers are
used, and the overall processing is very straightforward compared with planar devices. Czochralski
(CZ) wafers cannot be used, mainly because of the
problems in obtaining a high-quality, highresistivity wafer.
The fabrication process for high-power (planar) transistors on an FZ wafer is fairly complex
and involves an initial long (deep) two-sided diffusion step. One side is then ground back to the
high-resistivity material and polished prior to carrying out the further diffusion steps needed to form
the transistor. In order to leave the fabricated wafer
with a manageable thickness, especially for the
larger diameters, the starting wafer has to be relatively thick. Long diSiision times are needed to get
a low-resistivity diffusion layer with the required
thickness.
0009581
The altemative fabrication method is more
straightforward. An epitaxial film is deposited onto
a doped CZ wafer and the transistor fabricated in
the film. A lengthy diffusion step is eliminated and
relatively thin starting wafers, which are readily
available in large CZ wafer diameters, can be used.
The difficulty with this process is getting a thick,
good-quality epitaxial film onto which reliable
higher-powered transistors can be fabricated. To
date, most device manufacturers have stayed with
the FZ despite the complex processing involved.
Dataquest expects this situation to change
during the next five years as more reliable thick
films are developed to be used for emerging bipolar
and MOSFET appUcations. There is a definite
move toward the use of epitaxial films for the
emerging higher-performance transistor, and it is
generally believed that most of the new power
devices will use epitaxial films rather than
FZ wafers.
DATAQUEST CONCLUSIONS
Because of product mix and a relative difference in the use of epitaxial wafers, Dataquest
believes that the number of FZ wafers consumed in
Europe is approximately twice that of North
American consumption (see Table 3). Again,
because of the size of the discrete market and a
product mix that tends to favor FZ over epitaxial
wafers, the volume of FZ wafers consumed in
Japan is estimated to be about four times that of
North America. The demand for FZ silicon in the
Asia/Pacific region is estimated to be small relative
to the other regions of the world.
Historically, volume growth of siUcon wafers
has lagged the revenue growth of the integrated
circuit market by about 5 to 6 percent. Although
the same amount of difference in growth rates is
not anticipated for discrete devices, it seems
TABLE 3
Estimated 1990 FZ Wafer Consumption
(Millions of Square Inches)
Country
North America
Square Inches
11-13
Japan
43^5
Europe
23-25
Asia/Pacific
5-6
Total
82-89
Source: Dataquest (Maidi 1991)
®1991 Dataquest Incoiporated Maicb-Reproduction Prohibited
SEMMS Newsletters 1991 Materials—Silicon
FLAT FZ WAFER MARKET
reasonable to assume that silicon volume gjrowth
rate for power discretes wiU fall short of the corresponding discrete revenue growth rate.
Dataquest forecasts power discrete devices to
have a 10.5 percent CAGR during the next five
years. The high-power segment is estimated to
grow even more slowly, at a 5 to 7 percent
01991 Dataquest Ihcoiporated Maicb-RepioductiQn Prohibited
SEMMS Newsletters 1991 Materials—Silicon
CAGR. This trend, coupled with the fact that many
of the new power transistors wiU probably use
epitaxial films, points to a future FZ market CAGR
of less than 5 percent for the 1990-to-1995 time
frame.
Mark FitzGerald
00095S1
Dataquest
acompanyof
TheDun&Hradsti:[eet Corporation
Research Newsletter
T H E DECLINE OF THE
100mm WAFER MARKET
The glory days of 100mm (4-iiich) silicon
wafers are in the past. The rise of 150mm wafers in
the latter half of the 1980s and the current push to
200mm wafers have eclipsed the dominance of
100mm technology (see Figure 1). Dataquest is
forecasting the worldwide consumption of 100mm
wafer slices to shrink during the next five years at
a compound annual growth rate (CAGR) of
6.3 percent (see Table 1). Although the 100mm
wafer market is in decline, Dataquest e j e c t s
worldwide sales to total a healthy 27.80 million
slices or roughly $300 million to $400 million in
revenue in 1995.
