The Washington Post
April 1, 1990, p. C1
Last week’s allegations of a plot to smuggle nuclear warhead detonators from the United States to Iraq, together with reports that China was resuming its secret sales of ballistic missiles to the Middle East, have refocused attention on a frightening fact of modern life: Third-World tyrants, armed with missiles and A-bombs, are fast replacing the Soviets as the greatest threat to American cities.
Yet at the same time, a small group of Bush administration officials are working to heighten that threat — by exporting U.S. supercomputers to Brazil, Israel and India, three countries with secret nuclear weapon and ballistic missile programs. If the officials succeed, they will reverse the Reagan administration’s policy of keeping these machines away from countries that are trying to get the bomb.
In Brazil, the supercomputers could hasten the day when a nuclear-capable missile is sold to Libya or Iraq, and bring Brazil much closer to finally testing the nuclear weapon it is trying to produce. In Israel, supercomputers could design smaller, lighter nuclear warheads to attack more targets and could reduce the time required before an Israeli missile can bring Moscow within range. In India, the machines could accelerate the development of ICBMs to menace all of China with H-bombs.
The supercomputer was invented in the mid-1970s to design U.S. nuclear weapons. It has since become the most powerful tool known for designing both nuclear weapons and ballistic missiles. Today, no U.S. nuclear weapon or missile design is physically tested until it is optimized in computer models.
A supercomputer can simulate the implosive shock wave that detonates a nuclear warhead, calculate the multiplication of neutrons in an explosive chain reaction and solve the equations that describe fusion in a hydrogen bomb. For missile design, it can model the thrust of a solid-fuel rocket, calculate the heat and pressure on a warhead entering the atmosphere and simulate virtually every other force affecting a missile from launch to impact. Because of the billions of computations needed to solve these problems, a supercomputer’s speed is invaluable for finding design solutions in a practical length of time.
The lack of such machines will not stop a country from making its first atomic bomb. But with a supercomputer, a country can design more efficient nuclear warheads with a minimum of tests and design long-range missiles to carry the warheads to their destination. For countries with limited money and manpower, those advantages are crucial.
Proponents of the exports argue that there were no supercomputers when the first bombs were built, and thus a supercomputer is not the key to going nuclear. But in the early days the lack of computing power was made up for by tests. The Department of Energy estimates that about 180 physical tests were needed to design the high-explosive part of a 1955-vintage nuclear weapon. Today fewer than five tests are needed because of computation. According to DOE, a team of scientists using the calculators of the 1940s would take five years to do what a Cray supercomputer now does in one second. IBM wants to sell a supercomputer to Embraer, an arm of the Brazilian Air Force in the missile-for-export business. Through its ownership in another firm called Orbita, Embraer is now trying to turn Brazil’s Sonda IV space launcher into an intermediate-range nuclear-capable missile. In January 1988, Libyan arms-buyers offered to pay Orbita’s development costs in exchange for long-range missiles and the means to make them. Brazil publicly rebuffed U.S. protests of the Libyan offer. Embraer also exchanges personnel with the research arm of the Brazilian Air Force, called CTA, which West German intelligence says is secretly making nuclear weapon material. The U.S. supercomputer could wind up designing missiles for Libya and nuclear weapons for Brazil.
The University of Sao Paulo is IBM’s second intended Brazilian customer. West German intelligence says that one of the university’s own institutes is designing centrifuges to enrich uranium — a step leading to atomic bombs. Also on campus is a group called IPEN, which has secretly built lab-scale centrifuges and a plant for extracting plutonium (the nuclear weapon material that destroyed Nagasaki) and is planning a secret reactor that will create enough plutonium for one atomic bomb per year.
Cray — America’s other supercomputer giant — wants to sell a machine to Technion University, the Israeli MIT. In 1987, a Pentagon-sponsored study revealed that Technion was helping design Israel’s ballistic missile re-entry vehicle. According to U.S. officials, Technion’s nuclear physicists work at Israel’s secret nuclear weapon complex at Dimona, where a reactor makes plutonium for atomic bombs. According to a Technion brochure, the Cray that Technion wants to buy will be able to do in one month calculations that now take eight years.
Hebrew University would also get a supercomputer. The Pentagon study found that its physicists work at Israel’s nuclear weapon lab at Soreq, whose scientists were “developing the kind of codes which will enable them to make hydrogen bombs.” The study added, however, that the Israelis did “not yet have the capability to carry out the necessary calculations.” A U.S. supercomputer would provide exactly that capability.
At the Weizmann Institute, the third Israeli applicant, scientists are studying the high-energy physics and hydrodynamics needed for nuclear bomb design, as well as the use of lasers to enrich uranium — the most advanced method for making nuclear-weapon material.
Israeli Military Industries, the fourth Israeli applicant, makes Israel’s biggest rocket motors. This includes the ones that power the Jericho II intermediate-range missile and the big new Shavit space launcher. NBC News reported in October that Israel is selling the means to make these motors to South Africa. In return, Israel will be able to test the Shavit over the empty ocean off the South African coast. To reach its full range — which includes Moscow and Western Europe — the Shavit will need an accurate re-entry vehicle. This is exactly what IMI’s supercomputer could help design. IMI may even decide to share the computer with its South African customers.
In India, U.S. supercomputers would go to the Indian Institute of Science (IIS) and the Indian Institute of Technology (IIT). Both are studying stresses on rocket bodies and supersonic combustion. There is even a project to study the performance of solid rocket fuel through computer modelling. All this learning will go straight into missiles. When India launched its first intermediate-range missile in 1989, it used a first-stage solid-fuel rocket produced by the space program. Thus, Indian missiles could profit directly from the U.S. supercomputers.
