Discover more from ChinaTalk
CHIPS Act: The Battle for R&D Dollars
Where will the shovels hit the ground for the National Semiconductor Technology Center?
Yesterday on ChinaTalk we got into some of the tensions facing the manufacturing incentive side of the Chips Act. We’ll be exploring the $11bn going to the R&D side, which may in the long term end up being even more important to the future of American semiconductor competitiveness than the manufacturing incentives.
Our guest author today Arrian Ebrahimi, who recently launched the Chip Capitols newsletter, explores how decisions over the structure and geography of the NSTC will determine the trajectory of America’s next era of scientific research. He gets into:
How geography is destiny for R&D initiatives: a centralized structure likely biases towards basic research, while a more distributed approach will bend more towards applied work
What lessons the US can learn from German, Taiwanese, and even past US attempts to pour billions into semiconductor R&D
For officials at the Commerce Department, deciding where shovels ought to hit the ground for the National Semiconductor Technology Center (NSTC) is high on the to-do list. The CHIPS and Science Act of 2022 funded this and other microelectronics research programs to the tune of $11 billion over five years, but decisions made over the next few months are critical. The administration will either lay a firm foundation for the next American technology dynamo or doom the investment to be little more than upgrades at existing research facilities.
Whereas full centralization would concentrate all NSTC facilities at a single location, a fully distributed NSTC would be comprised of annexes to existing facilities across the country. Hybrid models lie at varying points on this spectrum, usually with a central hub and many differently located regional centers. This question is not only critical to how the Commerce Department will spend $11 billion of taxpayer dollars, but it is also the battlefield for a fierce, and enduring, debate in U.S. science and technology circles: Should the government fund basic or applied research? If the latter, who sets the agenda?
The NSTC is not the world’s first public-private research consortium. Governments around the world, including past American administrations, have launched such initiatives to both great results and instructive failures. As officials finalize the NSTC’s blueprint, they should look to the starkly different organizational choices of research consortia in Germany, Taiwan, and previous U.S. efforts as guides.
Geography is key. Decisions about where and how to locate NSTC facilities relative to each other will indicate the balance of power in setting the U.S. microelectronics research agenda.
A Research Consortium is Born
In August 2022, Congress began a new era of American industrial policy with its passage of the CHIPS and Science Act. In addition to $39 billion in manufacturing grants for the semiconductor industry, the law included $11 billion in funding over 5 years for new public-private research infrastructure, primarily for establishing a National Semiconductor Technology Center.
The vision of the NSTC was first laid out, however, in January 2021 in the annual National Defense Authorization Act (NDAA), and Commerce Department officials have been planning for its implementation ever since. According to the bill, the NSTC shall “conduct research and prototyping of advanced semiconductor technology to strengthen the economic competitiveness and security of the domestic supply chain” and will emphasize strengthening the U.S. capabilities in: (1) advanced test, assembly, and packaging; (2) materials; (3) semiconductor equipment; and (iv) metrology for security and supply chain verification.
This early statute was vague by design, leaving questions ranging from the NSTC’s geographic organization to its coordination with other grant programs up to government officials and private sector participants to decide.
One Thinks from Where One Sits
In theory, a program like the NSTC could be fully centralized, fully distributed, or follow a hybrid model. The Semiconductor Industry Association (SIA) notes that the full centralization of NSTC facilities would offer the most cost-efficient approach because it minimizes the need for redundant capacity at separate facilities. At the other extreme, full distribution would allow different centers to specialize in specific semiconductor technology verticals (i.e. focusing on logic, analog, or memory chips).
In reality, most stakeholders suggest a hybrid, wheel-and-spokes approach because existing academic and corporate facilities already have parts of the infrastructure needed for the NSTC. In these models, there is a core NSTC facility and several satellite “Centers of Excellence” (CoEs) serving specific functions. What is often lost in this conversation, however, is that dividing and co-locating the NSTC among existing institutions inevitably impacts its research agenda, workforce, and business model. The following subsection examines how different stakeholders have defined “hybrid,” and what this says about their goals for the NSTC’s research agenda.
A big wheel directing small spokes
Academics, federal labs, and the most R&D intensive semiconductor companies (namely, fabless design firms) advocate the most centralized version of the “hybrid” NSTC model. These groups are concerned that siloed R&D at one stage of the technology “stack” could cause incompatibility with advancements further up or down. They want to see “full-stack” innovation that coordinates R&D throughout all stages of chip development. (A prime example is heterogeneous chiplet integration, which is blurring the traditional boundaries between substrate design and packaging.)
Evolving semiconductor technology and supply chains at such a fundamental level requires early orchestration at the research and development stage. To that end, centralization advocates call for the NSTC to be led by a strong core with full-scale, end-to-end advanced fabrication capabilities. For cost efficiency, this hub would be built primarily on an existing advanced 300mm microelectronics research facility, with centralization advocates specifically recommending Albany NanoTech in New York. Home to the State University of New York (SUNY) Polytechnic Institute's College of Nanoscale Science and Engineering, such a hub would be Albany Nanotech’s second experience hosting a U.S. government funded public-private partnership for semiconductor research.
