What’s Good For Silicon Graphics is Not Necessarily Good for America: Some Super Computer Sales Imperil US Security

(Washington, D.C.): Earlier in this
century, the sentiment that U.S. national
interests were defined by those of
powerful domestic economic interests was
summarized by the expression,
“What’s good for General Motors is
good for America.” Although such
narrow parochialism has subsequently come
to be viewed as a generally unsound basis
for policy-making, it seems to have been
embraced by the United States’ computer
industry and its friends in the Clinton
Administration.

Specifically, the desire to improve
the international market-share enjoyed by
this well-heeled and politically active
industry prompted President Clinton on 6
October 1995 to effect a very substantial
liberalization of export controls on
powerful computing technologies and
systems. Unfortunately — as demonstrated
in recent hearings conducted by the
Senate Governmental Affairs Committee’s
Subcommittee on International Security,
Proliferation and Federal Affairs,
chaired by Senator Thad Cochran
(R-MS)⁽¹⁾
— what companies like Silicon Graphics
have deemed to be good for their
bottom-line is manifestly not good
for U.S. security. In the roughly 18
months following President Clinton’s
decision unilaterally to decontrol
supercomputers capable of making
2,000-7,000 million theoretical
operations per second (MTOPS), such
devices have wound up in several ominous
locales.

For example, according to a floor
statement made by Sen. Cochran on 19
June:

“Based on statements from
the Russian Minister of Atomic
Energy and from United States
Government officials, that there
are at least five American
supercomputers in two of Russia’s
nuclear weapons labs:
Chelyabinsk-70 and Arzamas-16.
Minister Mikhailov of the Russian
Ministry of Atomic Energy has not
been reluctant to proclaim what
these high-performance computers
will be used for, and he said in
a speech in January they will be
used to simulate nuclear
explosions, and that the
computers are, in his words, ’10
times faster than any previously
available in Russia.’ Four
of the five supercomputers we are
aware of publicly in Russia’s
nuclear weapons labs came from
Silicon Graphics…”

Sen. Cochran noted that the Chinese
military was also benefitting from such
American-made computers:

“We also know at
least 47 high-performance
computers have been exported
without licenses to the People’s
Republic of China. One
of the computers sold also by
Silicon Graphics is now operating
in the Chinese Academy of
Sciences. The Chinese Academy of
Sciences is a key participant in
military research and
development, and works on
everything from the DF-5 ICBM —
which, incidentally, is capable
of reaching the United States —
to uranium enrichment for nuclear
weapons. There can be no question
about the Chinese Academy of
Science’s status as a military
end-user.”

Sen. Cochran added on 7 July,
“…One Sun Micro systems machine
that we just learned [about] last week is
now running at a Chinese military
facility in Chungsha after being diverted
from Hong Kong.”

‘Selling
the Rope’

Such findings led Senator Cochran to a
grim conclusion:

“We learned that the
United States was a proliferator
of weapons technology that was
threatening the security of the
United States, and putting at
risk United States servicemen,
servicewomen, other interests,
and other assets and interests
throughout the world….We
were giving countries like Russia
and China and others the capacity
to improve the lethality, the
accuracy, and the capabilities of
nuclear weapons systems through
the exporting of technology that
they were using to simulate tests
— which they would not otherwise
be able to do — and to upgrade
the quality and accuracy of their
missile delivery systems and
weapons systems.”

The fact that Senator Cochran’s
amendment, which was co-sponsored by Senator
Richard Durbin (D-IL), had been
overwhelmingly approved in the House
meant that its defeat in the Senate
became the object of a concerted effort
by the Clinton Administration and its
friends in the U.S. computer industry. As
members of the conference committee
responsible for resolving differences
between the two versions of the FY1998
Defense authorization bill are likely to
be subjected to similar pressure from the
CEO of IBM and other interested parties,
it is useful to examine critically some
of the arguments made by opponents of the
Cochran-Durbin amendment:

Item: ‘These
Computers Are Not Really
Supercomputers’

Senator Barbara Boxer
(D-CA), who co-sponsored with Senator
Rod Grams (R-MN) a substitute
amendment that called simply for a
further “study” of the issue,
made a point repeatedly argued by
industry opponents of the Cochran-Durbin
proposal:

“…Computers in the 2,000
through 7,000 MTOPS ranges are
mid-level computers that are
widely available. They are not
supercomputers. Let me repeat
this because I know there is a
lot of confusion on this issue.
Computers in the 2,000 MTOPS
through 7,000 MTOPS range are not
supercomputers. In fact, many
computer servers will top the
2,000 MTOPS threshold next
year.”

