Chips on the table

In 1987 American chipmakers did what American firms rarely do: they asked the government for help. The result was Sematech, a consortium that received federal money from 1987 and, a decade after its founding, was “standing on its own without annual funding from the Government”.^{3DARPA innovation timelineSEMATECH} The programme is remembered as a success, in part because it ended. On 9 August 2022 President Joe Biden signed a far larger bill with the same purpose, and we shall see whether it learns the same manners.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022}

The CHIPS and Science Act puts $52.7bn into semiconductors. Of that, $39bn goes to manufacturing incentives, including $2bn for the older chips that go into cars and weapons; $13.2bn is for research and workforce development; and $500m is for international supply-chain work.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022} Sitting alongside is a 25% investment tax credit for factory and equipment spending.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022} Recipients may not use the money for share buybacks or dividends, and must not build certain facilities in China and other “countries of concern”.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022}

A shrinking slice. America's share of global semiconductor manufacturing capacity, %

	United States
1990	37
2021	12

Source: Semiconductor Industry Association and Boston Consulting Group, April 2021

The case for acting is not hard to make. According to the Semiconductor Industry Association, America’s share of global chip manufacturing capacity fell from 37% in 1990 to 12% by 2021, which it blamed largely on the subsidies of rival governments.^{2Semiconductor Industry AssociationSIA Urges U.S. Government Action to Strengthen America's Semiconductor Supply Chain5 Apr 2021} The White House says America makes about 10% of the world’s chips and “none of the most advanced”, while East Asia makes 75%.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022} Anyone who has watched a car plant idle for want of a small part will grant that concentration has a price. Free-market purists say a country should buy chips where they are cheapest. That is fine until the cheapest place is a short boat ride from a hostile neighbour. Insurance is not protectionism, so long as the premium is sensible.

Is it? The sums are large and not as large as they look. Micron alone has promised $40bn for memory chips, which is roughly the size of the whole manufacturing pot.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022} Companies announced nearly $50bn of new investment in a single week, bringing business commitments to nearly $150bn since Mr Biden took office.^{1The White House, via the American Presidency ProjectFACT SHEET: CHIPS and Science Act Will Lower Costs, Create Jobs, Strengthen Supply Chains, and Counter China9 Aug 2022} Whether the Act caused these pledges or merely arrived in time to claim them is the question every subsidy invites and none answers cleanly. Nobody can see the factory that would have been built anyway.

That is the case for the tax credit and against the grants. A credit is open to anyone who builds. It does not require an official to decide that this firm and not that one deserves help, and it cannot be lobbied into a favour. A grant, by contrast, is a negotiation, and officials have been handed $39bn of negotiating to do, with the leverage arguably on the other side of the table. Firms that can credibly say they will build in Taiwan or Korea instead have every reason to bargain hard. The bureaucrat who must decide how much is enough has no way of knowing.

The guardrails raise a second worry. Banning buybacks and restricting expansion in China sounds prudent and mostly is. But a firm that cannot grow in its largest market may reasonably decide that it would rather skip the money than the market, and Washington cannot be both a patron and a strict landlord without some recipients concluding that the rent is too high. The aim of the strings is good. Their drafting will matter more than their intention.

Defenders of the Act reply that this is an unfair standard, since every industrial policy in history has had to choose winners and America is simply late to the contest. That is partly true. The better reply is Sematech, which suggests that the right time to judge a subsidy is the day it stops. Sematech was a single consortium with a single aim, and in the end it was finished.^{3DARPA innovation timelineSEMATECH} The new Act has a long list of programmes, a generous credit and a stated hope that America will once more make the most advanced chips. It has no finish line.

We would keep the credit and the research money, scrutinise the grants and write the exit into the law. Congress should decide now what a successful chip industry looks like, and when the cheques stop. A subsidy without an end date is not insurance. It is a pension, and the chipmakers will ask for increases.