The spokes in this model would be the Centers of Excellence, which would play a more secondary role than in more distributed models. Centralization advocates envision CoEs organized by ecosystem stakeholders. These include (1) researchers focused on developing new technology and workforce, (2) providers of equipment or materials integrating new technology into commercial offers, (3) users collaborating in assessing semiconductor products and their use in downstream commercial offerings, and (4) manufacturers to scale new chip technology into commercial production. The CoEs would not be pursuing independent technology verticals but rather researching different stages of a planned technology goal.
Large spokes cranking a small wheel
Industry stakeholders, namely semiconductor manufacturing companies and their affiliated organizations, advocate a very different approach. One prominent response to a Commerce Department request for information on the NSTC says that a single NSTC location with full-flow capabilities is not necessary. Rather, the NSTC could function as a “distributed full flow” connecting nationwide hubs. Regarding Albany NanoTech, this paper says the New York location could house “portions of NSTC’s technical agenda,” while also noting that leading U.S. manufacturers could house NSTC annexes to their existing or future facilities.
Semiconductor manufacturers have a clear preference for decentralized hubs pursuing specific technology verticals. Intel has proposed setting up an Advanced Packaging Lithography Center and assisting with an Advanced Packaging Manufacturing Center, both of which would operate as standalone physical facilities tied to the broader NSTC ecosystem and agenda. Micron has urged the NSTC to create a Memory Center of Excellence built adjacent to existing leading-edge facilities, namely Micron’s own in Boise, Idaho. The manufacturers’ philosophy is captured by TSMC’s position, which calls for the NSTC’s Centers of Excellence to feature both specific focus areas and economically self-sustainable business models. Having largely self-sustainable business models would empower each center to have a different level of engagement with public and private sector partners to meet R&D needs. In other words, such a vision would empower CoEs to operate more or less independently of each other and the overall NSTC.
Lessons from History and Abroad
This is not the first time the U.S. or other governments have set out to build public-private research consortia. Since the NSTC was authorized, U.S. policymakers, companies, and academia have wisely looked to the successes and failures of America’s last semiconductor research consortium—SEMATECH. Unfortunately, there has been scant public analysis of how other countries’ research consortia tackled the same questions facing the NSTC today.
This section looks at what effect centralization or distribution had on research consortia in Germany, in Taiwan, and in America’s SEMATECH effort. Many other case studies exist (as indicated in the chart above), but these three provide most effectively illustrate the centralization versus distribution discussion.
While not limited to semiconductors, Germany’s Fraunhofer program is perhaps the best example of the distributed NSTC Centers of Excellence approach favored by industry advocates. Billed as an “application-oriented” research organization, Fraunhofer was founded in 1949 and currently operates 76 facilities across Germany. Each facility specializes in a specific technology or sector, ranging from automotive manufacturing and renewable energy to microelectronics and information technology.
This distributed model has powerful network effects. Each facility is responsible for establishing innovation clusters around Germany, usually involving local universities, small businesses, and large conglomerates. Critically, while the facilities pursue their individual technology verticals, Fraunhofer as a whole works to link facilities with each other and with other public research institutions. Even the most distributed proposals for the NSTC will follow a more coordinated agenda than Fraunhofer, but the German model shows that proactive networking is key for disparate parts to contribute to a broader national research ecosystem.
The biggest challenge with the Fraunhofer model echoes a central critique of applied research policy, namely, focusing public money on predefined verticals drives predictable but slow improvements to existing industries. Most industry and academia advocates agree that the purpose of the NSTC is to drive generational leaps forward, not to simply augment existing private-sector research efforts. Though Fraunhofer has undoubtedly driven advancement in a range of German industries, it has not birthed the technology underpinning entirely new sectors. Policymakers in the U.S. will need to consider whether distributed Centers of Excellence may relegate the NSTC to the same fate.
Taiwan: Industrial Technology Research Institute
Founded in 1973 by Taiwan’s Ministry of Economic Affairs, the Industrial Technology Research Institute (ITRI) took a sharply different approach than Fraunhofer. Most of its operations are concentrated at its headquarters in Hsinchu, with a significant satellite location, ITRI South, in Tainan. In previous interviews, ITRI managers expressed concern about spreading research locations too broadly, as geographic distribution could undermine the benefit accruing from dense innovation clusters.
The ITRI innovation cluster has had a remarkable impact on Taiwan’s semiconductor innovation productivity. According to local government estimates, firms within the Hsinchu technology cluster generate an average value-added of 50 percent, compared to 30 percent for Taiwanese manufacturers located outside the cluster. In addition to the workforce and knowledge sharing advantages that come with such clusters, participants in the ITRI ecosystem also compete against other companies, and underperforming suppliers, researchers, and customers can be easily replaced by local competitors. In contrast, research centers that are widely distributed tend to anchor themselves to local champions that face less direct competition from distant alternatives.
