Computer Pioneer Seeks Protection : Technology: Thinking Machines falters just as its ‘parallel processing’ is finally gaining acceptance.
Thinking Machines Corp., the company that revolutionized the esoteric science of supercomputers, said Monday that it will file for Chapter 11 bankruptcy protection--just as the “parallel processing” technology it pioneered is finally being accepted by the mainstream computer industry.
The Cambridge, Mass.-based company, which has sought investors for months to ease a cash crunch, said it will now lay off one-third of its 425 employees in an effort to return to profitability. The company will shrink its business to core operations such as government systems and seek to sell or license its patents and networking technology.
Thinking Machines’ troubles renewed concerns about the ability of small U.S. technology companies to survive over the long haul and reap the benefits of their technological breakthroughs. Such concerns could hamper efforts by other pioneering companies to sell in tough overseas markets, especially Japan.
Problems at other small supercomputing companies, coupled with rumors that Thinking Machines was in trouble, became self-fulfilling prophecies as sales of the company’s multimillion-dollar machines dried up.
“Every day I would meet with customers and I would have to distinguish ourselves from Kendall Square,” said Chief Executive Richard P. Fishman, referring to sharp sales declines and an accounting scandal at rival Kendall Square Research.
Thinking Machines’ problems come just as it was finally declaring victory in its long, uphill battle to persuade the computer world that so-called massively parallel processing is the wave of the future in high-performance computing.
Traditional supercomputers use extremely elaborate, custom-built central processing units that can execute computer instructions very rapidly. But such processors are very expensive to design and build, and they often need elaborate cooling systems and other infrastructure.
Massively parallel machines, by contrast, use hundreds or even thousands of low-cost microprocessors similar to those used in personal computers.
The computing task is divided up, with different pieces solved simultaneously by different processors.
When Danny Hillis, a computer scientist at the Massachusetts Institute of Technology, founded Thinking Machines in 1983, his concept was disparaged as conceptually interesting but hopelessly impractical.
Many computer designers believed the software complications inherent in massively parallel processing--figuring out a way to break up the problem and then reassemble it when the pieces had been solved--would be virtually impossible to overcome.
But the company gained recognition when it sold two systems to Dow Jones Co. for speeding through database searches. It also won a high-profile sale to American Express, which used the machines to predict buying habits of credit card users.
Gradually, other computer companies were convinced. Sales of parallel computers are expected to rocket from about $650 million last year to $5 billion in five years, according to Howard Richmond, director of high-performance computing at the Gartner Group, a consulting company in Stamford, Conn.
Thinking Machines was the market leader.
But all along, the company’s largest sales went to government agencies, which were interested in the technology for use in such high-tech defense initiatives as the “Star Wars” project.
“From its inception, it was the U.S. taxpayer that supported the company,” Richmond said.
Large government contracts helped create a company that was oriented toward pushing the envelope technologically without developing the commercial applications necessary to sustain sales, he said.
Meanwhile, computer makers such as IBM, Digital Equipment and Convex entered the market with cheaper systems. And Intel, eager to gain a foothold in an industry that promised to consume lots of microprocessors, also dove in aggressively.
Thinking Machines was mismanaged, too, concedes Fishman, who succeeded co-founder Sheryl L. Handler as chief executive in October. “We have this panoramic view of the Boston skyline.
There was this attitude that the sky is the limit--we pay twice what Lotus pays across the street and they are profitable,” Fishman said.
Thinking Machines also recently lost a key government contract and a big expected sale in Japan, leading to a loss of $20.5 million on just $82.3 million in sales last year.
The company’s fall could end up leaving mud on the face of the U.S. government--and eventually hurt other American companies trying to sell in Japan.
Japanese government institutions and corporations bowed to U.S. pressure and became major buyers of U.S.-made supercomputers, despite their worries about being able to sustain a long-term relationship with small, potentially unstable companies.
Fishman, who plans to resign when the company files for bankruptcy, tried to look at the bright side of the company’s plight. “The good news is that this technology that used to be disparaged as something from the lunatic fringe has proven itself,” he said.
Sales and profit or loss for Thinking Machines Corp., in millions of dollars: