Why Cray's Number Cruncher Got Crunched In JapanNeil Gross
It is a scene that by now is all too familiar: a war of words between Japan and the U.S. over high-technology trade. This time, the subject is supercomputers, and the precipitating incident is the decision by Japan's National Institute for Fusion Research to lease an NEC Corp. SX-3 system for $625,000 a month instead of a C90 from industry leader Cray Research Inc. After Cray volubly objected to the award last summer, the matter was reviewed by a panel of Japanese experts. Now, they've endorsed the selection of NEC. That prompted Cray CEO John A. Rollwagen to call a press conference in Tokyo and cry foul. The contest for the Fusion Institute contract is over: The experts have ruled, and there is no avenue for appeal. But on Oct. 9, U.S. Trade Representative Carla A. Hills promised to "scrutinize" the decision on Cray's behalf.
Once again, Tokyo and Washington are on a collision course that could have--indeed, should have--been avoided. In theory, Hills's investigation could lead to a renewed warning about Japan's barriers to supercomputer trade. But it goes beyond supers. The heart of the matter is the Japanese government's long-standing vow to support high-quality U.S. imports. If Tokyo were sincere about that pledge, Cray might expect to win many more supercomputer bids where the Japanese government holds the purse strings.
SPEC SQUABBLES. Unfortunately, the fusion lab dispute has devolved into a tit-for-tat over obscure technical matters. Rollwagen, waving four pages of benchmark test results, told reporters that Cray's system exceeds all but one of the Fusion Institute's speed requirements. The institute disputes those data, but Rollwagen's conclusions were supported by test scores leaked to a Japanese trade magazine last summer and confirmed recently by sources close to the bid.
His opponents go on to argue that the extra power available from Cray's machine is irrelevant: The fusion scientists would never use it. Give us a break, say Cray officials: What scientist would thumb his nose at additional supercomputing capacity? Finally, Japanese officials point to the bid's requirement that the machine work with specialized storage devices. Cray Research calls that a bogus requirement, designed mainly to stack the deck in NEC's favor.
This dialogue may be absorbing for techies in the audience, but it's a distraction from the more important issue. Subtle nuances aside, Japanese experts concede that, overall, the Cray and NEC computers are quite comparable in cost and performance. And that being so, the Japanese government, with a $40 billion trade surplus to work off and a long-standing pledge to boost imports, ought to be bending over backward to buy American in this case. Rather than setting the stage for endless squabbling, Japanese officials should be scouring the U.S. market for big-ticket purchases that meet or exceed Japanese requirements for quality and performance. Cray Research supercomputers, which play an important role in U.S. research and defense efforts, fit that bill precisely.
Indeed, Cray is the undisputed leader in large supercomputers, holding 67% of the world market. Fujitsu Ltd. comes in second at 20%, and NEC third at 6%. But look at Cray's track record in Japan: Thanks to the 1990 supercomputer trade accord, its market share in commercial installations there has inched up to 25%. Yet in the public sector, which includes government-funded universities and research labs, Cray is stuck at a trivial 8%. Rollwagen is hardly wrong in declaring: "In a fair competition, we win everywhere except Japan."
Even so, Rollwagen's casting of the issue as a matter of free trade may be doing his cause more harm than good. It's hard for Cray to maintain its free-trade oratory in Tokyo while Washington continues to discourage Japanese supercomputer sales in the U.S. No stateside government labs--the biggest supercomputer users in the world--have bought a Japanese machine yet. And whenever a U.S. university or government agency has flirted with buying a Japanese super, Cray has quietly lobbied to block the move.
FREEBIES. Last year, political pressure blocked Fujitsu from donating a $17 million machine to a Colorado consortium of environmental scientists. Congressional critics objected to the idea of a Japanese give-away. Yet last year, Cray itself donated an X-MP system to the Energy Dept. in support of a national high school supercomputer program. And nobody objected.
Clearly, Cray needs every sale it can get. It feels mounting pressure in the market it pioneered and has dominated for 15 years: On Oct. 15, the company began a restructuring that will eliminate 650 jobs. But Cray needn't resort to strident free-trade rhetoric to make its point. It would do better to focus attention on the more substantial issue: Japan's failure to live up to the goals it has outlined for increasing high-tech imports. Aside from the flak it would get from NEC and Fujitsu, the Japanese government risks hardly any downside in buying more C90s: It would be good for public relations, good for the balance of trade, and good for Japanese science.