The CHIPS Act, Microelectronics, and the 2021 NDAA
With an eye toward China, overdue help for the US semiconductor industry arrives
Welcome to Semi-Literate, a guide to the chip industry through the lens of public policy. After the previous post’s focus on China’s efforts to promote its domestic microelectronics industry, this post looks at US efforts. Feel free to forward to others you think may be interested.
Note: Several are receiving this email unsolicited given assumed interest in the content. If this assumption was mistaken, apologies. You can unsubscribe below.
BLUF: The CHIPS Act, passed as part of the 2021 NDAA, is the largest effort to coordinate US government support for the microelectronics industry since SEMATECH was established in the late 1980s. The CHIPS Act focuses on incentives to build or modernize fabs so that more chips, and more advanced chips, are made in the US. This focus on fabrication is appropriate, but it comes with trade offs: the Act underemphasizes innovation, calls for redundant reports, lacks detailed technical roadmaps, and mostly ignores workforce issues facing the industry. It also overburdens the Department of Commerce & undervalues the Department of Energy.
Introduction
Tucked in to the 4,517 page 2021 National Defense Authorization Act (NDAA) passed by Congress over a presidential veto on January 1, 2021 were the most important 46 pages of legislation for the US semiconductor industry in decades. The Creating Helpful Incentives to Produce Semiconductors (CHIPS) Act is now law. The CHIPS Act, which incorporates provisions of the American Foundries Act, is the largest coordinated effort by the US government to support the microelectronics industry since standing up SEMATECH in 1987. However the NDAA is an authorization bill (which directs how federal funds should or should not be used) not an appropriations bill, so final funding amounts will be decided by Congressional appropriators in 2021. They will likely mirror amounts in the original June 2020 version of the CHIPS Act.
The Problem
This legislation was drafted, much like FIRRMA, in response to a problem: how does the US respond to massively-activist Chinese industrial policies designed to dominate the technologies that will confer the greatest economic and national security benefit in the coming decades? The US government is concerned with China’s efforts to promote its microelectronics industry and the effect these efforts are having on the US industry. The US strategic response to “protect” and “promote” these technologies calls for carrots and sticks. FIRRMA was a stick. The CHIPS Act is a carrot.
The Solution
The top-level solutions proposed in the legislation are straightforward:
The US needs to make more chips in fabs physically located in the US.
The DOD needs more secure, advanced, and custom Made in the USA chips.
The US should be more activist & strategic in support for microelectronics R&D.
US allies need to share its thinking about the importance of microelectronics.
(There are no solutions, only trade-offs)
The basic tension in this legislation is how much to focus incentives on commercial priorities (profit, volume, economies of scale, innovation) vs. defense priorities (customization, reliability, trust, integrity) with respect to the microelectronics industry. US microelectronics firms face this tension all the time: It is much more profitable for a fab to produce 74 million of the exact same chip for an iPhone (high volume, low mix) than 600 chips for F-35s or 140 chips for THAAD air defense systems (high mix, low volume). US companies do microelectronics work that is purely commercial (ex. Intel, AMD), a mix of commercial and defense (ex. GlobalFoundries, CREE), and mainly defense (ex. BAE, SkyWater). This Act tried to balance incentives for both commercial and defense, yet was criticized by defense firms as too-focused on incentives for commercial chip firms. There are no solutions, only trade-offs.
Notable Provisions in the CHIPS Act
Sec. 9902. Semiconductor incentives.
These incentives are aimed squarely at commercial chip firms. The Secretary of Commerce will coordinate federal financial assistance to incentivize investment in facilities and equipment in the United States for semiconductor fabrication, assembly, testing, advanced packaging, and/or research & development. Incentives (grants) are capped at $3 billion per recipient and can only be used to finance the construction, expansion, or modernization of a facility or equipment to be used for semiconductors. There is a clawback provision if deliverables are not met on time (*cough* Foxconn), a prohibition on recipients engaging in a joint venture/licensing effort with a foreign entity of concern (*cough* AMD), and mandatory GAO audits of the program.
Sec. 9903. Department of Defense (DOD).
The DOD-related provisions in the NDAA amount to broad reforms to the DOD Trusted Foundry Program (managed by DMEA). This is good. It’s been clear for a while that the ~15 or so commercial fabs that provide trusted fabrication services to the DOD are not capable of meeting the full spectrum of DOD chip requirements.
The CHIPS Act tasks DOD with establishing a public-private partnership (PPP) to incentivize formation of consortia of companies “to ensure the development and production of measurably secure microelectronics [emphasis added],” with a focus on radio frequency, mixed signal, radiation tolerant, and radiation hardened microelectronics. Incentives can include, but are not limited to, grants under Section 9902. All participants in said consortia need to meet the standards of “trusted microelectronics” as established in the 2020 NDAA (see Section 224).
DOD is also responsible for standing up a National Network for Microelectronics Development to commercialize emerging innovations, encourage the participation of “nontraditional defense contractors [startups],” and provide a bunch of reports:
a plan to maintain production of modern and legacy DOD chips (start here)
a report written with the National Academies of Science, Engineering, and Medicine on “optimal” PPP structure by October 1, 2022; (copy this one)
a report detailing how DOD is going to accomplish all these requirements; and
biennial DOD Comptroller reports on how things are going financially.
Sec. 9904. Department of Commerce study on status of microelectronics technologies in the United States industrial base.
Commerce is to begin a survey of the US microelectronics industrial base. This survey and report will be done by its Bureau of Industry and Security (BIS), using Defense Production Act (DPA) authorities. BIS can compel mandatory survey responses under the DPA and issue fines to those who do not respond. It’s a “who, what, where, when, how, why” sort-of-survey. One novel angle is requiring survey recipients to disclose if they’ve gotten similar information requests from the Chinese government and if they have any commercial relationship with the Chinese military.
