DRAM Scam

If you want to see how "industrial policy" actually works in practice, take a look at Sematech. This semiconductor research and development consortium, funded jointly by private industry and the federal government, is probably the country's biggest foray yet into this new-style interventionism. While Sematech currently is seeking an extension of federal funding, it is also serving as the model for other initiatives.

On the industry side, Sematech's 14 member companies account for 80 percent of combined U.S. semiconductor sales. These companies contribute half of Sematech's roughly $240-million annual budget. Government participation consists of $100 million a year from the Defense Department's Defense Advanced Research Projects Agency (DARPA), as well as various state and local subsidies. The declared goal of this government-industry "partnership" is to develop new technologies that will help U.S. industry to regain (from Japan, of course) world "leadership" in this economically and militarily "strategic" industry.

From the standpoint of industrial-policy supporters, Sematech would appear to have all the makings of an ideal test case. "Sematech was an experiment, but also a good model," says Daniel F. Burton Jr., executive vice president of the Council on Competitiveness, an umbrella group sympathetic to a more active federal role in industrial policy.

In the first place, Sematech's beneficiary is not some declining smokestack industry; rather, the focus here is on the cutting edge of high technology. Thus, the usual knock against industrial policy—that it favors dying "sunset" industries at the expense of emerging "sunrise" ones—is seemingly inapplicable. Furthermore, the fact that Sematech is dedicated to "precompetitive" R&D apparently dispenses with the other stock argument against industrial policy—namely, that government should not be in the business of picking winners and losers.

And indeed, the basic idea behind Sematech—that government should extend its R&D spending into overtly commercial areas and should do so by working directly with private companies—has been gaining popularity. The Reagan and Bush administrations have generally opposed anything that looks like industrial policy. Due largely to the influence of presidential science adviser D. Allan Bromley, however, the White House has now signaled that it supports government funding for development of commercial technologies, so long as they are sufficiently "generic" to be considered "precompetitive."

One notable example of this kind of policy is the Advanced Technology Program (ATP), administered by the Commerce Department's Technology Administration. ATP, which gives grants to private companies conducting high-tech R&D, started with a 1990 budget of only $10 million; the budget rose to $36 million in 1991 and will increase $46 million in 1992. Meanwhile, the push is now underway to extend federal support for Sematech, which expires at the end of 1992, for another five years.

Sematech's actual track record, though, should serve as a warning rather than a blueprint. Cutting through the high-tech jargon and reassurances about "precompetitive" assistance, a close look at Sematech confirms all the darkest suspicions of industrial-policy critics. For as it turns out, even microelectronics has its sunset industries, and even precompetitive R&D has its winners and losers.

Semiconductors, or "chips," are electronic devices that store, retrieve, and process information. They provide the hardware "smarts" not only for computers but also for televisions, fax machines, telephones, microwave ovens, cameras, car ignitions, and antilock brakes—not to mention Patriot and cruise missiles and all the other dazzling high-tech weaponry on display in the Gulf War. Usually made from silicon, with microscopic aluminum wiring deposited on the surface in multiple layers, chips are miracles of miniaturization: A single chip the size of a postage stamp can contain millions of electronic components, with features on its surface measured in fractions of a micron (a human hair is about 75 microns thick).

Semiconductors may be divided into two basic categories: memory chips and logic chips. Memory chips, as their name implies, store and retrieve data. The biggest-selling of these is the DRAM (dynamic random access memory, pronounced dee-ram), which provides short-term data storage for computers. DRAMs are relatively simple to design but excruciatingly difficult to manufacture—at least at production yields high enough to make selling them commercially viable. Accordingly, the key to competitive success in DRAMs is high-volume, low-defect, low-cost production. DRAMs and other high-volume memory chips have been dubbed "commodity" chips to reflect their fungibility and fluctuating prices.