REGIONAL MARKETS
U.S. fabs have the largest l(X)mm wafer-start
c£^>acity, estimated at 18.01 miUion sQices. Moreover, U.S. production capacity is highly leveraged
on 100mm wafers; it accounts for 37.6 percent of
the total wafer-start capacity (all wafer diameters)
in 1991 (see Table 2). The dependence on 100mm
wafer technology can be understood in terms of
plant investment cycles. U.S. manufacturers
invested heavily in new 100mm plants from 1976
to 1984, the peak of 100mm process technology.
FiGintE 1
Worldwide Silicon Wafer Consumption
Percentage of Wafer Starts
Percentage (AN Diameters)
8070
60
¥V.\ 200mm
i ^ ^ 150mm
\%:%\ 100mm
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Note: Data do not add to 100 percent because 2-inch, 3-inch, and 125mm data are exciuded.
Source: Dataquest (February 1991)
01991 Dataquest Incorporated February-Reproduction Pndiibited
SEMMS Newsletters 1991 Materials—Silicon
0009425
The content of this report represents our interpretation and analysis cfin^rmation 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. Individual companies reported on and analysed by Dataquest
may be clients of this and/or other Datacptesl services. This information is not famished in connection with a sale or offer to sell securities or in connection with the solicitation of an
tyffer to buy securities This firm and its parent and/or their officers, stockholders, or members of their fomilies may, jktm time to time, have a long or shttrt position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
THE DECLINE OF THE 100MM WAFER MARKET
TABLE 1
Estimated lOOmm Silicon Wafer Consumption
(Millions of Slices)
United States
Japan
Europe
Asia/Pacific
Total
1990*
1991
1992
1993
1994
14.67
15.01
5.80
3.07
38.55
14.58
14.07
5.84
2.81
37.30
15.84
13.27
6.03
3.07
38.21
15.38
12.54
6.30
3.17
37.38
14.01
10.10
5.73
3.29
33.13
1995
13.84
6.42
4.97
2.57
27.80
C A G R (%)
1990-1995
-1.2
-15.6
-3.1
-3.5
-6.3
*Actiul
Souice: Dataquest (Febniaiy 1991)
TABLE 2
1991 Fab Capacity
(Millions of Slices)
United States
Japan
Europe
Asia/Pacific
Total
Total
Wafers*
47.89
65.37
27.61
18.79
159.66
100mm
Wafers
Percent
100mm
18.01
15.86
11.78
3.46
49.10
37.6
24.3
42.7
18.4
30.8
*A11 wafer diameteis
Souice: Dataquest (Febniaiy 1991)
As ISOnun technology evolved following the 1984
to 1985 recession, U.S. investment in new largerdiameter plants did not keep pace with the other
major regions of the world.
Japanese fabs also have a large lOOnun waferstart capacity, second only to the United States,
with an estimated 15.86 million slices. However,
unlike the United States, Je^anese 100mm wafer
capacity accounts for only 24.3 percent of total
Jq)anese wafer-start capacity. Japanese companies
invested heavily in lOOmm wafer plants from 1976
to 1984 but also continued to invest in new 150mm
plants after the 1984 to 1985 recession. The capacity of these larger-diameter plants quickly surpassed the installed 100mm wafer capacity. In
addition, Japanese companies have historically
been more aggressive in upgrading older fab lines.
European fabs have a smaller lOOmm waferstart capacity than do U.S. or J^anese fabs. Total
100mm capacity is estimated to be 11.78 million
slices, in part reflecting a much smaller total semiconductor capacity. Although European 100mm
capacity is relatively small, this region, like the
United States, is highly leveraged on 100mm wafer
technology, with 42.7 percent of its fab c£5>acity
dedicated to 100mm wafers. Europe, like the
0009425
United States, invested heavily in 100mm plants in
the 1976 to 1984 time frame. In the latter half of
the 1980s, European investment in larger-diameter
plants lagged behind the other regions of the world,
leaving European capacity exposed to 100mm
wafers.
The Asia/Pacific region is the new kid on the
block; as such, its semiconductor industry benefited
by growing up in the second half of the 1980s,
when ISOram technology was widely adopted.
Consequently, lOOmm fab capacity in this region
totals a mere 3.46 miUion wafers and accounts for
only 18.4 percent of wafer-start capacity, the lowest
of any region of the world.
