#769 1987 · SEMATECH (14 U.S. semiconductor companies) · Semiconductor manufacturing / industrial R&D consortium
US chipmakers stopped competing on how to make chips and started competing only on what chips to make
the problem
No single chipmaker could justify funding shared manufacturing research its rivals would get for free
background
By the mid-1980s, U.S. semiconductor manufacturers were losing ground fast to Japanese competitors, whose government-backed industry partnerships had driven rapid, coordinated improvements in chip manufacturing. American chip companies still competed fiercely against each other on chip design and market share, but on the underlying manufacturing process — the equipment, materials and techniques used to actually fabricate chips reliably at scale — no single U.S. company was investing enough to keep pace.
The reason wasn't a lack of will; it was that manufacturing process improvements tended to spread across the whole industry's shared base of equipment suppliers and skilled workers regardless of who paid to discover them, so any single company funding this kind of research alone was effectively subsidizing its competitors along with itself. A 1987 Defense Science Board Task Force report made the case to Congress that this collective underinvestment was a genuine national competitiveness problem, not just an individual corporate failure.
what everyone would do
Let each chip company invest in its own manufacturing process research independently and compete on process improvements the same way they competed on product design — the standard model for corporate R&D, and one where a company that developed a better manufacturing technique either lost the advantage once rivals reverse-engineered it or simply couldn't afford, alone, to match the scale of investment Japanese manufacturers were making collectively.
what they saw
Manufacturing process research — how to reduce defects, shrink circuit widths, improve yields — wasn't actually a source of lasting competitive advantage the way product design was, because once a technique existed it tended to spread through the industry's shared supplier base and workforce regardless of who discovered it first. Since no company could keep the benefits of its own process research exclusive for long, no company had a strong individual reason to fund enough of it, and every company underinvested in exactly the research the whole industry needed most. The fix wasn't to make process research more proprietary — it was to admit it was already effectively shared, and formalize the sharing so companies could pool their funding instead of each underfunding it alone.
the move
In 1987, fourteen U.S. semiconductor companies — including IBM, Intel, Texas Instruments and Advanced Micro Devices — formed SEMATECH, a consortium that pooled member funding specifically for shared manufacturing process research, restricting results to paying members while member companies continued competing exactly as before on chip design and everything built using the shared process improvements. The Defense Department matched the consortium's industry funding roughly dollar for dollar, and SEMATECH began operations in 1988.
why it works
By restricting SEMATECH's research results and technical advances to paying member companies, the consortium ensured that only firms contributing to the shared research fund got early access to what it produced, removing the pure free-rider risk that discouraged any single company from funding this kind of research on its own. Matching industry funding roughly dollar for dollar with a Defense Department contribution gave the consortium real financial scale no single company, or even the whole industry acting independently, was putting toward this specific problem, while member companies kept competing exactly as fiercely as before on chip design and everything built with the shared manufacturing techniques.
the payoff
By funding manufacturing process research collectively rather than leaving each company to underinvest alone, SEMATECH gave U.S. chipmakers the coordinated research scale their Japanese competitors already had, while restricting benefits to paying members preserved each company's incentive to keep contributing rather than free-ride. U.S. semiconductor manufacturers, whose global market share had fallen sharply through the mid-1980s, regained competitive ground over the following decade.
where it breaks
It only works if the line between "pre-competitive" research worth sharing and genuinely competitive advantage worth guarding stays clear and enforceable — a consortium that drifts into sharing product design or strategy, not just manufacturing process, erodes the very competition it was meant to preserve. And it depends on members trusting each other enough to actually contribute real funding and real engineers rather than free-riding on the consortium itself, which requires enough shared existential threat — in SEMATECH's case a common foreign competitor — to keep rival companies cooperating instead of defecting.
what came after
SEMATECH's model — pool funding for "pre-competitive" shared research while keeping product design and everything built on top of it fully competitive — became a template widely cited in innovation policy for how an entire industry can fund research no single company can rationally fund alone, without dulling the competition that drives the industry forward everywhere else.
references
- [1]Lessons from SematechMIT Technology Review, 2011technologyreview.com
- [2]The Tunnel at the End of the Light: The Future of the U.S. Semiconductor IndustryIssues in Science and Technology (National Academies / Arizona State University), 2019issues.org