Science & Technology
HUMBLE PIE FOR T.J. RODGERS
It was a reversal of fortune that T.J. Rodgers was slow to comprehend. Cypress Semiconductor Corp., the San Jose (Calif.) company he founded, had been a model of entrepreneurial success. By making ultrafast chips for small niches, it managed 37% annual growth and 15% profit margins even through chip recessions in the late 1980s. Then came 1991's fourth quarter. When earnings fell a stunning 27%, Rodgers blamed production glitches and vowed to compensate by aggressively pushing his broad product portfolio. "I felt that if we could crank the revenue up, we could float the rock again," he recalls.
He couldn't. In short order, Cypress found itself in a vicious price war and failed to get a key product to market. In this year's second quarter, it recorded its first-ever loss as a public company--a $1.7 million deficit, vs. income of $10 million a year earlier--as revenues dropped by 12%, to $65.8 million. Rodgers' critics had been right, it seemed: His company was ill-equipped to compete when its niches grew large enough to attract big rivals. "He was mugged by reality," says his former boss, Advanced Micro Devices CEO W.J. "Jerry" Sanders III.
That experience has led to a conversion of sorts. The CEO who once castigated his Silicon Valley brethren for failing to outduel foreign manufacturers has begun copying their methods. He is moving assembly to Thailand and laying off 400 employees, 21% of his work force, two things he once vowed never to do. And he promises to finally give up some control of day-to-day operations to concentrate on new products and markets. "It's a defeat," Rodgers concedes. "I've had to change my mind-set." Not to mention his expectations. Analysts see Cypress earning just $3.3 million this year, vs. last year's $34.2 million, on a 6% decline in revenues, to $270.9 million. Next year, they project a rebound--14% revenue growth, to $310 million, with profits rising to $19.5 million. After bottoming out at 738 in July, the company's stock has jumped four points, though it's still 40% below a year ago.
Back then, Cypress was riding high. Rodgers was boasting of his all-American company's efficient manufacturing and quick design turnarounds at his five subsidiaries. Pervasive computer monitoring kept him informed of every employee's and every product's progress, and he upbraided slipshod managers personally. That unusual style had sustained Cypress' winning streak even as Japanese behemoths such as Toshiba, Fujitsu, and Hitachi entered his market for fast chips called static random-access memory, or srams, which account for about 30% of company revenues. But the enfant terrible underestimated his competitors at home. Motorola Inc. and Micron Technology Inc. have both jumped into the sram market and early this year helped start a price war, analysts say. Result: One key Cypress sram, used primarily in high-end desktop computers, now sells for $3.70, vs. $8.10 then.
FLOODING? Disaster also struck another product called programmable logic devices (plds), a hot seller with a wide range of uses. Both Cypress and Altera Corp., from whom Rodgers licensed the pld design in exchange for manufacturing rights, accuse each other of flooding the market with plds in an effort to boost revenues. The resulting price war slashed profits at both companies, and now Altera CEO Rodney Smith has decided not to license his latest generation of plds to Cypress, turning instead to Japan's Sharp Corp. "I doubt if we would give Cypress any more products," Smith says, though he will keep making faster versions of his older chips at one Cypress subsidiary in which Altera still has a 16% investment.
The price wars showed that Cypress had to be more efficient. That prompted Rodgers to seek help from experienced executives such as Fred B. Bialek, formerly of National Semiconductor Corp. and now a Cypress director. Bialek persuaded Rodgers to shift assembly offshore, a move that will save $17 million a year when completed in March. Bialek also insists that Rodgers is "working real hard at not spending micromanagement time" on daily operations. "In the area of technology, the guy's fantastic. But I want him out of the things that he doesn't have that kind of insight into."
Indeed, as a manager Rodgers made classic mistakes. His engineers were spread too thin, developing too many new products while trying to update the 2,500 the company already had. That made everything late. Cypress has yet to produce the latest generation of sram, a 1-megabit chip, which Motorola and Micron have been selling for months. His microprocessor subsidiary, Ross Technology Inc., is a year late delivering a new-generation chip for Sun Microsystems Inc. So for now, Texas Instruments Inc. has won Sun's business. In response, Rodgers has hired consultants Thomas Group Inc., run by former Intel Corp. manager Alex W. Young, to help cut chip design time from 102 weeks to 65 weeks and order-to-revenue time from 180 days to 99 days--all by March. As a byproduct, Cypress' lineup will be pared by 25% and several new chips in development will be killed.
UNKNOWN FACTOR. These moves ought to restore Cypress to health. Getting it back on the fast track will take longer--and require finding more high-revenue niches than Rodgers has had until now. He is looking into flash memory chips, which are faster replacements for other erasable memory chips in computers, plus chips for multimedia computing that add sound and video processing. To offset the loss of Altera's new products, Rodgers has made an investment of undisclosed size in Quick Logic Corp., a startup making the latest generation of programmable devices, known as field programmable gate arrays.
The real unknown is the long-delayed sparc microprocessor being developed for Sun by Ross Technology, now due by mid-1993. Sun could buy $30 million to $80 million worth next year, says Daniel L. Klesken, an analyst at Robertson, Stephens & Co. Still, Rodgers is worried that other major computer makers don't use the Sun-designed chips--which retards growth at Cypress. Right now, Sun is selling fewer than 300,000 sparc engineering workstations a year, and ti may supply chips for most of those. "The idea of several suppliers is good," says David Ditzel, director of advanced systems at Sun Microsystems Laboratories. "But as the chips get more complex, that gets harder to do."
So, Rodgers is considering the once-unthinkable--grabbing the coattails of Intel, which he had vowed to thrash with his sparc microprocessor. He now says Ross Technology may adapt some chips that go with sparc, such as those that act as its short-term memory, to work with Intel microprocessors. It isn't quite the lofty goal he once aimed for. But it may beat flying too high, only to take a hard fall.Richard Brandt in San Jose, Calif.