CONTRIBUTION TO ARTICLE 305 STUDY ON SUPERCOMPUTERS
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP90T00114R000404380001-9
Release Decision:
RIPPUB
Original Classification:
C
Document Page Count:
21
Document Creation Date:
December 22, 2016
Document Release Date:
February 27, 2012
Sequence Number:
1
Case Number:
Publication Date:
February 25, 1987
Content Type:
MEMO
File:
Attachment | Size |
---|---|
![]() | 727.63 KB |
Body:
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9 5X1
DATE ~6-I O
DOC NO G1- M 97- t')O3(
OIR
PGPD J
Untral Intelligence Agency
2 5 FEB 1987
MEMORANDUM FOR: Glen Fukushima.
Director for Japanese Affairs
United States. Trade Representative..
Chief, Technology- and Industrial Competitiveness
Division
SUBJECT: Contribution to Article 305 Study on
Supercomputers 25X1
1. As requested, we are providing a contribution to the 305
study on supercomputers. The attachment provides a brief history
of the Japanese supercomputer industry and summaries of the three
Japanese supercomputersuppliers.
Attachment:
Contribution to Article 305 Study on Su er
rs
25X1
GI M 87-20036
Februar
1
8
25X1
,
y
9
7,
25X1
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
I I , 11 iIII11; 11- - ----
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9 5X1
SUBJECT: Contribution to Article 305 Study on Supercomputers
Distribution:
Orig - GlenFukushima, USTR
1 - Interagency Strike Force.for a.301 Investigation
on Supercomputers
SA/DDCI
ExDir
Executive Staff 25X1
DDI
DDI/PES
NIO/ECON
DD/OGI, D/OGI
CPAS/ISS
OGI/EXS/PG
CPAS/IMC/CB
C/TICD
C/TICD/TEC
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
1 II 1'111' II 1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
Contribution to Article 305 Study on Supercomputers
4. Japanese Supercomputer Industry
4.1 Brief History
4.1a Government Role: Questionable Short Term Impact
The current generation of supercomputers offered by Fujitsu, Hitachi
and NEC was developed without.direct support from the Japanese
government. Most of the circuit and systems design and production
technologies in Hitachi's and Fujitsu's supercomputers, and to a
lesser extent NEC's, are based on Japanese general-purpose mainframe
technology developed in the late 1970's. Much of the development of
this mainframe technology, however, was funded-by Japanese government
computer programs, including: the MITI-sponsored 3.75 general purpose
computer project, the Pattern Information Processing System (PIPS),
directly targeted towards the development of a supercomputer.
the-software production technology development program, as well as
NTT-sponsored programs such as the Denden Kosha Information Processing
Project (DIPS). Although these projects greatly contributed to the
overall current state of Japanese computer technology, none were
Branch, Office of Global Issues.
-------------------------
This memorandum was prepared by
GI M 87-20036
Technology Analysis
25X1
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
LI Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Future supercomputer developments, however, will benefit from direct
government support through several programs that are currently under
way, including the High Speed Scientific and Fifth Generation Computer
Projects. Although many of the stated goals of the programs are
ambitious and may not be realized, these programs could result in
important spin-offs as Japanese firms attempt to foster innovation in
supercomputer development.
High Speed Scientific Computer Project.
The High Speed Scientific Computer Project is an eight year program
initiated in 1982 to develop circuit technologies and system
architectures useful in supercomputer design and production. According
to the program guidelines, the project is designed to give the
Japanese clear-cut domination in the high-speed, scientific computer
area. The funding for this project,,is estimated to be about US $150
million, with matching funds coming from the six participating
Japanese firms, Fujitsu, Hitachi, NEC, Toshiba, Mitsubishi, and Oki.
The first six years of the program are aimed at development of
Josephson junction and gallium arsenide (GaAs) circuit technologies.
Concurrently, the project calls for the development of hardware and
software designs that are to be configured into a high speed system
during the final two years. The goals of the project include;
o Uniprocessors capable of 100 million floating point
operations per second (MFLOPS) and a multiprocessor capatileeof
10,000 MFLOPS.
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
o High-speed logic and memory devices, including; a 3,000 gate
array chip with 10 picosecond delays (a picosecond is a
trillionth of a second), implemented in Josephson junction
transistors or low temperature gallium arsenide; a 3,000 gate
array chip with'30 picosecond delay and a 16K memory device with
a 10 nanosecond access time (a nanosecond is a billionth of a
second), both implemented in room temperature gallium arsenide.
o Software design and system architecture capable of supporting
100 individual processing elements for maximum parallel
operation.