Critically, investing a larger share of NSTC resources in a single location raises the stakes of site selection. To imitate ITRI’s success, policymakers should consider how the presence of universities, suppliers, supportive local governments and other factors in any prospective site would contribute to creating an innovation ecosystem around the NSTC hub.
United States: SEMATECH (SEMiconductor MAnufacturing TECHnology)
First established in 1987, the SEMATECH consortium received funding from the U.S. government to increase American semiconductor competitiveness against Japan. It operated an independent research facility in Austin, Texas through partnerships with thirty-one universities, American semiconductor companies, the Semiconductor Industry Association, and others. In 2003, the consortium began a partnership with Albany Nanotech at SUNY Poly. However, by 2010 financial constraints forced it to shut down its Austin facility to fully relocate to Albany, using facilities provided by the university. In 2015, SEMATECH was formally absorbed into SUNY and ceased to operate independently.
SEMATECH today counts a mixed legacy. The consortium’s initial success can be attributed at least in part to its original organizational structure. By owning and operating its own large R&D facility in Austin, SEMATECH could drive its long-term agenda without a single member company or university host towering over it. Industry and academia researchers could collaborate on more or less equal footing, and all members had reasonable access to SEMATECH’s facilities. An NSTC with a strong central hub would similarly be governed by a balanced board of industry, academia, and government representatives that could more effectively pursue long-term, rather than parochial, goals.
It is also worth noting that SEMATECH’s identity as a standalone institution gave its research staff a sense of loyalty to the consortium’s mission. Although these engineers were mostly on assignment from member companies, they clearly understood upon walking into SEMATECH’s doors that this project was distinct from their home companies’ typical work.
After its move to Albany, SEMATECH lost the benefit of its independent facility in Austin. Reliance on Albany Nanotech facilities established an uneven power relationship with SUNY, where SEMATECH had less and less control over its finances, and therefore its research agenda, until it was fully absorbed into SUNY Poly in 2015. As it is nearly certain that any manifestation of the NSTC will have annexes at existing company and university facilities, policymakers should make sure these NSTC facilities have business models or funding structures that can maintain their independent research agenda and avoid becoming arms of their host institution.
Geography is a Proxy Battle
As companies, universities, and federal labs clamor for their preferred vision of the NSTC, one should not lose sight of the wider shift in American science policy. Funding for the CHIPS Act floundered for nearly a year due to divergent approaches to science policy between the House and Senate.
While both chambers supported passing a bill to bolster American technology and its competitiveness against China, House Science Committee leaders sought to continue the U.S.’s “traditional” approach of supporting basic research in physics, chemistry, and other disciplines. In contrast, the Senate Commerce Committee and Majority Leader Chuck Schumer advocated a stronger federal focus on applied research. The National Science Foundation (NSF) became the primary battleground for this debate, with the Senate bill proposing several new technology directorates to focus on “use-inspired and translational research” for key technologies. Though the Senate triumphed in this legislative battle between basic versus applied research, the debate is far from over. Decisions on a centralized or decentralized NSTC are the administrative front of this ongoing debate.
A more centralized NSTC would represent a victory for advocates of basic research. There are two common paths for obscure scientific theories without clear paths to commercialization to get funding: basic research grants and insertion into long-term moonshot efforts. Certainly, any manifestation of the NSTC will be a far cry from the NSF’s model of awarding grants to an uncoordinated portfolio of basic research projects. However, if this research consortium has a strong central plan that orchestrates full-stack shifts in how chips are designed, manufactured, and packaged, it would provide more opportunities for the “long-lookers” –academics and non-commercial labs– to implement their risky, paradigm-shifting ideas as part of NSTC moonshots.
In contrast, a more distributed NSTC would stand as a victory for advocates of applied research. Dividing the NSTC into Centers of Excellence with unique technology verticals would allow the U.S. government and industry to spur American technology leadership in areas they identify as critical. Similar to the NSF’s new technology directorates, this distributed route gives policymakers more levers to pull in steering the ship of U.S. technological development. To the extent that private companies will benefit from these incentive levers, they will be happy to implement a national security research agenda. It also goes without saying that companies will have outsized influence over the agenda of NSTC annexes collocated with their own facilities.
This article does not advocate for a more centralized NSTC versus one focusing on separate applied technology verticals. Policymakers may justifiably consider advancements in predetermined areas of economic and national security interest to be more urgent than full-stack changes to how the chip industry operates.
As officials from Commerce, Defense, OSTP, and other offices finalize plans for the NSTC’s launch in the first half of this year, however, they would do well to consider the impact of geography on the fundamental nature of this endeavor. History teaches that early decisions regarding centralization versus distribution impact not only a consortium’s research agenda, but in some cases, its viability as a distinct public policy endeavor.
ChinaTalk is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.