When the Chips Are Down
It's time for Samsung to exit the DRAM business
By ASSIF SHAMEEN
Wednesday, October 31, 2001
Web posted at 11:20 a.m. Hong Kong time, 11:20 p.m. GMT
Some months ago, I advocated in these columns that the South Korean government and creditor banks let Hynix Semiconductor, the memory-chip maker, die. Never mind that 15,000 jobs might be lost or the fact that Hynix's overseas sales in its heyday made up 5% of Korea's total exports. My proposal was fairly radical, and it earned me a few stinging e-mails. It doesn't give me any pleasure to be able to say that a lot of people have come round to thinking the same way. Sure, the events of Sept.11 have helped move them toward my viewpoint, but, believe me, Hynix has had trouble written all over it for years now. With $8 billion in debts, the company seems destined to join all those others that thought they had some magical formula to defy basic business logic. This week's news that creditors are prepared to bail out Hynix doesn't make me change my mind.
But it's not just Hynix. Memory-chip prices are now so low, oversupply is so great, and demand is so weak that almost every company that makes DRAMs might soon be in trouble. Gartner, the IT industry market-research firm, forecasts that global sales of DRAMs are likely to be down 67% this year to $10.5 billion from $31.5 billion and will fall another 19% next year. Other analysts are more pessimistic, with projections of over 70% down this year and up to 25% down next year. Five Japanese DRAM makers alone could lose up to $6 billion on DRAMs alone in their current fiscal year ending March 2002. That's a lot of red ink even for Japanese electronics giants.
Time was when memory chips were made by U.S. giants like Intel and Texas Instruments -- until it dawned on them that the commodity chip business wasn't anything like as rewarding as making microprocessors or customized chips. In the 1980s the business was dominated by the "fabulous five" Japanese giants -- NEC, Toshiba, Fujitsu, Mitsubishi Electric and Hitachi. In the late 1980s and early 1990s, Korean giants such as Samsung Electronics, as well as Hyundai and LG Semicon (who merged to form what is today's Hynix) became dominant players. Samsung is by far the world leader. In the mid-1990s, the Taiwanese jumped on the bandwagon and became large DRAM makers. Taiwanese names such as ProMos, Powerchip, Winbond, Mosel Vitalec, Nan Ya Technologies now account for some 27% of total global DRAM production. Since the mid-1990s, two Western companies have emerged -- Micron of the U.S. (No. 2) and Germany's Infineon (an affiliate of Siemens), now the 4th-largest DRAM maker and the fastest growing among the major players.
The semiconductor industry goes from feast to famine to feast. In the mid-90s, the big DRAM makers were raking in more than a billion dollars a quarter in operating profits from memory chips. What a difference a year or two can make. Two years ago, standard memory chips were selling for over $20 in the spot market. That was in the aftermath of Taiwan's September 1999 earthquake. Today, the price for the standard 128M DRAM in the spot market is around 90 cents. It costs between 3 and 4 bucks to make those things, so the more chips you make and sell, the more you lose. Some Japanese DRAM makers have drastically cut production and other memory-chip makers have shut production plants for days on end to keep a floor under the prices. But as the OPEC oil cartel knows, production cuts work -- but only to an extent.
Fujitsu has gotten out of the DRAM business. Toshiba, which is talking to Infineon about a joint venture, says it too will shut its DRAM plants if those talks fail. Hitachi and NEC have formed a joint venture dubbed Elpida, which now controls the DRAM plants of both those companies. But just as two wrongs don't make a right, a bigger merged company selling DRAMs at 90 cents a pop isn't going to lose any less money than two separate companies selling them for 90 cents a pop -- as long as production costs are closer to 4 bucks per chip. Elpida is reportedly looking to shut the old Hitachi DRAM plant in Singapore. Unless the shutters come down at Hynix before Christmas, the betting among industry insiders is that none of the Japanese electronics companies will be in the DRAM business after results for the current fiscal year ending March 2002 are announced next May.
With the Japanese out of the market, there would be just three big players -- Samsung, Infineon and Micron -- plus a bunch of Taiwanese, many of whom would probably set up new plants in China to keep making those cheap DRAMs you and I need for our PlayStation and laptops. We'd probably still need DRAM plants in Europe and America to supply manufacturers in those regions, so that means efficient producers like Micron and Infineon will probably survive and prosper.
That brings us to the big question: Does Samsung really need to be in the DRAM business? Isn't it wasting its time, money, and effort in a cyclical market where over the long run the risks are heavier than the rewards? Already Samsung is selling half its semiconductors and 80% of its DRAMs at below cost price. Only its high-end DDR (double-data-rate) DRAMs and Rambus DRAM modules are making money for it right now. True, Samsung is about to roll out 512 megabit DRAMs commercially and widen the gap between itself and the rest. And it is investing in state-of-the art 0.12 micron technology (as opposed to the current 0.15 micron), but the company would do well to learn from the mistakes of Toshiba, NEC and Hitachi.
Samsung should be concentrating on designing new products and building its brand, instead of messing around with DRAMs. Why doesn't it let Taiwanese foundries such as TSMC and UMC and the new Chinese operators look after the problems of etching the chips on 10 in or 5 in wafers? Its about time chairman Lee Kun Hee, who took Samsung from a bit player to the center stage of global electronics, let his men do what they do best, rather than bet the company's fortune and his reputation on commodity chips.
asiaweek.com |
