The nation's leading supercomputer company announced yesterday that it is canceling its most advanced computer project and that its chief designer would be leaving the company to attempt to develop the technology on his own.
Cray Research Inc. said it had decided to discontinue its MP supercomputer project because it appeared that the attempt to develop the machine would overtax the company's resources. "The project has grown significantly beyond our original vision, both in terms of technological risk and budget, and we believe that it no longer meets the objectives or style of Cray Research," the company's chairman, John Rollwagen, said in a statement.
Analysts said the departure of the designer, Senior Vice President Steve S. Chen, could have significant long-term ramifications for Minneapolis-based Cray, which makes high-speed, high-power supercomputers used for such varied tasks as predicting weather, designing automobiles and complex electronic circuitry, and creating highly sophisticated computer animation effects.
"It is a blow, no question about it, when you lose a star like Steve Chen," said Ulric Weil, who follows the computer industry for Gartner Securities Corp. "There's a big hole there that's not easily filled."
"Long-term, it's sad and unfortunate," said Richard Schaffer, an analyst at Technologic Partners, a New York research group. "Steve Chen is very important."
Cray stock, one of Wall Street's high flyers in recent years, skidded $9.25 on the news, closing at $103.75.
Chen headed a 200-person team at Cray that was developing the MP project based on parallel processor technology, in which dozens of smaller computers would be linked in parallel to simultaneously process information, allowing unprecedented computer power and speed. The MP project was foreseen as a successor during the 1990s to Cray's existing Cray-2 and X-MP computer lines, its soon-to-be-released Y-MP computer and its under-development Cray-3 machine.
Analysts estimated that Cray had sunk about $35 million into the MP project and had about $17 million budgeted for research and development on it this year. But it apparently became apparent that development of the MP would involve more than Cray's typical strategy of developing new machines by taking advantage of basic computer technologies developed by others. Instead, Rollwagen said in a telephone interview, Chen became increasingly involved in researching fundamental technology he felt was needed to develop the MP model.
"The project was moving away from the way we do stuff here," Rollwagen said. "Steve, on the other hand, believes firmly that some basic research needs to be done into the technology itself in order to create a new generation of computers."
The two-year-old project had "started to take on a life of its own," said analyst Gary Smaby of Piper Jaffray & Hopwood, a Minneapolis brokerage.
Chen could not be reached for comment.
The development of the MP also apparently was going to be extremely expensive. Smaby estimated the price tag of the project at $100 million over the next few years -- as much as Cray plans to spend on research and development in 1987 and far more than it spent on any of its previous computers -- with less certain chances of success.
"It will take substantial resources -- money -- to do," Rollwagen said.
The departure of Chen does not leave Cray high and dry. Its founder, Seymour Cray, continues to work on new products for the company, and currently is putting the finishing touches on the Cray-3 model, to be introduced later this decade. "He is beginning to think about what's next, and I can tell you that what's next is not retirement," Rollwagen said.
Other researchers at the company are developing other machines, analysts said, and the company hopes to continue working on some of the technology that was involved in the MP project, according to Rollwagen.
"This work continues, and certainly fills any gaps that were left by the cancellation of the MP project," Rollwagen said. "But I'd be foolish to say we're not disappointed."
Analysts said the abandonment of the MP could slow Cray's growth as it hits the 1990s, and some observers suggested that the decision may reflect a new caution on the part of the company. "That's the hallmark of a maturing company, that they become risk-averse," Weil said. "There may be that aspect to it, that Cray feels, 'We'd better be safe than sorry.' "
And Schaffer suggested that Cray's admission that it felt it could not handle the complexity of a project like MP could slow the overall development of complex parallel processor supercomputers. "If Cray isn't going to push that frontier, who will?" he said. "This seems to suggest that that's much too risky, and not as certain as it seems."
But Smaby said the future of supercomputing clearly would continue to be in increasingly complex parallel processors. "Both users and vendors believe parallel processing is the future of supercomputers, and I don't think that Cray strategists would dispute that," Smaby said.
Chen was given the opportunity to stay with Cray to work on a different project, a Cray spokeswoman said, but chose instead to leave the company and attempt to develop the computer on his own. Cray said Chen would be allowed to work on the technology outside the company as long as he does it for his own company. He will lose rights to the technology if he joins a competitor, Rollwagen said.
"Certainly, Steve has the brain power and the technical competence to lead such a project, but as a practical matter there is a question if it can be carried forward," Rollwagen said. However, he added, "We don't want to put anything in Steve's way. ... Slavery went out 100 years ago. What Steve does for a living is design supercomputers, and he's going to design supercomputers."
Analysts said Chen should have little trouble raising money to start his own company to develop the parallel processor technology. "I can assure you there would be dozens of venture capitalists that would be glad to stake him to start his own company," Weil said.
Chen's departure from Cray is in some ways reminiscent of a similar move made a decade ago by Seymour Cray, who left Control Data Corp. to pursue his dream of a supercomputer and founded Cray Research to do it. "It sounds like he's taking a leaf out of Seymour Cray's book," Weil said.
But Smaby said the rapid advancement of the supercomputer industry, and the great expense involved in developing the next generation of supercomputers, made an independent effort much more difficult now than it was when Cray left Control Data. "The ante is significantly higher today," he said. "You can't just walk out and start a company in a garage. The capital requirements are much greater today."