While the concept of parallel processing isn't new, turning it into a profitable business is.
In the past two years, more than 30 established companies and start-up firms have introduced their versions of parallel-processing machines for what they hope will be the next multibillion-dollar market in computers.
However, this proliferation of companies and architectures, combined with the current industrywide computer slump, has created the perfect ingredients for a shakeout in a market that barely has gotten off the ground.
"I think the shakeout is going to come very, very quickly over the next two years," said Jeffry Canin, a senior official at San Francisco's Hambrecht & Quist Inc. "It's definitely a market that's large enough -- and growing -- for three or four major vendors, but not for a dozen or more. Any company that doesn't have a product out in the next 12 months is in trouble."
Canin estimates the potential of the "minisupercomputer" market, which includes both parallel processing and other approaches to high-speed, high-cost supercomputing, will grow from about $125 million this year to more than $1 billion by the turn of the decade.
While bubbling with growth potential, the nascent parallel-processing market may relive the computer industry's experiences with minicomputers and personal computers. Initially, dozens of large companies and venture-capital-fueled entrepreneurs plunged into a perceived growth technology. The result was market chaos and too many companies chasing too few customers.
"Just as in the other new-wave computer architectures, the change gets oversubscribed," said David Rogers, vice president of marketing for Sequent Computers, an Oregon-based parallel-processing computer start-up firm that recently has begun to establish a presence in the market. "You saw that in PCs and, if you look way back, remember there were once 50 minicomputer companies."
Of course, those shakeouts led to a mere handful of companies becoming the dominant, and profitable, suppliers in fast-growing market segments.
But what makes the future particularly perilous for parallel-processing firms is the variety of approaches to the technology, each with its own claims to price and performance, that have led to intense confusion by potential customers. That, in turn, has led to far-slower-than-expected growth.
Indeed, these problems already are taking a toll. Denelcor of Aurora, Colo., a parallel-processing supercomputer company, went out of business in October after after six profitless years.
"There's been more media hype than market penetration," said Joe Rizzi, chief executive officer of ELXSI, a San Jose-based parallel machine company. "While everybody's excited about parallel processing, they're still not quite sure what it means -- and 15 companies each offering different definitions simply confuses things. Many potential customers have gone from active curiosity to wait-and-see."
There is an extraordinary variety of parallel processing offers.
Three-year-old Encore Computer, Sequent Computer and Alliant Computer all offer high-speed scientific computing power through a parallel architecture that can handle more than a dozen processors simultaneously. Other companies, such as Bolt Beranek and Newman, offer a machine capable of coordinating up to 256 processors. Thinking Machines Corp., a venture company, plans to offer its "massively parallel" Connexion, a computer which is supposed to handle thousands of processors simultaneously.
Of course, International Business Machines Corp. and other established companies have parallel projects underway.
"From the ashes of the shakeout will rise some very strong companies that will be setting the pace for scientific computing in the 1990s," said Sequent's Rogers.