While it is true that the
state-of-the-art in computing has
inexorably increased the power of
advanced computers, machines in the
2,000-7,000 range still represent
extremely potent machines. The best
personal computers on the market are
presently in the 250-300 MTOPS range.
Relatively small numbers of workstation
systems used for specialized applications
such as computer-aided design and
manufacturing — not office
networking and other common
administrative functions — are in the
low 1,000 MTOPs range. Systems
capable of performing between two and
seven thousand million theoretical
operations per second represent a
computing power that qualifies them
clearly as supercomputers.

Item: ‘These
Computers Are Not Militarily
Relevant’

Opponents of the Cochran-Durbin
amendment went on to make an even more
startling assertion: They argued
that supercomputers in the 2,000-7,000
MTOPS range had no military value.
As Senator Kit Bond put
it: “Supercomputers that do military
work these days are 20,000 MTOPS to
650,000 MTOPS. They are talking about
computers 10 times — 10 times — the
range that would be covered by this
[proposed requirement for export
licensing].” This argument reflected
the Clinton Administration’s assessment
of the situation as expressed in an 8
July letter to Senate Majority Leader
Trent Lott from Secretary of
Commerce William Daley. It
argued that “critical defense
applications that justified export
controls were clustered at levels above
7,000 MTOPS.”

Of course, the relevant test
is not how fast the fastest computers can
perform, but how much value would
2,000-7,000 MTOPs supercomputers be to
potential adversaries’ militaries.
In fact, such devices are substantially
more powerful than those are being used
today in various U.S. weapon systems —
and that are likely to be used for years
to come. The utilization of these
supercomputers by the Russian and Chinese
nuclear weapons design bureaus speaks
volumes about the considerable military
significance of this sort of computer
technology — especially when
combined for use in parallel processing.
As Sen. Cochran pointed out on 19 June:

“In 1986 [when the upper
bounds of supercomputing
performance were considerably
lower than today’s], the
Department of Energy published an
unclassified report entitled, The
Need for Supercomputers in
Nuclear Weapons Design. The
report’s conclusion included this
statement: ‘The use of high-speed
computers and mathematical models
to simulate complex physical
processes has been and continues
to be the cornerstone of the
nuclear weapons design
program.'”

Interestingly, Sens. Grams and Boxer went
to some lengths to express their
commitment to preventing, as Senator
Grams put it,”computers above the
7,000 level from reaching military end
users, that’s for sure.” Sen. Boxer
waxed even more emphatic: “Does [the
Grams-Boxer amendment] mean we should
allow companies to sell any computer at
any level to any country notwithstanding
our national security interests? Of
course not. Our national security
interests are paramount. They are
paramount. Our export policies absolutely
must ensure that our foreign policy and
security objectives, particularly as they
relate to non-proliferation and
counter-terrorism, are maintained.”

It will be worth watching to see how adamant
these and like-minded Senators will be
about preserving the 7,000 MTOPS limit
when Dr. Seymour Goodman — the academic
and long-time advocate for decontrolling
computers, whose sole-source-contracted
study was used to justify ending
licensing requirements for 2,000 to-7,000
MTOPS supercomputers — completes a
follow-on study that is expected to
argue for decontrolling still faster
machines.

Item: ‘These
Computers Are Available from
Foreign Sources’

A number of Senators argued that
foreign availability ensured that the
only effect of denying export licenses to
U.S. firms would be to give overseas
competitors the business. As Sen. Boxer
put it:

“We know the Japanese make
these computers. We also know
companies in France, Taiwan, the
United Kingdom, and Germany all
manufacture computers in the
2,000 through 7,000 MTOPS range.
And how about this? China is
producing computers at the 13,000
MTOPS level, far above the level
which the Cochran amendment seeks
to control.”