In May 1989, CIA Director William Webster told a congressional committee that India appeared to be working on a hydrogen bomb. Over the next few years, India will be trying to perfect an efficient fusion warhead and an accurate long-range missile to carry it. Supercomputers could powerfully aid both endeavors. To sugar-coat the sales, proponents are proffering security plans. U.S. government inspectors — whose identity and competence is yet to be defined — could visit computer sites to see who had used the computer and for what. But that may not be possible. Embraer is free to design aircraft — to compute fluid flows around aircraft noses, fuselages and wings. But the computations are basically the same as those used for missiles. Hebrew University is free to study nuclear fusion by using exactly the same computations one needs to design hydrogen bombs. It would be an extraordinary inspection that could detect a violation under such conditions.
To make matters worse, Brazil, Israel and India are already violating inspection agreements made in the past. In order to import West German nuclear equipment, Brazil promised to allow the International Atomic Energy Agency to verify that the equipment was not used for weapons. But Brazil has prevented inspection by refusing to report a single one of the numerous German deliveries over the past 10 years. In order to import heavy water to run its Dimona reactor, Israel promised to allow Norway to verify that the water would not be used to make atomic bombs. However, Israel has repeatedly rejected Norway’s demands for inspection. And India promised — as a condition of importing two reactors from the United States for its site at Tarapur — to restrict all the reactors’ plutonium to peaceful use. However, India is now threatening to declare the plutonium — enough for about 320 Nagasaki-sized bombs — free for use in nuclear weapons because of an absurd interpretation of the nuclear sales agreement.
Inspection would also be costly. The sale to Embraer is worth only $ 400,000, the price of two “vector processors.” The processors would raise Embraer’s existing IBM mainframe to supercomputer speed. After U.S. government inspectors visit Brazil several times a year for the next few years, the costs of inspection will easily exceed the profits from the sale. U.S. taxpayers will thus foot the bill for IBM’s decision to sell computers to people whose promises are suspect.
Computer-makers argue that the growth market is now overseas, but the sales figures tell a different story. Cray has sold about 140 supercomputers in the United States and exported about 100 to developed countries and NATO allies — none of which is a proliferation risk. Each machine has a security plan. IBM is estimated to have sold about 300 vector processors to the same market. Compared to these sales, the prospective handful to the proliferators is a drop in the bucket. To get the drop, however, the exporters are ready to put the whole world at risk.
To avoid such a risk, the Commerce Department has issued specific regulations. Five criteria determine whether a country can import U.S. computers: It must belong to the Nuclear Non-Proliferation Treaty; have all of its nuclear activities under international inspection; have an agreement for nuclear cooperation with the United States; support in its public statements and policies the goal of nuclear non-proliferation; and be generally cooperative on non-proliferation policy matters. None of the candidate nations — Brazil, Israel and India — meets a single one of these criteria. This was why the Reagan administration wouldn’t approve the exports. The proposed sales would make a mockery of U.S. nuclear non-proliferation policy. The United States is now trying to stop France from selling the technology for the “Viking” rocket motor — a powerful, ICBM-sized booster — to Brazil. Our government justly fears that Brazil will use it to make an intercontinental missile. Can we still credibly oppose the sale after hawking supercomputers to the very Brazilians who make such missiles? And about one year ago, the United States stopped West Germany from helping Libya build a poison gas plant. Could the United States credibly do that again, after selling supercomputers to the very Israelis who are working on hydrogen bombs?
If the need to restrain France and Germany were not enough, there is the need to restrain Japan. In 1984, the United States and Japan — the world’s only supercomputer suppliers — agreed not to sell the machines to developing countries that had rejected the Non-Proliferation Treaty. In 1986 they renewed the agreement. All of the proposed exports will breach that accord. If the United States does break faith with Japan by making these deals, the result could be a no-holds-barred race to sell supercomputers to the Third World. Moreover, the loss of the Japan accord would make it impossible to bring new suppliers into it. This would be a disaster for U.S. nonproliferation policy and for world security.
The Commerce Department is already chipping away at the Japan agreement. It wants to raise the agreement’s definition of a supercomputer from 100 to 150 megaflops (million floating-point operations per second — a measure of mathematical computing speed). This would let the University of Sao Paulo get its machine (110 megaflops) without a security plan. The more cautious Japanese want to stay at 100. To justify the boost to 150, IBM and Cray say that their lower-end machines are now nearing the 100-megaflop threshold. Soon, they say, even desktop workstations will have near-supercomputer speed, and the limit should be raised to reflect the advance of technology. By the same logic, however — since the MX missile is better than the early Atlas and the neutron bomb more advanced than our earlier fission devices — we should sell Atlas missiles and ’50s-vintage A-bomb designs to the Third World.
Commerce and State, the departments pushing the deal, are suffering from export-mania and from heavy lobbying by IBM and Cray. Even the Pentagon now favors the exports, despite the fact that it vetoed every one of them under Reagan. Only the Arms Control and Disarmament Agency — with the help of the Department of Energy — seems determined to follow the export rules and preserve U.S. credibility.
President Bush promised after his election to come out swinging against nuclear and missile proliferation. But if export advocates get their way, he will take a dive in the first round.
Gary Milhollin is professor of law at the University of Wisconsin and director of the Wisconsin Project on Nuclear Arms Control, which tracks the spread of nuclear weapons to developing countries.