By Stephen Bryen
The Chips Act of 2022 is a $52.7 billion subsidy to semiconductor companies in the United States. The funding, mostly aimed at the construction of new chip fabrication facilities, called Fabs in the trade, can be awarded in up to $3 billion chunks of cash to any US-based company making “advanced” semiconductors.
The only US-based companies excluded are those owned by China, although the Act does not actually explicitly name China. However, if China invests through proxies or cutouts, Chinese companies could still get US money.
In theory, the CHIPS Act subsidy is intended to fix a serious problem which is US dependence on two Asian companies, Taiwan Semiconductor (TSMC) and Samsung. The US has experienced severe supply chain problems which has been very costly to the US automobile and truck industries and probably has delayed the manufacture of some weapons, such as the Javelin anti-tank missile.
Both auto makers and weapons makers were caught by surprise. Had they pre-stocked enough integrated circuits to deal with potential supply issues, there would not be a supply problem. Even now, as chip production has regained momentum in Asia, it would be sensible to build up a chip stockpile, especially given the extreme political volatility in Asia because of China’s incessant and growingly reckless threats against Taiwan. Unfortunately the CHIPS Act offers no funds to subsidize building stockpiles, not even for the military. Recently Raytheon told the Pentagon that replacing Javelin and Stinger missiles, used up in the Ukraine war, would take a few years. One of the reasons was the need to procure the right chips from Asia.
The CHIPS Act theoretically solves one problem, the fact is that US-based companies have not invested in advanced semiconductor manufacturing. Until the first new plant in Ohio comes on line in a few years, US companies will remain dependent on Asia for advanced chips.
Chip production is considered advanced if its output has a feature size of 7nm (nanometers) or less. The Intel Ohio plant, once it is operational, will produce integrated circuits at 5nm. It is likely Intel will target the auto industry and maybe 5G devices. However, there is no assurance that Intel will be able to compete against TSMC or Samsung. Both companies use the most modern extreme ultraviolet photo lithographic processes and have perfected production techniques through long experience. The US companies, such as Intel and other American fabs, are inexperienced in applying 5 nm technology, meaning that one can rightly expect their wafer yields will be lower than their Asian counterparts, and chip cost will be higher while profitability lower. Meanwhile, while US companies will be learning how to produce 5nm chips, companies such as TSMC are already moving on to 3nm and (possibly) 2nm feature sizes. US companies will also have to train new workforces, which means another steep learning curve.
On a positive note, there are funds in the CHIPS Act for worker training, even though that money comes with some strings attached.
GPU Chips for Data Mining
There is nothing in the CHIPS Act that tells a US semiconductor what to make, aside from making advanced integrated circuits. This means that US-based companies will seek to service the US commercial market and will have little interest in custom projects for US defense companies.
The commercial market does include chips that perform artificial intelligence (AI) functions. AI is of increasing importance for military applications. Three types of chips are significant for AI –field programmable gate arrays, applications specific integrated circuits, and graphics processing units. In addition, there is a trend to produce advanced SoCs (system on a chips). The military can rely on commercial chips, but unless these chips are somehow controlled, many will wind up in China and Russia as in Ft. Worth or Detroit.
Security standards for subsidized US advanced chip production are not part of the CHIPS Act. While the Act excludes Chinese ownership of subsidized fabs, it does not exclude foreign workers, including from China. Nor does the Act recommend measures to protect intellectual property, or require cyber security for the design centers and fabs that will be funded under the CHIPS Act. The Act also does not take up the issue of US export controls. It is an observable truth that the failure of the US export control system, the dismantlement of the international COCOM system, has proven to be a bonanza for China and it has, along with substantial foreign espionage and cyber spying, enabled China to leapfrog the US in many military categories, and match the US in most others. The increasing parity of China’s military and its proximity to potential conflict areas, has the US scrambling to see what it can do with older, worn systems to counteract the growing menace of an aggressive China.
The CHIPS Act failure to specify the types of integrated circuits needed for US defense also is lamentable. In fact, the Act specifies almost nothing and leaves it up to the US Commerce Department to hand out goodies to US and foreign companies who want to build plants in the United States. At the most elementary level, one would need to ask why TSMC or Samsung, for example, would build a fab in the United States in order to compete with itself? Instead, with subsidies in hand, these companies will offload older product production so they can optimize the newest at home.
There also is a key concern is that Washington will reward its political friends and shun all others. The penchant of the Biden administration to play favorites with grants, like underwriting failed solar and wind energy ventures while strangling the US oil and gas industry, including refineries and pipelines, is a case in point.
The Act likewise makes no provision that companies that get subsidies stop outsourcing high end products abroad. So, for example, a US company could keep buying AI chips from Asia while selling other advanced chips to customers. Others already have remarked that this corporate handout lacks any system of accountability. Here is the money, build the plant and produce chips. No other questions asked. No requirements, even if ventures fail.
In the end the US semiconductor industry will be rewarded for failing to invest in its own future. The US military won’t be any less dependent on Asian production, or will have to pay through the nose to get the parts it needs. China will continue to have nearly unfettered access to US know how and technology. And the US probably will not solve its competitive problem vis a vis Asia.