Masters Of The Clean Room

For Jin Daeje, it's astonishing how little--and how much--things have changed in 10 years. Then and now, Jin could revel in the prestige of doing vital research on semiconductor design for the premiere producer of those ubiquitous memory chips known as DRAMs. Only then, the company was IBM. Now, it's Samsung Electronics Co., a chip nonentity in 1983. And Jin is no longer just another researcher with a Stanford PhD. Since deciding in 1985 that, as a native, he had "an obligation to be back in Korea," Jin has become Samsung's managing director of product development. He has shepherded three generations of memory chips that have sent Samsung's chip fortunes soaring.

Jin's chips, in fact, have lifted the Seoul company to No.1 among the world's suppliers of DRAMs, the semiconductor industry's best-seller. Samsung's DRAM business shot up 35% last year, to $1.2 billion, besting DRAM rival Toshiba Corp. by $69 million and boosting the Korean company's total chip sales by 29%, to $1.9 billion. By matching the Japanese on quality and delivery--and beating them on price--Samsung is likely to claw its way into the top 10 of chipmakers this year. "Samsung is one of the amazing stories of the chip industry," says Bart Ladd, manager of U.S. DRAM marketing for NEC Corp.

It's a lot more than that. Samsung's rise proves that Tokyo's approach to industrial policy needn't be peculiar to Japanese culture. Copying the tactics of Japan's Ministry of International Trade & Industry (MITI), the Seoul government used low-interest loans and pro-export policies to spur innovation. And it offered financial incentives to lure home Korean engineers and scientists working abroad. By the mid-1980s, Samsung and other Korean companies had a critical mass of research-and-development infrastructure and talent. Given Korea's success, says Jim Handy, an analyst at Dataquest Inc., the MITI model seems sure to be copied by others--India, Malaysia, maybe even China.

And to some extent, perhaps, by the U.S. as well. Michael G. Borrus, co-director of the Berkeley Roundtable on the International Economy (BRIE), argues that Washington has too often tried quick-fix protectionism to make American industries more competitive. The exception is Sematech Inc., the government-backed consortium in Austin, Tex., charged with restoring U.S. prowess in chipmaking equipment. Borrus is encouraged that the Clinton Administration sees Sematech as the pattern for similar efforts in emerging and risky technologies, such as flat-panel displays.

Skeptics question how durable Samsung's lead may be. Its chip business is more vulnerable than most to the price swings that periodically pummel the industry because 75% of its revenues are from low-margin commodity chips. U.S. and Japanese companies, by contrast, have diversified into microprocessors and other specialized chips, which normally produce fatter profits.

FOOTING THE BILL. Samsung's leaders reply that they intend to go on confounding the skeptics. Jin and his peers have so far avoided the shortsightedness of Silicon Valley, which during the 1980s reflexively chopped budgets for production equipment during cyclical downturns. That left them short of capacity in the upswings that followed like clockwork. Meanwhile, Toshiba, NEC, and Hitachi, having invested throughout the slump, would grab more market share.

This time around, it's Samsung with the deep pockets--and the victims are the recession-mired Japanese. With demand for chips languishing in their huge domestic market, wary Japanese managers have been slow to build new chip plants. That has left them ill-equipped to meet soaring demand for DRAMs. Not so Samsung, which exports about 80% of its output to the U.S. or fast-growing Asian markets. For the past five years, it has pumped $500 million or more annually into new facilities--as much or more than some companies twice its size. And this year it will invest $980 million, more than any producer except Intel Corp., the world's richest chipmaker.

Most of this year's spending is earmarked for facilities that will crank out 16-megabit DRAMs in volume. These hold more than 16 million bits of data, equal to about 700 typewritten pages. Samsung also has budgeted $700 million for R&D on futuristic chips that will store 256 megabits. When IBM, Toshiba, and Siemens looked at their estimate of a $600 million design tab for 256s, they decided to share the costs. But Samsung Electronics, backed by its $50 billion parent, Samsung Group, is going it alone. "Money has been no problem," says Jin. "We ask for $1 billion, and we get it."

HEAD START. Such ambitious investments have already led to some stunning milestones. One came in 1990. Samsung wowed IBM, Digital Equipment Corp., and other customers by showing off a fully functioning 16-megabit chip, one of the first that really worked. It began shipping commercial samples of these chips last year--for the first time launching its latest DRAM generation in virtual lockstep with the Japanese.

Samsung's prospects with this product are excellent, analysts say, because it has better equipment. Its most advanced production line, in Kihung, will crank out a typical 20,000 wafers a month. And because each round silicon slab is 8 inches in diameter--vs. 6 inches at most facilities in Japan--Samsung gets up to 75% more chips from each wafer. The higher productivity lets Samsung charge less. Tack on recent worldwide price increases, and an observer stationed in Tokyo by U.S. industry figures that Samsung could reap close to $3 billion in total chip sales this year, a 50% jump. By contrast, NEC Corp. Senior Vice-President Hajime Sasaki says his company's chip revenues will grow just 7%--because NEC doesn't have the capacity to satisfy all its DRAM demand.

The Japanese can't be kept down forever. Toshiba expects to narrow the gap in DRAMs with its new 8-inch plant in Mie Prefecture, which came on line earlier this year. NEC, meanwhile, is shifting more DRAM production to its U.S. factory in Roseville, Calif., where output is sheltered from the effects of the strong yen. Also, 80% of the machinery in Korea's chip plants is imported, much of it from Japanese suppliers. "That means Samsung is dependent on Japanese technology to raise productivity," says Kanro Sato, general manager of Toshiba's memory division.

Korea also faces growing competition from modern chip factories going up in Malaysia, Singapore, Taiwan, and China--some with Japanese funding. Most worrisome, the profit structure of the memory business may be facing a sea change. Starting with the 64-megabit generation, which should hit the market in limited quantities next year, some analysts believe that demand will begin to atrophy because the storage capacity of each chip will be so cavernous. "Then the old strategies won't work anymore," predicts Dataquest analyst Akira Minamikawa in Tokyo. Similar warnings have been sounded before, however--and new demand has always emerged to sop up whatever memory is available.

In any event, Samsung is already anticipating such problems. It has set up design centers for more advanced, software-intensive chips in the U.S., Germany, Taiwan, and Hong Kong. And by cultivating alliances with U.S. equipment companies such as Applied Materials Inc., Korean industrialists hope to lessen their dependence on Japanese production gear. Indeed, bureaucrats at Korea's Science & Technology Ministry broached the idea of a sweeping collaboration between the Korean and U.S. high-tech sectors when President Bill Clinton visited Korea last July.

Absent such a grand alliance, Samsung's growth may slow as it branches into other kinds of chips. That's because innovation in these higher-tier markets is driven by design. And so far, says Borrus of BRIE, "the Koreans have demonstrated no particular ability in creative design." That doesn't mean they won't, he adds. But even the Japanese have yet to make it over this hurdle. For now, Samsung has a cushion: Even if DRAM demand begins to wane with the 64- and 256-megabit generations, the company can count on about five more good years to scramble for a new beginning. Considering what it has achieved in the past decade, a half-decade could be long enough.