The Man Who Can See The Future For Chips

Dauvin of SGS is uncanny at calling industry swings

It's Friday afternoon in Paris. Jean-Philippe Dauvin glances out the window of his corner office and predicts strong winds and cloudy skies for the coming weekend. Weather forecasts are old hat to Dauvin, the chief economist for chipmaker SGS-Thomson Microelectronics. While doing his military duty 30 years ago, he did a stint with the French weather service, and he insists that projecting chip industry cycles isn't so much different.

Every major semiconductor company has its own resident forecaster. Several market researchers and Wall Street analysts also maintain computerized financial barometers. Then there is the cooperative statistical effort by the Semiconductor Industry Association and its counterparts in Europe and Asia. So there is no lack of semiconductor oracles. Yet, in semiconductor circles, Dauvin is famous for gazing at his crystal ball and foretelling industry swings ahead of the pack. SGS has been exploiting his insights, grabbing market share from competitors and steadily climbing in industry rankings.

Jim Eastlake, associate director of research at Dataquest Inc.'s London office, pays Dauvin the ultimate chip-watcher tribute: "His analyses are more advanced than ours." Dauvin's current forecast again breaks with the consensus. If he's correct, the chip business is in for a bigger boom than Dataquest figures.

The previous example of Dauvin's outfoxing other seers happened in mid-1995. Chip sales were soaring, on their way to a 40% gain for the year. Other mavens were calling for a repeat surge in 1996. So producers were merrily pumping money into new manufacturing capacity. Not SGS. The French-Italian company began nursing its resources in June because Dauvin foresaw--correctly--a sharp downturn at yearend. Not being saddled with excess capacity helped SGS-Thomson remain profitable last year, unlike most chipmakers.

Bucking the tide in 1995 was a big gamble for Pasquale Pistorio, president of SGS. If Dauvin had been wrong, SGS could have lost a year or more in its drive to become one of the world's top half-dozen chipmakers before 2010. Getting there could mean nothing less than the company's survival. Dauvin sees a massive shakeout looming in the next decade, because chip factories--wafer fabs, in industry jargon--will grow too costly for all but a select few. Half of the world's top 25 chipmakers could be forced to toss in the manufacturing towel, he figures. Thanks to Dauvin's foresight in 1995, SGS reached its 1996 target of climbing into the Top 10, up from No.14 in 1993. But if the coming crunch proves as severe as Dauvin envisions, 10th place could still be on the fringe of vulnerability, so SGS hopes to move up a few more notches for insurance.

LOTS OF HITS. Dauvin credits his soothsaying wizardry in part to the work of Marc Vodovar, an economist now with Texas Instruments Inc. In 1973, Vodovar and Dauvin were both at BIPE, an economics consultancy in Paris, where Vodovar pioneered a model of the chip industry for the French government. In 1982, when Dauvin jumped ship, Vodovar laid out his model's principles, and the two have been conferring regularly ever since. "Basically," says Vodovar, "the concepts we use are the same."

However, Dauvin is constantly tweaking his model to boost its ability to call turning points in advance. Vodovar says Dauvin is usually proved right, "even though some of his ideas are early." Overall, he figures Dauvin is batting about .800. In the high-stakes semiconductor game, that practically gives him a license to mint money.

To investors and outsiders, the chip business may seem fraught with violent, capricious swings. But a big picture emerges from a plot of the industry's past three decades. Every few years, the market peaks, slumps or slows, then bounces back. But on average, chip sales have grown 16.4% annually. Dauvin sees that as the overall climate.

Aberrations can upset the long-term trend temporarily, but the $140 billion chip industry often makes matters worse by overreacting. "The minute oversupply kicks in and prices start plummeting, companies put new fabs on ice," he observes. Once things pick up, "they stampede to ramp up new plants," which inexorably leads to yet another glut.

One reason for the knee-jerk reactions, Dauvin believes, is that managers are drowning in statistics--and missing the big picture. "You don't need a lot of data" to understand the chip business, he says. "It's all in the capacity." His presentations to Pistorio are limited to two charts. The simple-is-better approach also carries over into Dauvin's outside activities. He teaches a course in semiconductor economics at the Ecole Superieure d'Electricite (Supelec) in Paris and is the chip consultant for the United Nations Industrial Development Organization. To initiate neophytes into the intricacies of semiconductor economics, he often spouts parables: Chips are like the classic pork cycle in economic textbooks, he says. It takes 15 months to grow a pig, the same as building a fab--and pork supplies drive price fluctuations that affect the investment in new pig farms.

NO COMPUTER. Even his economic model is disarmingly simple. It consists of a few equations that can be solved with a calculator: A computer is notably absent from his office. "I'm like an old-fashioned doctor with one or two good tools," he quips. Chip supplies are the key parameter, but other variables include chip prices and sales of semiconductor manufacturing equipment. His latest model adds a new component: wafer fab capacity utilization, which is important because it drives investment. Currently, utilization stands at 82%, so prices are fairly soft. But a 95% level would bring sizzling price increases and spur investment.


So what does this new model foretell? For starters, the chip market will double by 2000 and hit $290 billion to $300 billion. Dauvin holds to that outlook, despite the recent plunge in memory-chip prices that has prompted many gurus to lower their sights to $250 billion or less. That's too low, Dauvin insists, because "there is no disconnect in the long-term trend" of 16.4% annual growth. This year's 6% growth rate is way below the line, but Dauvin asserts it will soon be made up, with sales climbing at least 20% in 1998 and then zooming 30% higher in 1999.

As for the next downturn, Dauvin says it's not in the cards before mid-2001, when a gradual buildup of capacity could sap prices again. Or it might hold off until 2002. Should things change, Dauvin trusts his model to sound an alert. By reacting sooner than its rivals, SGS expects to continue enlarging its piece of the silicon pie.

    Before it's here, it's on the Bloomberg Terminal.