Logic chips, on the other hand, juggle and manipulate data. They make decisions, route information to different destinations, perform calculations, and relay instructions. The best-known logic chips are microprocessors, which act as central control centers in personal computers. Since logic chips perform complex functions and frequently have highly specialized or even customized applications, the premium is on design rather than raw manufacturing efficiency. Unlike memory chips, logic chips compete in the marketplace based on what they can do, not how much they cost.

The origin of Sematech goes back to the mid-1980s, when U.S. companies staged a wholesale evacuation from the DRAM business. Although DRAM technology was pioneered in the United States (Intel invented the DRAM in 1971), by the early 1980s American chipmakers were seeing their profits and market share slip away in the face of fierce Japanese competition.

A glut in worldwide DRAM capacity, combined with a sharp drop in demand, caused prices to plummet in 1985. Faced with huge and mounting losses (though not as large as the losses being suffered by Japanese producers), one U.S. chipmaker after another bailed out of DRAM production. By 1986, only two American-owned companies were left making DRAMs for sale, Micron and Texas Instruments. The American industry, which had virtually monopolized the world market only a decade before, now claimed less than a 10-percent share. The Japanese, on the other hand, had increased their market share to more than 80 percent.

(By the way, these statistics do not take into account the U.S. "captive" producers—namely IBM and AT&T—that produce chips for their own use rather than for sale on the open market. Both companies have continued to manufacture DRAMs, and IBM remains the world's largest producer.)

The loss of DRAMs sent the industry into a panic. Many regarded commodity memory chips as the key "technology driver" in semiconductor production. In this view, the unceasing race to cram more and more memory onto less and less silicon at lower and lower cost spurs the innovations that are needed to stay competitive in all areas of chip manufacturing. Thus, the state-of-the-art DRAM in the late 1970s contained 16,000 bits of memory; today companies are beginning to sell DRAMs with 16 million bits of memory—a thousandfold increase. Drop out of this race, it was thought, and competitiveness in other more specialized semiconductors would soon falter as well.

Having failed in the marketplace, the big U.S. chipmakers turned to Washington for help. They prevailed upon the U.S. government to bring antidumping cases against Japanese producers, accusing the Japanese of selling below cost. (Since both American and Japanese chipmakers were losing money in the mid-1980s, they were all selling below cost in a sense. This is all that's required to trigger antidumping tariffs.) The antidumping cases threatened Japanese chip imports with punitive duties as high as 108 percent. To avoid this outcome, the government of Japan struck a deal with the U.S. trade representative in July 1986.

The agreement imposed worldwide controls on Japanese semiconductors. It established price floors for sales to the United States and third countries and targeted 20 percent of the Japanese market for U.S. and other foreign suppliers. To implement the agreement, Japan's Ministry of International Trade and Industry leaned on Japanese chipmakers to reduce output and shipments of DRAMs. The result was an acute worldwide shortage of DRAMs during 1988 that raised prices, bestowing windfall profits on Japanese chip companies and inflicting serious harm on U.S. computer makers, computer buyers, and anyone else who needed DRAMs. (Last summer this agreement, in somewhat altered form, was extended for another five years.)

Sematech represented the next step in the Washington strategy. After using political means to restrain foreign competitors, industry leaders now campaigned for outright government assistance. In March 1987, 14 U.S. chipmakers announced the formation of a consortium to take on the Japanese in developing advanced manufacturing techniques. They also announced that they wanted government funding for the project.

Initial planning for the consortium, called Sematech for "semiconductor manufacturing technology," had envisioned actually manufacturing DRAMs for sale. This idea was scuttled, first because the remaining American DRAM producers didn't want to create a new competitor, but also because IBM was afraid it would get stuck buying Sematech chips if other purchasers could not be found (not exactly a ringing endorsement of the consortium's prospects). Accordingly, project planners settled on the more limited goal of cooperative R&D, which members could then use in their own manufacturing operations. Even this degree of collaboration marked a dramatic shift from the rugged entrepreneurship that had always characterized the American microelectronics industry.