The project has not met with great success. With less than three years
to project completion, expectations appear to have been drastically
scaled back, and Japanese officials concede that they have yet to
Ai'
settle on the basic architecture for the prototype. Much of the
problems are related to controlling parallel operation efficiently.
Several Japanese engineers say that even if a prototype is developed,
they doubt that it could result in a commercially usable machine in
the near future. Additionally, research on Josephson junction
technology and GaAs devices has not moved as fast as was originally
expected. Designers say that they would now plan to use conventional
silicon chips for all but a few of the new machine's components.
Fifth-Generation Computer Project
In 1981, the Japanese initiated a new program, the Fifth Generation
- 3 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Computer Project, to define new levels of performance for computer
systems in the 1990s. It is an eight year program; calling for
funding of about US $500 million by . the. participating Japanese
computer makers (including all. of the firms in. the High Speed
Scientific Project) and the Japanese government. The program'is not
explicitly intended to promote the development of Japanese
supercomputers. It could, nonetheless, have implications for Japan's
capabilities in the supercomputer field.
An integral part of the high performance needed to implement these
fifth generation systems will be the use of parallel processing
techniques. Much of the hardware and software technology developed in
this project could be used in the design of future generation
supercomputers. Fifth generation systems (and supercomputer systems)
will be increasingly dependent on parallel processing to realize high
performance. However, mechanisms that control massively parallel
systems have been difficult to design and build.
There will likely be no
short-term benefits of this program for the Japanese supercomputer
manufacturers, although future systems may borrow on the technology
that is currently under development.
I.1b Fujitsu and Hitachi Entry in Late 1983
Fujitsu's entry in the supercomputer market were the VP-100 and VP-200
with claimed peak performance ratings of 250 and 500 MFLOPS
respectively. (Table 1 outlines-,the full supercomputer product lines
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
1 1 LI. 1J11 LL . _ -A-1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
of the three Japanese firms.) These machines were first delivered in
December 1983. A low-end model, the VP-50 (160 !FLOPS) and a high-end
machine, the VP-400 (1,140 MFLOPS) were ready for delivery in late
1985, and mid 1986 respectively.
Hitachi entered the supercomputer market at about the same time as
Fujitsu, with the introduction of two supercomputers, the S-810/10 and
S-810/20, with a claimed peak performance rate of 315 MFLOPS and 630
MFLOPS respectively. In late 1985, the company introduced a low-end
version, the S-810/5. The firm is also planning a high-end processor,
similar to the Fujitsu VP-400, that will have a maximum performance
rate over 1,000 (FLOPS.
5 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
In achieving their levels of performance, Japanese supercomputer
designers demonstrated some insights into critical factors affecting
supercomputer performance.
they
avoided potential bottlenecks by providing high-capacity memory
expansion units, improved channels,from main memory to the processors,
and increased memory storage outside and inside the processors. The
Japanese also benefited from basic software research performed in the
United States, but not yet applied by US manufacturers. For example,
Fujitsu's automatic vectoring FORTRAN compiler received critical
acclaim, not just for its vectorizing capabilities, but also for its
well designed interactive user interface.
The VP and S-810 machines also show that Japanese supercomputer
designers made some mistakes because of their relative inexperience in
supercomputer development. The designers gave too much capability to
their vector processors, which handle large arrays of data, compared
with their scalar processors which manipulate single units of data.
-6- 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
As a result, important elements of the Japanese machines sat idle
during many applications, while other portions ran too slowly.
Although the Japanese decision to base their supercomputers on their
general-purpose mainframe computer technology lessened development
costs and risks, it also resulted in diminished performance.
The supercomputers from Hitachi and Fujitsu-had an important
difference from the original supercomputers offered by their US
counterparts at Cray Research and Control Data. These Japanese
systems were designed to run IBM-compatible operating systems. They
could also use standard IBM-compatible peripherals. This was not
surprising since Fujitsu and Hitachi benefited greatly from having
IBM-compatible mainframe lines. Besides selling to the traditional
target market for supercomputers---the scientific community--Hitachi
and Fujitsu believed that they could make inroads into more
traditionally oriented commercial computing environments. These were
believed to represent about 20 percent of IBM compatible mainframe
operations.
the Japanese enjoyed reasonable success in the sale
of their supercomputers. At the end of 1985, Fujitsu had installed 17
machines---three were at Fujitsu plants, one was at Amdahl in
Sunnyvale, with another 18 systems on order. Fujitsu's low-end
machine, the VP-50 proved to be its most popular model and is likely
to comprise almost half of its future orders. Twenty-two of the
25X6
25X6
25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Fujitsu's 35 total orders were in Japan. The other 13 were overseas.