DATAQUEST CONCLUSIONS
Many merchant silicon-wafer manufacturers
are moving their production out of 100mm into the
larger-diameter wafers. The 150mm and 200mm
products have better margins, and the largerdiameter wafers are the fastest-growing segments
of the silicon wafer market. Yet the demand for
lOOmm wafers, although declining, is going to
persist weU into the laiter half of this decade.
In the short term, this trend is creating a
shortfall in the supply of 100mm wafers. Dataquest
anticipates that the supply/demand imbalance wiU
grow more serious through 1993, which we forecast to be the top year in the next silicon cycle.
Spot-noarket purchasers and those semiconductor
manufacturers that do not have an established relationship with a wafer vendor are at greatest risk.
In the long term, Dataquest expects prices to
edge iq) 5 to 10 percent per year during the next
five years. Dataquest expects the average selling
price of 100mm wafers, now $10, to rise to the
$13 to $15 range by 1995. The higher prices will
be a large e n o u ^ incentive for some of die smaller
wafer vendors to maintain and even add lOOmm
wafer-production cjqjacity.
Mark FitzGerald
<S>1991 Dataquest Incoiporated February-Repnxhictian Prohibited
SEMMS Newsletters 1991 Materials—Silicoa
. * — ,
W W
| M ^ —
1
lll^t^AV
I' I I
#W
2
• + * L'l^Wi.-wi'VjwiiifciLryrtKi*
^^mm
Dataquest
accHniranyof
The Dun & Bradstieet Corporation
• ^ ^ f l ^ i ^ ^ ' l ^ j ' * . - . 4JV-^
•ti^-fe-pl:
•BS
Research Newsletter
HEALTHY FORECAST FOR U.S. POSITIVE RESIST
INTRODUCTION
In 1990, the number of gallons of photoresist
used in front-end semiconductor processing grew
7.7 percent in the United States. The United States
was the second largest regional market for photoresist after Japan with sales totaling $97.5 million
or 366,700 gallons.
The growth in the U.S. market is attributable
to positive-type resists. In 1990, consunq)tion of
positive resist totaled 266,600 gallons (see
Table 1), an increase of 11.9 percent. The consunq)tion of negative resist totaled 100,100 gallons
(see Table 2), a decrease of 2.2 percent. Dataquest
expects positive-resist volumes to continue to grow
during the next five years; negative-resist consumption is e}q>ected to decline.
TRENDS
The average selling price (ASP) for negative
resist is estimated to be $93 per gallon. Dataquest
expects negative-resist prices to remain flat because
they are used in less-critical applications (e.g.,
discrete devices). Negative-resist prices are largely
TABLE 1
1990 U.S. Optical Photoresist Market—Positive Resist
Company
Gallons (K)
North America
J.T. Baker
Dynachem
K n Chemical
OUn Hunt
Shipley
Subtotal
Japan
Mitsubishi Kasei
Sumitomo
Tokyo Ohka
Subtotal
Europe
Ciba Geigy
Hoechst
Subtotal
Total
Share (%)
Sales* ($M)
* 2.2
30.0
19.5
54.0
91.0
196.7
0.8
11.3
7.3
20.3
34.1
73.8
0.7
9.9
6.4
17.8
30.0
64.9
0.2
0,5
20.7
21.4
0.1
0.2
7.8
8.0
0.1
0.2
6.8
7.1
1.5
47.0
48.5
266.6
0.6
17.6
18.2
100.0
0.5
15.5
16.0
88.0
•ASP = $330
Soiuce: Dataquest (Jamuny 1991)
01991 Dataquest Incoipotated Januaiy-Reproduction Prohibited
0009246
SEMMS Newsletteis 1991 Photoresist
77K conleni of this report represents our interpretation and analysis of information generally ayailable to the public or released by responsible individuals in the subject companies, but
is not guaranteed as to accuracy or completeness. It does not contain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is not furnished in connection with a sale or offer to sell securities or in connection with the solicitation cfan
cffer to buy securities This firm and its parent and/or their cfficers, stockholders, or members of theirfomiliesmay, from time to time, have a long or short position in the securities
mentioned and may sell or buy such securities
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
HEALTHY FORECAST FOR U.S. POSITIVE RESIST
TABLE 2
1990 U.S. Optical Photoresist Mariiet—Negative Resist
Company
North Amedca
Dynachem
KTI Chemical
Olin Hunt
Subtotal
Japan
Tokyo Ohka
Europe
Ciba Geigy
Total
Gallons (K)
Share (%)
8.0
25.0
57.0
90.0
8.0
25.0
56.9
89.9
0.1
0.1
10.0
100.1
10.0
100.0
Sales* ($M)
0.8
2.4
5.4
8.6
1.0
9.5
•ASP = $95
Source: Dataqiieit (Jamiaiy 1991)
set on a cost-plus basis. The estimated ASP for
positive resist is $330 per gallon. This figure is
e^)ected to increase because consumption of the
new higher-priced submicron optical positive resist
is forecast to grow faster than the overall positive
resist market.
