Imagine waking up to this paragraph in your favorite newspaper:
The willingness of the U.S. government to eschew partisanship and undertake a bold experiment -- an experiment based on cooperation as opposed to traditional procurement, and with accountability standards rooted in trust instead of elaborate regulations -- has led the U.S. to a position of preeminence in an industry which is vital to our nation's security and economic well-being.
You'd likely be incredulous, right? Private-public collaboration, as if! Bipartisan investment in a long-term project, sure! Redesigning out-of-date systems and ditching unnecessary regulations, in our dreams!
That is not a quote from a utopian novel.
I lifted that language from the Final Report to the Department of Defense on an initiative known as Sematech. In the late 1980s, the US recognized that it was in danger of missing a pivotal technological moment. Japan had seized a substantial advantage in semiconductor research and development. As of 1989, the US share of semiconductor equipment sales plunged from 69 percent in 1983 to just 51 percent in 1988. Over that same five-year span, Japan surged ahead--increasing its share from 25 percent to 40 percent. Rather than pursue a patchwork of state-driven approaches and in place of relying solely on private US firms to magically catch their Japanese rivals, a diverse set of stakeholders coordinated a national response.
Congress acknowledged that the status quo marked an "unacceptable situation" and formed SEMATECH--a private-public consortium of 14 private companies committed to putting the US back on the vanguard of this critical industry. In practice, this involved “developing, demonstrating and advancing the technology base for efficient, high yield manufacturing of advanced semiconductor devices.” The consortium embarked on research covering everything from lithography to plasma etch and deposition equipment. The idea was that this kind of basic research would lead to the US reclaiming its status as the market leader.
Such an intensive, ambitious, and cooperative research agenda would not have happened without a willingness for public and private participants to invest—literally and figuratively—in a more competitive, productive America rather than attempt to regulate our way back to the top. The deal required all stakeholders to put significant skin in the game. More specifically, each company pledged to pay one percent of their semiconductor sales revenue per year, albeit with a floor of $1 million and a ceiling at $15 million. The federal government threw in a cool $100 million of matching funds to help things along.
But the buy-in from the private side was far more than financial--as Chris Hughes explains in his book, “Marketcrafters.” Companies sent some of their top talent to Austin, home to Sematech’s state-of-the-art fab facility (which was built in just 32 weeks). Many of those brilliant engineers stayed for two years--demonstrating the importance of the cause as well as the esprit de corps that SEMATECH’s leaders managed to cultivate.
SEMATECH firms also agreed that the resulting research would be shared and commercialized, which meant that even non-members could experience downstream benefits from the project’s advances. Intel later reported that its cumulative $17 million in SEMATECH saved the firm $300 million via production efficiencies. On the whole, economists suspect that the return on investment was somewhere between $1.40 and $2.80.
We’re in dire need of a SEMATECH for AI. More than three years into the current generative AI revolution, there’s a shortage of basic AI research -- studying new AI paradigms beyond large language models, developing evaluations to test model capabilities, and generally advancing the AI frontier so that we can keep pace with geopolitical rivals and realize the technology’s potential to increase human flourishing.
It’s no secret that some of America’s greatest scientific and technological advances have come about through getting a bunch of really smart folks together in one place and charging them with an ambitious mission that’s personally meaningful, politically important, and, yes, lucrative. A SEMATECH for AI would reduce duplicative work that’s going on at the labs, across universities, and within nonprofits.
Two key lessons should be gleaned from the SEMATECH example: first, government should not pick winners or micromanage innovation; and, second, the state can sometimes do the most good by clearing space for voluntary coordination, aligning incentives, and then getting out of the way. A modern AI consortium should be narrow in scope, time-limited, and disciplined by real private capital—not a new regulatory superstructure, not a permanent bureaucracy, and not a vehicle for industrial policy. Its charge would be simple: fund pre-competitive research that the market undersupplies, share results widely, and sunset once its mission is complete. That approach respects what folks such as Nobel Laureate Joel Mokyr have long argued—that durable technological leadership comes from trust, experimentation, and decentralized problem-solving, not compliance checklists or precautionary bans.
Kevin Frazier is a Senior Fellow at the Abundance Institute, directs the AI Innovation and Law Program at the University of Texas School of Law, and is an Affiliated Research Fellow at the Cato Institute.
Trump’s ‘Just for Fun’ War Talk Shows a Dangerous Trivialization