Four of the overseas orders were booked by Amdahl, which is selling
the Fujitsu supercomputers in the United States, but under the Amdahl
company label.
Hitachi has not done as well as Fujitsu. At the end of 1985, Hitachi
had only shipped seven systems and had two additional systems on
order. Three of the installations were within Hitachi. There were
several reasons why Hitachi sales were significantly below that of
Fujitsu. Performance tests indicated that the Hitachi machine was not
as fast as its Fujitsu counterpart. Additionally, Hitachi was slow to
introduce a low-end counterpart to the VP-50. Finally, and perhaps
most importantly, Hitachi restricted its sales to the domestic market,
while Fujitsu was selling its supercomputers on a world-wide basis.
In-the future, the Japanese supercomputer'suppliers will face
increasing competition When-the Japanese supercomputers
were introduced, a major selling point was that they were IBM-
compatible, while Cray and CDC supercomputers were not. The Japanese
strategy received a setback at the low-end when IBM introduced a
vector facility for its 3090 processors announced in early 1985. IBM
used a different instruction set than Fujitsu and Hitachi for their
vector processors. This will not be a major problem in Japan, but
could severly limit the sales of the Japanese in offshore markets.
The Japanese could have real problems in the low end of the
supercomputer performance spectrum because of IBM's competitive'
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
I __L_ _L_ .. II. .1 Jll.l.l; _ . -1_1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
hardware and superior. marketing power.
4.1c NEC Enters the Market in 1985
NEC announced and delivered two new supercomputers almost two years
after Fujitsu and Hitachi. In the middle of 1985, NEC had available
two systems, the SX-1 with a claimed maximum performance of 570
MFLOPS, and the SX-2 that had a performance of 1,300 MFLOPS. At that
time, the SX-2 was the highest claimed peak performance of any
commercially available supercomputer. In early 1986, NEC shipped its
first SX-1E, the low end of the SX series with a maximum performance
of 285 MFLOPS.
The NEC machines were quite different, both architecturally and
physically from the machines offered by Fujitsu and Hitachi.
Specifically, the NEC machines represented some of the first attempts
by any Japanese computer manufacturer to address some of the tough
engineering problems facing next-generation supercomputer suppliers.
Unlike Fujitsu and Hitachi, the NEC supercomputers were not an
extension of the firm's mainframe line. The NEC supercomputers used
water cooling---a first for any Japanese computer manufacturer--as
well as an advanced multilayer ceramic chip package. The system had
an extremely fast cycle time (6 nanoseconds), fastest available at the
time from any Japanese or US manufacturer. Like the other Japanese
manufacturers, the NEC supercomputers were heavily dependent on
pipelined vector operations to achieve high performance,, thus limiting
its range of applications. Additionally, the SX supercomputer had one
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90T00114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
of the fastest scalar processors then available. In general, the
hardware of NEC's SX series represented a giant step forward by the
Japanese in supercomputer development.'
Although the SX systems were impressive on paper, there were many
disadvantages to the machines. Most importantly, like the rest of the
NEC data processing line, the SX systems were not IBM-compatible.
This greatly limited the commercial acceptability of the systems.
Also, the operating systems and support software that was supplied
along with the system was considered weak, and diminished the
capability of the system to perform at high rates.