In point of fact, at this
juncture, foreign availability for
supercomputers is very limited.
As Senator Cochran advised his colleagues
on 19 June: “It is not like some
other country has these systems available
for sale on the market. They do not.
We are the state-of-the-art producer of
the supercomputers. Japan has the
capacity to produce supercomputers as
well, but their export policy is more
restrictive now than ours is.”
(Emphasis added.)

It is to the credit
of the Japanese government that it
continues to exercise stringent controls
on the sale of supercomputers. It has
done so in the face of “foreign
availability” created by the Clinton
Administration’s irresponsible decision
to make many of these devices available,
even to nations that have been enemies of
Japan and America — and that may revert
to that status in the future. Thanks to
this Japanese restraint, it is
not too late for the United States to
restore effective export controls on
supercomputers in the 2,000-7,000 MTOPS
range.

Unless corrective action is promptly
taken, however, the Clinton policy of
indiscriminately transferring advanced
computer chips and manufacturing
technology to the vast majority of
nations around the world will, in due
course, give rise to a situation in which
there are indeed many alternative foreign
suppliers of supercomputers. The wisdom
of laissez-faire export policies with
respect to both of these key technologies
needs to be urgently reassessed.

Item: ‘Alternatives
Exist to Government Licensing of
Supercomputers’

1) The Honor System: Opponents
of the Cochran-Durbin amendment claimed
that the present arrangement — in which
computer exporters are left to judge the
character of the prospective end-user —
is working satisfactorily. As Sen.
Cochran pointed out on 19 June, though,
this amounts to an “honor system,
[whereby companies are] policing
themselves and deciding themselves
whether or not the end user is going to
be a military entity or will be putting
the supercomputer to a military
use.” He went on to observe that:

“Unfortunately, some
companies have already been
tempted to take a chance.
Maybe they were not sure; maybe
they were tempted by the profits
of the transaction. Whatever the
motivations and the
understandings or lack of
information, or for whatever the
reason, we have known that some
transactions have involved the
sale of supercomputers, without
objection from our Department of
Commerce or our Federal
Government to those who may be
putting computers to a military
use, or maybe military entities
themselves.”

This reality prompted Sen. Cochran to
declare with evident exasperation:

“According to the CEO [of
Silicon Graphics], Edward
McCracken, it was his company’s
understanding that the computers
were for environmental and
ecological purposes. It may
be that Silicon Graphics was
unable to determine whether a
Russian nuclear weapons lab was
going to be the military end-user
or if its supercomputers would be
put to a military end-use. But it
seems from the statements made by
the Atomic Energy Minister in
Russia that they certainly are
available to them for those
purposes….

“The Commerce Department
maintains that President
Clinton’s supercomputer export
control policy is working.
Commerce continues to make this
claim despite the fact that the
administration’s policy has
allowed American supercomputers
to be shipped to Russia’s and
China’s nuclear weapons
complexes, and who knows where
else. If this policy is
working, what would a policy that
wasn’t working look like? Would
there be more
supercomputers in Russia and
China, or would we know
absolutely that our
supercomputers were in Iran,
North Korea, or other terrorist
states?”

The following day, Senator Cochran
elaborated on the folly of this approach:

“Computer companies
do not have the capacity to make
determinations on their own about
the use to which the computers
they are selling in the
international market will be put,
or the relationships between
prospective purchasers and
governments, particularly in
the case of China or Russia.
The U.S. government, though, has
the capacity, through its
contacts worldwide, to do a much
more reliable and accurate job of
assessing whether or not someone
would be a purchaser who would
use these computers to enhance
the [lethality] of nuclear
weapons or missile technology to
put our own citizens at risk, the
lives of Americans at risk, in a
way that they would not otherwise
be, but for the sale of our
computer technology.”

2) Commerce’s Selective,
Public List of Bad Actors: The
evident salience of the Cochran argument
prompted his opponents to suggest a
no-less-defective approach: having the
Commerce Department publish a list of
end-users who are security risks. On 20
July, Sen. Cochran analyzed the flaws in
this scheme:

“…In our hearings in the
Governmental Affairs Committee
the Administration officials
talked about the fact that the
reason they did not publish and
make available a list of end
users or potential purchasers of
these computers at this time was
because of diplomatic
considerations and the questions
about whether it puts in jeopardy
our intelligence-gathering
capabilities and a
number of other issues that
concerned them enough so that
they do not now make available
this list even privately to
exporters of
supercomputers.”