The prices for submicron optical positive
resist run in the range of $450 to $500 per gallon.
Prices are significantly above the positive-resist
ASP because oftiixeincreased costs associated with
meeting tougher specifications. Resist siq)pliers are
faced with tighter metal and particulate specifications. In addition, semiconductor manufacturers are
requiring more uniform performance characteristics
of a resist from batch to batch. These trends are
driving up the ciq>ital investment required to
manufacture submicron resist.
At the same time, the volume of resist used
per mask step is dropping off significantly because
of improvements made in track equipment.
Leading-edge track equipment uses 1.5 to 2.0 milliliters of resist per mask level on a 150mm wafer;
earlier generations of track equipment typically
delivered 4.0 milliliters of resist per layer. New
track equipment installation will have the greatest
impact on submicron optical resist volumes
because the lion's share of new track equipment is
purchased for submicron semiconductor processes.
A counterbalance to the intact of reduced
dispense volumes is the trend toward more mask
layers driven by increasing process con^lexity and
increasing levels of metallization. Devices such as
DRAMs and ASICs will have many more masking
steps than do previous generations. For exanq>le,
0009246
Dataquest forecasts that the 64Mb DRAM will
have more than 25 mask levels (see Figure 1). The
increase in mask steps will partially offset the
impact of smaller volumes of resist used per layer.
DATAQUEST CONCLUSIONS
Taking into account these trends and using
Dataquest's 1991 silicon wafer forecast for the
United States, we estimate that the total dranand for
photoresist in sendconductcnr front-end processes
will reach 534,800 gallons in 1995 (see Tkble 3).
The compound annual growth rate (CAGR)
for photoresist wiD be 10.7 percent from 1990 to
1995. Positive resist wlU total 442,800 gallon
and negative-resist demand will shrink to
92,000 gallons.
North American companies dominate the
positive-resist market, accounting for 73.8 percent
of the gallons sold. Shipley has tqiproximately onethird of U.S. positive-iesist market share. Ranked
just behind it are two other U.S. companies, Ohn
Hunt md Dynadxem, and one European company,
Hoechst, having shares that fall in the
lO-to-20-percent range. Also noteworthy is Tbkyo
Ohka, a Jf^ianese entrant to the U.S. market, which
has won a 7.8 percent share of the positive market
since 1987.
North American conq)anies also dominate the
negative-resist market, accounting for 89.9 percent
of negative resist Olin Hunt and KTI Chemical
own most of the negative market with 56.9 and
25.0 percent shares, respectively.