By the middle of 1986, NEC has only shipped seven machines, three
internally to NEC. NEC, though not traditionally an aggressive force
outside Japan, spent some effort trying to market the machine
Ai-
themselves in the United States. NEC only sold one machine in the US,
toga university research consortium in Houston.
the company announced last year that they would begin a joint
venture with Honeywell to market NEC supercomputers in the US. NEC
hopes to sell more than 50 supercomputers through the joint venture
over the next five
25X6
25X6
25X6
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
4.2 Principal Companies and Market Performance
4.2.1.a
Fujitsu
Fujitsu is Japan's leading computer company and second overall behind
NEC in Japanese industrial electronics, with more than US $6.5 billion
in revenues in fiscal 1985. (See table 2 for a financial comparison
of the three Japanese computer firms.) The firm is first and foremost
a computer firm; approximately 60 percent of Fujitsu revenues are
derived from sales of computers and data processing systems. (See
table 3 for a breakdown of the firms' data processing revenues.) For
the rest of its revenues, 21 percent comes from semiconductor sales,
13 percent from communications equ~pments sales, and the remainder
from sales of products including automobile electronics. It is much
less diversified than the other giants of Japanese electronics.
Compared with Hitachi annual revenues (US $20.9 billion in 1985) and
NEC (US $ 9.9 billion in 1985), Fujitsu is a relatively small firm.
Despite its size, Fujitsu has a leading-edge position in computers,
and semiconductor technologies. In computers, the company offers a
wide spectrum of products, ranging from supercomputers to mini- and
micro-computers. Fujitsu currently holds 29 percent of the Japanese
mainframe market, and more than 18 percent of the domestic
microcomputer market. Fujitsu has also recently announced some of the
most impressive Japanese mainframe systems offered to date
- 11 -
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
J _l.l~AJLLL. . JLI
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
In peripherals, Fujitsu has announced impressive
developments in erasable optomagnetic disks,.and sputtered thin film
magnetic disks. Through the Fifth Generation Project, the company is
involved in the development of systems based in artificial.
intelligence and parallel processing techniques.
In semiconductors, Fujitsu is a leading supplier of high density
memory components, and is particularly strong in developments related
to large computers--- high speed logic and gate arrays, and fast
bipolar, memories. Fujitsu is also a leader in the development of GaAs
semiconductor devices. GaAs devices offer the potential to have
higher-speed, lower-power performance than their silicon counterparts.
1
Fujitsu, along with many other Japanese and US electronics firms,
believes that GaAs components could play an increasingly important
role in the production of future supercomputer and mainframe systems,
as well as advanced telecommunications systems.
In telecommunications, Fujitsu has one of the fastest optical
transmission systems available on the market, and appears to hold the
lead in the field of optoelectronic integrated circuits. A major area
of Fujitsu interest has been ISDN technology for central office
switching. Fujitsu has announced developments in the entire range of
telecommunications systems components, including new switching and
transmission techniques as well as terminal development.
25X6
25X6
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
NEC offers a wide range or important electronics products and is a
major player in the computer, telecommunications and semiconductor
markets. NEC products are well balanced across the spectrum of high.
performance electronics. NEC revenues from computers and industrial
electronics is about 32_percent of total revenues; communications
equipment account for about 28 percent, microelectronics for about 27
percent, home electronics and other products generate the remaining 13
percent of total revenues. NEC products include:
o A full range of computers. In microcomputers,, it holds 70
percent of the domestic market for 16-bit machines. In
supercomputers, it currently offers one of the highest
performance systems available in the market.
o An extensive line of subscyiber switching products--key
telephone systems, private branch exchanges, and hybrid KTS/PBS
systems.
o A complete selection of networking equipment including
central office switches, packet switches, and a wide spectrum of
transmission equipment. For its communications equipment, NEC
relies on the quasi-privatized Nippon Telegraph and Telephone for
about 25 percent of its total sales, sales to the US
telecommunications market account for another 25 percent.
o A broad range of semiconductors and components--NEC is the
largest semiconductor manufacturer in the world merchant market,
- 13 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
I I H., _1 Jl.LL . , . 1_1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
surpassing the US-based Texas Instruments in 1985.
A-wide variety. of other equipment including consumer
.electronics,.facsimile,equipment, direct-broadcast-satellite
receivers, cellular radio sets, and semiconductor manufacturing
equipment.
In data processing, NEC has carved out a niche market in which IBM
compatibility does not matter. Its ACOS operating systems reflect NEC
historic ties with General Electric and Honeywell. Unlike Fujitsu and
Hitachi, NEC has been able to concentrate on mainframe technology and
performance instead of anticipating IBM product announcements. This
has, however, resulted in NEC being the least active Japanese large
systems supplier outside of its domestic market.. Strategic alliances
with France's Bull and Honeywell in the United States may help to
increase NEC foreign participation in the near future.
Hitachi generated over US $20 billion in revenues in fiscal 1985.
