For Convex Computer Corp., life has been one big blast since 1982, when it began building the first high-speed number-crunchers for engineering and science labs that couldn't afford $10 million supercomputers. Convex's minisupercomputers can do everything from design drugs to simulate car crashes, but they cost as little as $300,000. Their success has enabled Convex, based in Richardson, Tex., to outlive a raft of competitors, win some 400 customers, and grow to $209.3 million in sales last year.
Round Two. To keep revenues growing at 15% to 20% a year, Chief Executive Robert J. Paluck reckons he now must push into the larger, more demanding market for midrange supercomputers, which have up to five times the speed of minisupers. By now, companies such as Digital Equipment Corp. and IBM have responded to Convex' older machines by adding special hardware to their computers. But the move into full-fledged supercomputers puts Convex on a collision course with the industry's dominant force, Cray Research Inc. Cray, searching for growth, is also pushing into the midrange market--if only to keep Convex from winning customers who one day might want a full-size Cray. Late last year, Cray began shipping the Y-MP2E, a $3.3 million midrange super.
FIRST BLOOD. Already, the sparks are flying. Knowing that Convex planned a May introduction of its first midrange super, the $2 million-to-$8 million C3, Cray made a preemptive strike. On Apr. 8, it announced the Y-MP4E, a $5.5 million-and-up big brother to the Y-MP2E. Both machines are aimed at customers who felt they couldn't afford Crays. "We must do more to make our products cheaper," says Cray CEO John A. Rollwagen. Cray figures that the two small Y-MPs together will account for 50% of its 1991 installations.
Convex, meanwhile, plans to launch the C3 on May 7. It's not providing many technical details yet, except to say the eight-processor machine will be 2.5 times faster than its current C2 line of minisupers for four times the price. It will be the industry's first machine to rely fully on ultrafast gallium arsenide chips.
Seymour Cray's Cray Computer Corp. has been working on a gallium arsenide design since it was spun off from Cray Research in 1989. Analysts say each C3 processor will perform about 125 million arithmetic operations per second, putting it within reach of the Y-MP4E (table). "The C3 is clearly going to be the company's engine of growth for the next several years," says Barry F. Willman of Sanford C. Bernstein & Co., a securities firm, who predicts it will account for about 25% of Convex' revenues this year and 40% in 1992.
Convex and Cray Research bring different strengths to the midrange market. Because it sells cheaper machines, Convex has more customers. And they are more likely to move up to a C3 than switch to Cray. However, Cray is a $804 million company with enormous prestige and clout in supercomputing. "Where the battle will be is for the not-as-yet committed," says analyst Jeffry Canin of Montgomery Securities.
Convex faces mounting pressures. It says that first-quarter earnings will be lower than analysts' expectations. And it will have to sell to a new audience, including the CEOs of its customer companies, rather than just to engineering departments. Selling pricier systems will also take more time and patience. "Convex is going into a marketplace that they aren't as familiar with as we are," says Rene Copeland, Cray's marketing director.
MEMENTO MINI. Talk like that doesn't intimidate gung ho Paluck, who co-founded Convex with former Data General Corp. computer architect Steven J. Wallach. Paluck is confident that Convex can outmaneuver Cray. And he's hardly ignoring the still-lucrative minisuper market: When it takes the wraps off the C3, Convex is also expected to announce an enhanced C2 minisuper.
As they go head-to-head with Cray, Paluck and Wallach are very mindful of the vagaries of the computer business: On a patio at the company's headquarters, there are more than 20 names etched in cement. Chopp Computer, ETA Systems, etc. are the companies that started alongside Convex in supers, but fell by the wayside. The lesson, says Paluck: "You've got to have a brilliant strategy, and you have to actually execute it. Otherwise, you become a tombstone."