We have been speaking with a number of policy people in recent months on a whole range of semiconductor topics. In preparation for that work, we solicited friends and contacts for good ideas that the US should implement to drive the industry, and the US economy more broadly. By far, the most compelling idea we heard was the establishment of a National R&D and Prototyping Center. To be clear, this is not our idea. we will keep the author’s name out of this discussion, but he is someone who knows how to run a fab.
From what we can tell, the policy organizations in Washington are developing a muscle that has atrophied in the US over the past fifty years, namely how to craft a coordinated Industrial Policy. There is considerable debate as to whether the US can or should engage in this kind of planning. Setting aside purely ideological objections (please), a big concern with such efforts is they risk creating perfect versions of old technology. The USSR spent much of the 20th century building a formidable 19th century economy – trains! hydro-power! coal mines!- to name just one example. We think there is some risk of that happening in the US, or at least a lot of subsidy dollars going to older, established technologies. For us, a much more compelling idea is to focus on all the things that come next, aka innovation. The US government is not going to provide that – except of course for the Internet, jet engines, rockets, satellites, etc. Or more precisely, the US government cannot and should not fund all of the innovation. But it can take some fairly easy and cheap (only single digit billions) steps to greatly facilitate innovation which others can extend and ultimately commercialize.
There are several approaches for this in semiconductors. When discussing this topic, the knee jerk response is “quantum”, but that field probably has enough commercial backing for the moment. Instead, we think the focus should be on new materials and new packaging technologies. A big problem any company, start-up or established, encounters in this space is the significant upfront costs of just proving out new technology. Getting capacity from a foundry for small batches is expensive and highly restricted by capacity. And for start-ups, a huge portion of seed capital has to flow right out the door to IP licenses and manufacturing partners, throttling the ability to actually work on new technology. We have written (a lot) about US VC’s reluctance to invest in hardware start-ups, and much of that reluctance hinges on these upfront costs.
So we propose a National R&D and Prototyping Center. This would be a fab from science fiction – kitted out with all the latest machines from the leading equipment makers. Not just EUV machines to build leading edge standard silicon processes, it would have the capacity to manufacture on a whole range of materials – SOI, GaN, SiC, GaAs, III/V, CMOS and all the rest. It would also have advanced packaging capacities, including photonics. This is feasible because the goal of this plant is not to produce commercial volumes – small batches only. This greatly reduces the amount of equipment required. The plant would be optimized to run these small batches, something that is possible when no one cares too much about the accounting costs of depreciation.
Now to be fair, something similar to this already exists at SUNY Albany, which has a joint research facility with IBM. The National Center would be different in that it would build a far larger network of partners. This would include the major foundries, the IP and EDA providers, big fabless companies, and anyone else who wants to participate. The benefit for these companies would be early access to all the participants, who would enter the facility too small for the majors to work with but exit as major new customers. Venture investors would be presented with a stable of start-ups whose technology has been de-risked and now only needs capital to grow. Start-ups would be given licenses to play around with software, IP libraries, early integration with the foundries’ PDK tools and the ability to rapidly iterate on prototypes. We think this could be incredibly powerful, allowing small companies to explore, develop and prove out new technology.
This does not have to be wildly expensive. Again, we are not proposing a giga-fab producing 100,000s of wafers a month. We estimate the whole facility would cost ‘only’ $5 billion to build, and would probably cost $1 billion a year to operate, a bargain at semis-scale. Critically, the government would need to commit to funding the center for ten years, to assuage long-term concerns about viability. In theory, the facility could offset these costs by allocating capacity to large companies’ research needs, accepting sponsorship, or some other mechanism, although we would argue the benefits to the broader economy so greatly outweigh these costs, the facility could forego revenue entirely.
That being said, the most critical piece of this plan would be the governance model. We would argue strongly that the entity has to be run in a public-private partnership, free of political management. This could be achieved with a strong Board comprised of government employees (please not political appointees), CTOs from large corporates, academics, and a few professional venture investors. The board would establish a public selection process with rigorous standards, via an application not designed by a government employee. Ideally, the goal would be to let in everyone who qualifies, and the application would exist only to filter out the unscrupulous and un-serious. Further, the government would make no claim on any participant’s IP, although maybe the government could accept small equity stakes in participants.
It is hard to overstate how powerful this could be. The source of so much of the US’s success has rested on its ability to bring immense technological innovation. But that process has had its share of disappointments, and many of those disappointments rests on situations where US researches invent something new, but established companies and investors overlook it, allowing others to reap the benefits (looking at you TVs). A center like the one we describe here would bridge that gap, allowing inventors to prove out their innovation here in the US. The truth is that the big companies are at best moving glacially when it comes to truly new technology, they have too many more immediate concerns (like competing and making payroll). Someone has to step in to make innovation easier. Put simply, big companies do not seem to care much about innovation and small companies cannot afford to make dreams real. A coherent government policy could change all that.