01991 DaUqueit Incorponted Jamuiy-Reproductioa PiohiUted
SEMMS Newtletten 1991 FhotoniUt
HEALTHY FORECAST FOR U.S. POSITIVE RESIST
FIGURE 1
DRAM Mask Levels
Number of Masks
3025-
10-
s
1Mb
4Mb
64Mb
16Mb
DRAM Density
Source: Dataquest (January 1991)
TABLE 3
1991 U.S. Photoresist Market Forecast
(Thousands of Gallons)
Volume
1989*
1990*
Positive
238.2
NA
Negative
Total
1991
1992
266.6
11.9
272.0
2.0
102.3
NA
100.1
(2.2)
340.5
NA
366.7
7.7
CAGR (%)
1990-1995
1993
1994
1995
309.2
13.7
366.6
18.6
420.6
14.7
442.8
5.3
98.6
(1.5)
98.2
(0.4)
102.0
3.9
95.9
(6.0)
92.0
(4.1)
(1.7)
370.6
1.0
407.4
10.0
468.6
15.0
516.5
10.2
534.8
3.5
7.8
10.7
*AcUial
NA s Not available
Souioe: Dataqueit (Jamury 1991)
However, Dataquest believes that Japanese
suppliers will increase their market share over the
next five years by targeting die i-line resist market
Jf^an Synthetic Rubber, Sumitomo, and Tokyo
Ohka have been successful in Japan with their
i-line products. Dataquest expects Japanese
conqjames to use these resists to gain share in the
i-line segment, ^iiich will be die fastest-growing
area of the U.S. photoresist market during the next
ten years.
Mark FitzGerald
The topics coveted by SEMMS newsletters ate selected for their general interest to SEMMS clients, vAucb include wafer fab eququaait
suppliers, semiconductor wint>"i^<if conqtanies, and semiconductor device manufacturen. t h e topics selected indicate die btoad range of research
that is conducted in the SEMMS gnnq>. Clients, however, often have specific infotmalion lequuements ttiat eidier go beyond the level of detail
contained in the newsletters or beyond flie scope of vrfiat is narmally published in die newsletters. In ortler to provide conq>letB decision support
to our clients, Dataquest has a consulting service available to handle these additional informatioo needs. Please call Stan Bmederle at (408)
437-8272 or Joe Grenier at (408) 437-8206 to discuss your custom requirements.
01991 Dataqnect Incoipoiated Janoaiy^Reproductiai PnriuUted
SEMMS Newdetteii 1991 Fhotoieiift
0009246
DataQuest
.4-
fss
fft^
acompanyof
The Dun 8t Brad^iect CcMporatKxi
Research Newsletter
CALIFORNIA FABS WILL SURVIVE THE DROUGHT
The fifth year of drought in California poses
more of a long-term threat to the state's semiconductor industry than an immediate problem for the
66 production fab lines (see Table 1), which
currently account for 17.6 percent of the North
American wafer capacity. Although the semiconductor manufacturing process is water intensive,
Dataquest believes that local water authorities have
worked closely with most companies in the past
two years to cut their consumption. Most companies wiU easily achieve current targets. However,
the drought is expected to further tarnish California's appeal as a site for future production fabs or
expansion of existing fabs.
The water shortage has not affected all areas
in CaUfomia to the same degree. The preponderance of the state's semiconductor fabs are located
in Silicon Valley and the Los Angeles basin, two
areas that are suffering the most serious shortfall
because they import the lion's share of their water.
On the other hand, the Sacramento area, where
NEC recently built a fab and is planning to expand,
is not as severely a£fected because local sources are
more plentiful.
WATER USAGE
Dataquest estimates that a production-scale
semiconductor plant uses from one to three acre
feet of water a day (an acre foot equals 326,000
gallons). Numerous factors determine an individual
plant's usage. Device production levels, the age and
design of the water system, and past spending on
capital-intensive reclaim programs are a few of the
more prominent factors that determine usage.
Although deionized (DI) water is a critical
material in the manufacture of semiconductors, the
DI water plants located at a £ab account for onfy
25 to 30 percent of the water consumed in that fab.
Secondary services such as cooling towers and
loops, wet scrubbers, and ancillary services use far
more water than does the actual manufacturing
process.
TABLE 1
Production Fabs in California by Company
Company
Location
AMCC
San Diego
AMD
Santa Clara
Sunnyvale
Avantek
Santa Clara
Newark
Cypress
San Jose
ECI Semiconductor
Santa Clara
Exar
Sunnyvale
Exel
San Jose
FEI Microwave
Sunnyvale
Foxboro ICr
San Jose
GI Quality Tech. Corp.