Sec. 9905. Funding for development and adoption of measurably secure semiconductors and measurably secure semiconductors supply chains.
A “Multilateral Semiconductors Security Fund” will be established by the Department of Treasury and administered by State to develop/promote secure semiconductor supply chains with allies. The Congressional intent here is that State use this money, in concert with US allies in Europe and Japan, to de-risk semiconductor supply chains away from China. There is also language about promoting shared investment screening and export control standards among allies with respect to microelectronics.
Sec. 9906. Advanced microelectronics research and development.
The National Science and Technology Council (part of the White House Office of Science and Technology Policy) will write a national strategy on microelectronics
The Secretary of Commerce will establish:
an industry technical advisory group for microelectronics (a BIS TAC);
a National Semiconductor Technology Center, which will function as a PPP, to promote R&D in to semiconductor manufacturing equipment;
a National Advanced Packaging Manufacturing Program;
microelectronics research & ManufacturingUSA programs at NIST.
Sec. 9908. Defense Production Act of 1950 efforts.
The president will submit a report to Congress detailing how DPA Title III provisions (which give the President authority to ensure availability of domestic industrial resources essential to national security through economic incentives) may be used to support microelectronics production in the US.
Trade Offs
The aforementioned provisions are all good and the CHIPS Act is a commendable effort. With that said, below are some critiques and suggestions.
Fabrication > Innovation:
The way many of these incentives are structured belies a tacit emphasis on support for large existing semiconductor firms (ex. Intel, GlobalFoundries, AMAT) pursuing traditional metrics of innovation (chasing Moore’s Law). In general, the biggest winners are going to be companies that build fabs and the equipment that goes in fabs. Thats good - hardware is expensive. But there should be more of a focus on incentives for fabless companies doing innovative work as well as companies re-thinking how to make chips in general: ex. photonics, EDA vendors not named Cadence, Mentor, or Synopsys, open-source initiatives like RISC-V, and figuring out how to make lithography equipment in the US that can compete with ASML and TEL.
Lack of Detail:
The CHIPS Act provisions do not seem to incorporate specific recommendations of the many reports written about this industry in the last five years. To take just one example, the PCAST report put out at the end of the Obama Administration had detailed recommendations (see sections 3 & 4) improving: workforce, pre-competitive research, fab permitting processes, new approaches to integrated circuit design, and moonshot challenges. Industry has provided even more detailed recommendations.
The Act also missed the opportunity to revive big-picture ideas like the Global 450 Consortium. Leading edge chip manufacturing transitioned from 200mm to 300mm wafers in the mid 2000s. This allowed for more chips per wafer, enabling economies of scale that resulted in cost reductions greater than 20% per unit area. Yet there was no call out to incentivize pilot production at a US fab of 450mm wafers in the CHIPS Act. If not in this once-per-generation legislation, when? Maybe this will all be done by an implementing agency, but Congress could’ve just been explicit.
Analysis Paralysis: Too Many Reports
The Act calls for myriad reports which will do nothing to directly help the industry, little to inform the government, and are certain to arrive at redundant conclusions. There are plenty of good reports on the state of the US industry, Defense Department microelectronics needs, and at the strategic level. Yet Commerce, the Executive Office of the President, GAO, the NSTC, State, and DOD are all on the hook for reports.
In particular, Section 9904, which requires a Commerce Department industrial base survey/report, is duplicative and unimaginative. There are links to ten-plus chip industry reports sponsored by the US government and published since 2015 in the previous paragraph. The BIS website says they concluded a survey of this industry last year. Its hard to believe BIS can draft, distribute, collect, and analyze a survey of all microelectronics firms in the US and then write a report that will offer information about the US microelectronics industry in 180 days thats not already available.
Commerce Department > Energy Department:
More generally, the CHIPS Act overburdens the Commerce Department with a lot of deliverables. This is because Commerce has a lot of statutory authorities. But, “the department of everything else” doesn’t have the technical depth to do everything. Even though BIS, ITA, NIST, and USPTO all have some microelectronics equities, Commerce doesn’t have the volume of technical experts needed to effectively administer these taskers. In contrast, The Energy Department has world-class technical talent and clear microelectronics-related equities (from NNSA to the national labs to DOE-SC), very few of whom get a direct call-out in the CHIPS Act.
Workforce Support Lacking:
Except for a perfunctory call that the Secretaries of Labor and Energy coordinate with the National Academy of Sciences to support microelectronics workforce efforts, the CHIPS Act underemphasizes the most important thing this industry needs: a stable talent pipeline of electrical engineers, chemists, and physicists. This omission is strange because the original version of the Act had more explicit language on this issue (section 5). Specific requirements for affordable access to prototyping and low volume manufacturing (ex. modernizing MOSIS and expanding nanoFabs), essential resources for training the next generation, are also lacking.
Conclusion
Many of the critiques above may be mitigated depending on the competence and capability of the government agencies that actually implement this legislation. Or maybe these problems are addressed somewhere in the other 4,471 pages of the NDAA. But the more that CHIPS Act provisions delegate to Executive Branch and/or implementing agencies, the more Congress’s intent becomes subject to interpretation. A recent report found that PPPs, such as the ones the CHIPS Act calls for, need to be funded at $200-300M/year at a total budget of $2-3B on a 10-year timeline. This CHIPS Act is a great start, but Congress needs to make sure its 2021 appropriations work matches the scope of the Act if the US is going to win this race by innovating faster.
Note: the views here are my own and drawn solely from the documents I cite here.