Although the foundation of the firm is based on the older technologies
of steel, chemicals and electricity, its long term goal is to
transform itself into a company increasingly based on electronics.
Total revenues generated by the industrial products category ---which
includes computers and semiconductors-- accounted for less than 20
percent of Hitachi revenues in 1981, had risen to 30 percent or`nly four
years later.
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
1 .1 1 i.I ? 1L , .. .
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Hitachi is the most broadly diversified Japanese supercomputer
supplier, revenues are generated from five major product areas:
o 30 percent--industrial electronics including semiconductors,
computers and telecommunications equipment.
o 22 percent--consumer products.
o 18 percent--wire and cable, metals; chemicals.
o 16 percent--industrial plants and machinery.
o 14 percent--power systems and equipment.
Hitachi strategy in industrial electronics is simple; the company has
isolated semiconductor technology as a key to its future success in a
wide range of electronic goods. By focusing money and engineering
talent on the semiconductor-business, Hitachi has moved into the
number two spot in the world semiconductor industry, behind NEC.
Because about 70 percent of Hitachi semiconductor revenues comes from
the sales of semiconductor memories, the firms financial performance
is closely linked to the rise and fall of that segment of the
industry. Hitachi is attempting to diversify into other segments of
the industry, with principle targets being microprocessors and
microcontrollers.
Historically, telecommunications equipment has ranked a distant third
in Hitachi's corporate strategy, behind semiconductors and computers.
However, telecommunications will.likely play an increasingly important
- 15 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
role in Hitachi's future. Recently, Hitachi has introduced several
new digital PBX systems designed to move it into some new vertical
markets such as hospital and general business applications. Hitachi
is also the sole supplier for cellular radios to AT&T.
In Japan, Hitachi is a full-line supplier of computer systems, from
personal computers to supercomputers. The most important task it now
faces is successfully moving that product line into export markets.
Traditionally, Hitachi has used original equipment manufacturer
agreements to sell its computers abroad. Virtually all of the Hitachi
computer products sold in North America carry another company's name,
or are built into another companies product. In the North American
market, Hitachi continues to rely on National Advanced Systems to
market its mainframe products. This agreement is tenuous, and Hitachi
may be looking for alternate distributors and seeking ways to bolster
its US computer sales and service network.
16 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
1 I . _L J_.- . L I . . . 1 . 1 1 1 1 L . . . _1L I
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Table 1
Japanese Supercomputer Product Lines
Maximum Livermore Memory Announced FCS*
Performance Loops Capacity
(MFLOPS) (Harmonic (MBytes)
Average,
.MFLOPS)
Fujitsu
VP-400 1440 na
VP-200 500 19.8**
VP-100 250 18.7
VP-50 160 na
256 04/85 1H/86
256 07/82 4Q/83
128 07/82 12/83
128 04/85 11/85
Hitachi
S-810/20 630 14.7 256 08/82 4Q/84
S-810/10 315 na 128- 08/82 4Q/83
S-810/5 160 na 128 09/85 1Q/86
NEC
SX-2 1300 na
Sr-1 570 37.7
SX-1E 285 na
256 04/83 06/85
256 04/83 1986
128 10/85 03/86
* First Customer Shipment
* Harmonic average for Cray X-MP-1 is 13.7 MFLOPS
- 17 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Table 2
Comparison of Leading Japanese Computer Firms Financial Status
Fujitsu
Hitachi
NEC
1985 Total Revenue
6,563
20,919
9,899
1985 DP Revenues
4,309
2,885
3,761
1985 Total R&D
524
1,223
249
1985 Net Income
235
768
249
DP Export In percent (est)
19
20
14
PP&E as a percentage of
Revenues over last 3 years
15.4
8.8
14.0
All number in $US millions unless otherwise noted
- 18 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
LL i I L IL_ ILL, _ .. AL---- - -- --
-Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9
Table 3
Data Processing Revenue Breakdowns
Fujitsu
Mainframes
38*
29
32
Minicomputers
10
0
0
Microcomputers
4
2
9
Datacommunications
9
0
12
Peripherals
25
49
28
Software
6
7
10
Maintenance
9
10
9
Other
0
3
0
* percent total data processing revenues for Japan fiscal 1985
- 19 - 25X1
Declassified in Part - Sanitized Copy Approved for Release 2012/02/28: CIA-RDP90TO0114R000404380001-9