Palo Alto
Gigabit Logic
Newbury Park
Harris Microwave
Milpitas
Holt
Irvine
Hewlett-Packard
Santa Rosa
Palo Alto
IC Sensors
Milpitas
EDT
Santa Clara
Santa Clara
Salinas
IMP
San Jose
Intl. Rectifier
Rancho Califomia
El Segundo
Linear Technology
Milpitas
Milpitas
01991 Dataquest Incorporated February-Reproduction Prohibited
SEMMS Newsletters 1991 Materials—Other Materials
(Contittued)
0009S26
The content of this report represents our interpretation ami analysis of information generally available to the public or released by responsible individuals in the subject compimies, but
is not guaranteed as to accuracy or ctmipleleness It does not contain material provided to us in confidence by our clients. Individual companies reported on and analyzed by Dataquest
may be clients of this and/or other Dataquest services This information is notjumished in connection with a sale or q^r to sell securities or in connection with the solicitation of an
G0er to buy securities. This firm and its parent and/or their officers, stoc^diolders, or members of their families may, from time to time, have a long or short position in the securities
mention&i and may sell or buy such securities.
Dataquest Incorporated, 1290 Ridder Park Drive, San Jose, CA 95131-2398 / (408) 437-8000 / Telex 171973 / Fax (408) 437-0292
CALIFORNIA FABS WILL SURVIVE THE DROUGHT
1 (Continued)
Production Fabs in California by Company
TABLE
Company
Litton Microwave
LSI Logic
M/A-Comm Phi, Inc.
Micrel
Micro Power Systems
Microwave Tech.
National Semiconductor
NEC
Northern Telecom
Nova Sensors
Opto Diode
Oibit Semiconductor
Raytheon
Rockwell
Rohm
Samsung Semiconductor
Semicoa
Semtech
SI-Fab
Siemens
Signetics
Sihcon General
Sihcon Systems (TDK)
SiUconix
Spectra Diode Labs
Spectro Labs (Hughes)
TRW
Unisys Components Group
Vitesse Elect.
VLSI Technology
Xicor
Source: Dataquest (Febnuiy 1991)
0009526
Location
San Jose
Milpitas
Torrance
Sunnyvale
Santa Clara
Fremont
Santa Clara
Santa Clara
Santa Clara
Santa Clara
Santa Clara
Santa Clara
Roseville
Rancho Bernardo
Fremont
Newbuiy Park
Sunnj^ale
Mountain View
Mountain View
Newbury Park
Suimyvale
San Jose
Costa Mesa
Newbury Park
Scotts Valley
Cupertino
Sunnyvale
Sunnyvale
Garden Grove
Santa Cruz
Tustin
Santa Clara
San Jose
Sylmar
Redondo Beach
Rancho Bernardo
Camarillo
San Jose
Milpitas
Milpitas
The secondary uses of water, which account
for 70 to 75 percent of water usage, provide
manufacturers an opportunity to reclaim and recycle large voliunes of water. Unlike the DI water
used in manufacturing processes, the purity of the
water used in the secondary services is not critical
and can be reclaimed cost-effectively. Because such
a large volume of water is used in these secondary
services, manufacturers can dramatically lower
their water consumption by implementing reclaim
programs.
Little opportunity exists to use reclaimed
water as a feedstock for the DI water systems. The
expense of operating these systems is closely tied
to the quality of the water fed to the system.
Savings associated with using reclaimed water can
quickly be offset by the higher resin usage required
to polish a lower-quality feedstock. For this reason,
most fabs will continue to use water delivered from
the local municipality to supply their DI water
plants.
Many semiconductor plants in California have
achieved water conservation over the last several
years through careful planning. IBM in south San
Jose recycles water four to five times prior to
discharging it. Cypress Semiconductor has reduced
its water consumption by a factor of two since
1987 by upgrading its water system.
LONG TERM
Although Dataquest believes that semiconductor companies are positioned to meet the
planned cuts in water consumption for 1991, a
statistically imlikely sixth year of drought may well
leave some con^anies scrambling to implement
reclaim programs. The gravity of the overall water
situation can easily be understood by reviewing the
water balance sheet—supply and demand—^for
Santa Clara County. Santa Clara County encompasses most of Silicon Valley, so the drought's
impact on this area is representative of other areas
in the state with concentrations of high-tech companies.