It is further evidence of the
readiness of the Clinton Commerce
Department to subordinate national
security to trade interests that it
decided on 30 June to publish what it
called an “Entities List.”
On 7 July, Senator Cochran dissected its
shortcomings:

“…The ‘Entities List’
consists of 13 locations in 5
‘Tier 3’ countries that can
receive an American supercomputer
only…subject to a license. So,
now the total list of proscribed
end users consists of 15
entities. On this list are
Chelyabinsk-70 and Arzamas-16 in
Russia which have already
received at least five American
supercomputers and parts of the
Chinese Academy of Sciences,
which also is now manufacturing
more modern nuclear weapons with
America’s finest technology.

“Because of this list, now
America’s computer exporters know
that they need a license to ship
a high-performance computer to
any of these entities. What
about other entities, though?
What about the Chinese company
that shipped ring magnets to
Pakistan last year for use in its
nuclear program? Why isn’t that
company on the list? It has been
subjected to sanctions imposed by
our Government, and it is not on
our Government’s list as a
prohibited end user. What about
the Chinese company or government
entity that shipped M-11 missiles
to Pakistan and now, according to
press reports, is helping
Pakistan build a factory for the
indigenous manufacture of M-11
missiles? Why isn’t that entity
on the list? What about the
Russian company or government
entity helping Iran to upgrade
its nuclear program and ballistic
missile programs, why aren’t they
on the list?”

“…This list does
not solve the problem. If
anything, it makes it more
confused, it makes it more
difficult for American exporters
to determine who should or should
not receive American
high-performance computers. In
many ways, this list is worse
than nothing.”

The Bottom Line

In the end, the salience of the arguments
made by Senators Thad Cochran and Richard
Durbin proved insufficient to overcome
relentless lobbying by the executives and
employees of companies like IBM and by
senior Clinton Administration officials.
The Senate agreed by a vote of 27-72 to
the substitute amendment offered by
Senators Grams and Boxer. The
Center for Security Policy commends Sens.
Cochran, Durbin and their 25 colleagues,
however, for putting the national
security before the parochial
self-interest of a relatively small
number of computer firms.⁽²⁾

The issue now goes to a House-Senate
conference committee where similar
pressures will doubtless be brought to
bear. If Senate conferees have any
question about whether to recede to the
House position, they would be
well-advised to bear in mind a
development that occurred immediately
after the Senate vote: On 14 July, Agence
France Press reported that the PRC’s
official China Daily served notice that:

“China will open up
its defense sector to foreign
investors next year as a
way of bringing electronic
warfare technology to its army.
The decision, which the newspaper
called ‘crucial to ensure rapid
technical renovation’ of the
People’s Liberation Army,
received the endorsement of one
of China’s top army directors,
General Liu Huaqing.”

The report went on to note that Liu
called for “strengthening
international military-related electronic
technology exchanges and upgrading
China’s military electronic
equipment” and that “the
decision was adopted by the Commission of
Science, Technology and Industry for
National Defense and by the PLA’s
equipment department. ‘China’s
defense-related electronics should no
longer be hidden from foreign investors,’
Su Huining, an official of the national
defense commission, said. He called on
foreign investors to display their wares
in Beijing next May during the China
International Defense Electronics
Exhibition.” (Emphasis added.)

The time has come to discourage American
companies from seeking short-term profits
at the expense of long-term national
security. As Sen. Cochran put it on 20
June: “[The Administration’s
supercomputer decontrol decision] was
supposed to be a policy that both
enhanced our ability to compete in the
international computer market but at the
same time protected our national security
interests. It worked on the one hand, but
it has failed on the other.” We can
ill-afford any more such failures.

– 30 –

1. See the
Center’s recent Decision Brief entitled
Profile In Courage: Peter
Leitner Blows The Whistle On Clinton’s
Dangerous Export Decontrol Policies
(No. 97-P 82,
19 June 1997).

2. According to
the sponsors of the Cochran-Durbin
amendment, only 6.37 percent of
American computer exports would be
affected by the amendment’s licensing
requirement.