Cray Approaches Moment of Truth in Shaky Market

ant the true measure of a company's character? Forget how it deals with success. Look instead at how it copes with the downturns, when Wall Street analysts begin lining up scapegoats and the business press launches long, dirge-like features on The Fall of a Once-Great Company: It got complacent, it missed a product cycle, it squandered money on vain projects, it lost focus.

This moment of character-revealing truth may be coming at Cray Research, a trail-blazer in the now-troubled supercomputer market since 1972. Profit and revenue growth is fast losing momentum. Mainstay defense and science customers are broke. Computing behemoth IBM has finally decided to lumber into the market -- siphoning off huge chunks of potential and real business from weaker entrants. But perhaps most troubling, the company's cash-cow, the old-line vector supercomputer, is widely thought to be a Jurassic throwback.

When Steven Spielberg adapted Michael Crichton's Jurassic Park for movie audiences, he swapped out the Cray Research supercomputer featured in the book for a slick, new massively parallel Thinking Machines hot rod. Thinking Machines even rented out a movie theater for the film's debut and threw a bash afterwards to crow about its marketing coup to reporters.

Still, Cray has been tested before. It was Thinking Machines, after all, and not Cray, that recently filed for bankruptcy. And at this year's supercomputing show nearly everyone pegged Cray, along with IBM and Silicon Graphics, as survivors of the supercomputer industry shakeout.

"In the life of a company, everyone hits the wall," says Chairman and CEO John Carlson, an accountant by training who took over the helm of Cray from John Rollwagen in 1993. That was just after Cray crashed into its first wall: revenues dropped to $797 million in 1992 and the company racked up $15 million in losses.

At the same time, Rollwagen left the Eagan, Minn., headquarters of Cray for Washington to defend his nomination as Department of Commerce Secretary, only to be thwarted by charges of conflict of interest.

The intervening period, with Carlson in charge, has been one of spotty success and frequent frustration. On the one hand, the company has filled in huge gaps in its product line, working with companies such as DEC and Sun Microsystems to enter the booming workstation server market and so-called "massively parallel processor" business.

In its first year with products to sell, the firm's commercially oriented server business will tally $50 million in sales -- an impressive achievement for a new venture, much less one guided by a company hardly known for its marketing savvy. This is a firm, after all, that relied for years on a handful of intelligence and defense customers.

Still, Cray has been unable to grasp Wall Street's holy grail of sustained profitability and revenue growth. Its stock has been punished accordingly, trading at nearly one-third its peak price in 1989.

Carlson, who joined Cray as employee No. 75 in 1976, admits the quest for corporate redemption is far from over. But he also exudes a scrappy confidence, noting that Cray has put in place an impressive and thorough lineup of products -- one that should deliver top- and bottom-line gains by 1996.

"The company has a wonderfully designed new machine," agrees Michel McCoy, with Lawrence Livermore National Laboratory, who has had a far easier time using Cray's new massively parallel computer than those of the firm's competitors.

But as the recently bankrupted Thinking Machines or Kendall Square Research can attest, technical excellence does not a business case make. The equation is as simple as it is challenging: Cray's business model has always been to provide more processing power at a higher cost. Its customers could afford it. But now Cray must provide more power at a lower cost. Essentially, Cray must replace machines it sold for tens of millions of dollars a few years ago with machines that cost a few hundred thousand -- and still make a profit.

In tech-business parlance, this act is referred to as self-cannibalization -- the conscious destruction of one product line with a new one that is cheaper, but has the potential for higher-volume sales. Intel is a master at this in its chip business, as is Microsoft in software.

So ultimately the real test of Cray's character may be its willingness to eat its own lunch -- and emerge lean and mean enough to fend off the likes of IBM and Silicon Graphics.

Cray has attacked this challenge in three ways, each corresponding with a distinct segment of the supercomputer industry: vector processors, massively parallel processors and symmetrical multi-processors.

The first category covers old-line supercomputers, the hulking, air-cooled monsters put out by Cray and a few Japanese companies. The second category -- a cheaper potential replacement for the first -- gained prominence in the mid-1980's, with the founding of firms such as Thinking Machines, nCube and Kendall Square Research. These companies try to tether hundreds and even thousands of processors to achieve mastodonic improvements in performance at a fraction the cost. The final category attempts a loose-coupling of powerful workstations to perform jobs often done by mainframes. These SMP systems, as they are known, are not supercomputers in the traditional sense of being the most powerful machines on earth. But they have the potential to replace other machines at the low end of the supercomputer market.

Carlson admits it will be a challenge to sustain R&ampD across all three efforts, especially as the company dips into R&ampD to focus more on marketing and sales. Historically, Cray has spent about 15 percent of its revenue stream on R&ampD and about 15 percent on marketing and sales, far higher for R&ampD than the industry average.

Carlson's job, in part, is to put that ratio in line with the more common split of about 18 percent for sales and marketing and about 12 percent for R&ampD.

That move will require some finesse. Engineers, not marketers, have called the shots at Cray. And the nature of the kinds of problems Cray solves is changing. As the firm moves into commercial markets, developers used to throwing prodigious computing resources at solving profound scientific problems will now be absorbed in less prosaic tasks: improving database marketing tools, building video servers for the infobahn, divining critical correlations in stock-trading data. The firm's approach to this problem has been to separate its commercial efforts as much as possible from its scientific and technical computing operations. Cray's commercial server group -- Cray Research Superserver -- is based in Beaverton, Ore., and it is the first Cray operation to be set up as a separate company, with its own budget.

Cray has also leveraged the work of others for its new efforts. The firm uses the DEC Alpha chip for its T3-D massively parallel processors. And it has allied with Sun Microsystems to essentially produce and market Sun's high-end computers. In embarking on this strategy, Cray may avoid having to do the increasingly expensive job of microprocessor design and manufacture on its own -- and it can leverage other company sales and marketing forces.

Initial results at Cray have been promising. Cray is anticipating about $100 million in sales for its MPP line. Meanwhile, the company continues to work on the next generation of old-line vector processor. It is also crafting offerings for the mid-range technical and scientific market -- and whisperings at the recent supercomputer conference in Washington, D.C. say resulting products have begun to eat into high-end sales for Silicon Graphics.