Sematech
The 1987 government-industry consortium credited with helping America win back the chip industry from Japan, and the template every chip subsidy since has invoked.
In the mid-1980s the United States was losing the industry it had invented. Japanese firms had taken the memory chip market with better yields and cheaper capital, American memory makers were exiting one by one, and the Pentagon was contemplating a future in which its weapons depended on foreign silicon. The response, in 1987, was Sematech: fourteen American chip companies, accounting for the bulk of the country's manufacturing capacity, agreeing to pool research with matching money from the Defense Advanced Research Projects Agency, roughly a hundred million dollars a year from each side.
Fixing the machines, not the chips
Its first instinct was wrong. Sematech began by trying to build a shared demonstration fab in Austin that would show American firms how to manufacture, and its members, fierce competitors, were not eager to share much of anything. Under Robert Noyce, the co-inventor of the integrated circuit who left Intel to run it and died in the job in 1990, the consortium changed course toward the part of the problem nobody owned: equipment.
American chipmakers were buying tools from American suppliers who were smaller, worse funded, and increasingly behind their Japanese rivals. Sematech redirected roughly half its budget into those suppliers, sending members' engineers to work on their tools, funding development, qualifying equipment jointly so that no single company bore the cost of proving a new machine. It also standardized how tools were tested and how fabs measured themselves, and published the industry roadmap that told every supplier what would be needed and when.
Did it work?
American chipmakers regained the lead in the 1990s, and Sematech is routinely credited with it. Economists are more careful: the recovery owed a great deal to Intel's pivot from memory to microprocessors, to the personal computer boom, to the collapse of Japan's asset bubble and the capital that had funded its fabs, and to the 1986 semiconductor trade agreement with Japan. What can be defended is narrower and still useful. Sematech measurably strengthened the American equipment sector, cut duplicated research, and created a place where rivals could agree on standards and a technology roadmap. Federal funding ended in the mid-1990s, when the members decided they could pay for it themselves, and the consortium opened to foreign firms in 1998, which was both an admission that the problem was no longer national and the beginning of its loss of purpose. Its later years were spent on problems it had once dismissed: a Sematech lithography task force ranked extreme ultraviolet last of four candidate technologies in 1997, and after the Intel-led consortium that proved EUV wound down in 2003, Sematech took over the unglamorous work of making defect-free mask blanks for it, eventually at a dedicated center in Albany. It moved there in 2007 and was absorbed into the State University of New York system in 2015.
Sematech is now cited far more often than it is examined. Every recent argument for industrial policy in chips, the CHIPS Act's National Semiconductor Technology Center in the United States, Europe's chip plans, China's national funds, reaches for it as the proof that governments can rebuild an industry. The record supports something more modest: consortia are good at fixing shared bottlenecks that no single firm will pay to fix, and are not a substitute for a company deciding to make a better product.