The Santa Clara Valley Water District
manages the water supply for the county. There are
16 retailers supplying water within the county; 13
of these are city municipalities and 3 are private
companies. The Water District's job is to balance
the county's supply and demand. The use of
imported water must be balanced with the use of
local water supplies. In its position, the Water
District sets pricing and sourcing policy for the
retailers.
®1991 Dataquest Incorporated February-Reproduction Prohibited
SEMMS Newsletters 1991 Materials—Other Materials
CALIFORNIA FABS WILL SURVIVE THE DROUGHT
The Water District has three sources for
imported water—other counties with a surplus of
water; the state of California, whose water
resources are managed by the Department of Water
Resources; and the federal government, whose
water resources are managed by the Bureau of
Reclamation. Both the state and federal governments are planning to severely curtail their sale of
water to Santa Qara County. The county's supply,
both local and imported (see Table 2), should just
meet demand, assuming a 30 to 35 p^cent conservation level. The county's forecast water usage in
1991 totals 255,000 to 275,000 acre feet. The forecast usage assimies a 30 to 35 pocent savings over
1987 actual consumption. Water consumption in
the SiUcon Valley peaked in 1987 and totaled
393,000 acre feet per year; consequently, 1987 is
used as the base year for the county's planning
purposes.
The downside to the worst-case scenario is
that two of the largest sources for the county will
be depleted. If 104,000 acre feet of ground water
are used in 1991, the county will be perilously
close to a ground water level at which point ground
subsidence begins. In addition, the federal government's sources will be completely depleted except
for water levels required to sustain wildlife
habitats. If the drought were to continue into 1992,
the loss of these two critical water sources would
cause a severe shortage, assuming that no other
sources are found.
The Santa Clara Valley Water District is pursuing a conservative forecast strategy. However,
two wild cards may put the county in a better
position than the oirrent analysis might suggest.
First, there is considerable upside to conservation efforts that have not been realized. The shortage could be solved simply by setting strict priorities on water use, recognizing that one-half of
water usage in the county goes to landscaping.
Second, the county is dumping large volumes of
water from water treatment plants into the San
Francisco Bay. The water being dumped almost
meets drinking-water standards. In the future, this
water could be used for most industrial and agriculture applications, provided changes were made in
TABLE 2
Santa Clara County Water Sources and Use
(Thousands of Acre Feet per Year)
Normal
Entitlement
1991 Estimated
Entitlement
1991
Worst-Case
152
25-50%
38
State Government
97
15-50%
14
Yuba County
26
100%
26
Hetch Hetchy Reservoir
76
50%
38
351
NA
116
Natural Inflow
NA
NA
40
Coyote Canyon
NA
NA
5
Gound Water
NA
NA
104
Local Subtotal
Total
NA
NA
NA
NA
149
265
Sources Imported
Federal Government
Imported Subtotal
Sources—Lxxal
Base Year 1987
Actual Consumption
Use
Santa Clara County
393
1991 Targeted
Percentage Cuts (%)
30-35
1991 Targeted
Consumption
255-275
NA = Not available
Souice: Dttaqueat (Febniaiy 1991)
01991 Dataquest Incoiporated February-Reproduction Prohibited
SEMMS Newsletters 1991 Materials—Other Materials
(»09526
CALIFORNIA FABS WILL SURVIVE THE DROUGHT
state laws that currently prevent local water
districts from recycling this water.
DATAQUEST CONCLUSIONS
Because water has been so inexpensive to
date, Cahfomians have paid little attention to the
consumption of this resource. The current drought
is expected to force the state to seriously pursue
conservation strategies for residential, agricultural,
and industrial end users. It is also expected to result
in a permanent increase in water prices.
Dataquest believes that it is unlikely that the
operations of the state's semiconductor plants will
0009526
be adversely affected by the drought. Semiconductor companies are at the forefront of conservation
efforts. Moreover, significant savings can still be
realized by pursuing reclaim programs.
The drought is e:q>ected to add to the growing
list of reasons that argue against locating future
semiconductor production facihties in California.
However, Dataquest beUeves that California will
continue its leadership position as the corporate
headquarters and research and development center
for the North American semiconductor industry.
Mark FitzGerald
01991 Dataquest locorponled Febniary-ReproductioD Prohibited
SEMMS Newsletter! 1991 Materials—Other Materials
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement