In the wake of tremendous shortages and soaring demand for the basic building blocks of the computer industry - silicon memory chips - the US semiconductor industry is tiptoeing back into a market it had all but abandoned to the Japanese. The world's fourth-largest semiconductor manufacturer, Motorola, this month announced it would return to making dynamic random-access memory (DRAM) chips after dropping out of that market in 1985.
Another big US chipmaker, National Semiconductor, announced recently it was considering reentering the field. A spokesman for the Semiconductor Industry Association (SIA) said other chipmakers were also weighing reentry into the DRAM market. And the only two US companies that still make and sell DRAMS say they will beef up production.
These tentative steps back into a world market still dominated by Japanese semiconductor companies (which captured 90 percent of the DRAM market in the past decade) are more than merely symbolic.
``We think there is some future for US DRAM manufacturers,'' says Victor de Dios, an industry analyst with Dataquest, a market research firm in San Jose, Calif. ``We strongly encourage [US firms] to reenter. We think conditions are right.''
A chipmakers' supermarket
With the current trend toward long-term supply agreements between computermakers and US semiconductor manufacturers, companies like Motorola and Intel are apparently seeing the need to have a full line of products.
That means participating, once again, in the DRAM market, the largest chunk of the $34 billion world computer chip business. DRAM sales were $2.4 billion (about 7 percent) of the world chip market last year, according to the SIA.
For big US chipmakers, not having DRAMs to sell computermakers is ``like having a grocery store without a meat counter,'' says Juan Benitez, president of Micron Technologies Inc. Micron last week announced a long-term agreement to supply much larger Intel with DRAM chips. Intel, along with a host of other semiconductor-makers, fled the highly volatile DRAM market to invest its energies in more profitable application-specific chips and microprocessors.
Micron and Dallas-based Texas Instruments Inc. are the only US companies selling chips to computermakers. International Business Machines and American Telephone & Telegraph have their own chip foundries to supply their needs.
``It wouldn't surprise me if US computer manufacturers started favoring US chip suppliers - as long as they get the same quality and the same price as they get from the Japanese,'' Mr. de Dios says. He and other analysts cite several reasons that a window of opportunity may exist for US companies:
The stronger yen, which has made the previous Japanese production cost advantage over US companies ``inconsequential.'' US-produced DRAMs have sometimes cost 30 percent more to make than those by Japanese companies.
Japan's Trade Ministry is reportedly trying to change the attitude of Japan's big chipmakers away from a market-share-driven concept to a profit-driven concept. This might discourage ``dumping'' chips below cost on the world market. It remains to be seen whether the effort to change attitudes will remain strong in a glutted market.
The SIA is trying to keep the political heat on the Japanese to open their large DRAM market to US companies. There was some encouragement as US makers' share rose to about a 10 percent level late last year. The rate of penetration has leveled off, however, and the SIA has said it is concerned.
Tighter relationships between US chipmakers and US computermakers may insulate DRAM-makers somewhat when the next glut of chips and consequent cutthroat pricing occur.
Since charging the Japanese with ``dumping'' DRAM chips, the United States has tried to level the playing field by forcing Japan to sell them at a reasonable profit instead of below cost. Dumping has been squelched, but a shortage of chips has brought higher prices.
``We've been scrambling,'' says Frederick Van Veen, vice-president of Boston-based Teradyne Inc. His company makes automatic test equipment that uses DRAMs. ``It has been tight, but so far it hasn't affected our shipments,'' he says.
But the main reason for the shortage is that both US and foreign chipmakers are shifting from 256K (256,000 bits) to 1 megabit (1 million bit) chips and 4 megabit production. Eight bits of memory equals one letter character of memory. Analysts predict the DRAM shortage will end sometime between the end of this year and the middle of 1989.
While many other types of silicon chips are used in computers for memory and computing functions, DRAMs are the most basic and widely used. Anywhere from one to hundreds of these fragile silicon wafers are used in every computer. Complex, yet highly repetitive patterns representing thousands of transistors are etched onto them.
In the early 1980s the 16K chip, representing 16,000 bits of memory, was the basic building block. Yearly advances packing the equivalent of thousands more transistors onto the same tiny space have made 256K the current DRAM standard.
At present, the inglorious DRAM has reached the commodity status of pork bellies, corn, or wheat - produced by the bushel and sold for anywhere from 30 cents to $70 each, depending on demand.
The shortage has resulted in a jump in price of 256K chips from about $1.80 last October to as much as $8 on the spot market. Prices of new 1 megabit chips have soared from $15 to $40 each.
A key `technology driver'
Analysts say another reason semiconductor companies may return to making DRAMs is that it may help them keep a slight edge over the Japanese in the more lucrative production of custom chips and microprocessors.
``No doubt that DRAMS are a big important market,'' says Daryl Hatano, director of international trade and government affairs at the Semiconductor Industry Association. ``But for US manufacturers it has an important strategic role as a technology driver.''
Because these chips must be manufactured in large volumes at very exacting tolerances, the DRAM is considered a key ``technology driver.'' Other chips can be technology drivers. But because the DRAM is of a relatively ``regular'' design, it is a bit easier to pinpoint defects in the manufacturing process.
Fine-tuning a production line to DRAM quality and volume gives a company the superior manufacturing expertise needed to produce lower-volume, more-sophisticated microprocessor chips, and chips for specific applications.
``These processes are very difficult to master,'' says Stanley Victor, a spokesman for Texas Instruments. ``They require exceptional discipline to manufacture. Making DRAMs drives manufacturing excellence, which, in turn, gives other products a great advantage.''
The cycle of shortage and glut
Despite the current profitability, chipmakers getting back in the DRAM business have to look to the long run. ``You don't make a decision like that based on market demand at any one time,'' says Ken Phillips, a spokesman for Motorola.
Mr. Phillips and others say the money-losing gluts, such as occurred in 1985 and 1986 with the 64K chip, may be less severe in the future. Agreements with the Japanese and closer links between chipmakers and computermakers may act as a buffer, or even defuse future chip gluts. This is because chipmakers are more able, these days, to trust that computermakers' orders for chips are not inflated.
That may be true, but there is a caveat to the current spate of optimism and profitability in this volatile industry.
G. Dan Hutcheson, president of VLSI Research Inc. in San Jose, sees a strong likelihood that past patterns of five-year feast-and-famine cycles will be repeated.
``It's a textbook case of pure competition,'' Mr. Hutcheson says. ``It is real easy to overbuild capacity.'' This, he says, is due to the long lag time between a corporate decision to build a plant and the 2 years it generally takes to get the plant up to speed. During that time shortages persist, prices are high, and more companies jump on the bandwagon building plants.
Nevertheless, excess capacity is not right around the corner,'' Hutcheson contends. Right now factories are working at 95 percent of capacity and are stretched to their limits. Still, only about 20 efficient plants are needed to supply the world with memory chips, and there are almost that many now. So adding many more plants could lead to a glut in 1 megabit chips, he says.
``Everyone always says we're not getting in memories again,'' he says. ``But they all begin building when things look promising.''