The Do’s and Don’ts of industrial policy

Scientist holds a chip in a laboratory. (Max Acronym via Shutterstock)

Industrial policies are back. Such policies – which try to pick winners and losers among industries and firms, and which fell out of favor in the United States three decades ago – are in vogue again. These policies were the beating heart of Bidenomics, and the Trump White House is enthusiastically adopting them too, though in a form different from anything the United States has attempted before. 

Both administrations have targeted the semiconductor industry as a key focus for intervention. The sector also figured prominently during the heyday of U.S. industrial policy, which ran from the 1960s to the 1990s. The Biden Administration’s signature measure in this area, the 2022 CHIPS and Science Act, aimed to increase America’s market share in semiconductor manufacturing by subsidizing U.S. construction of new plants, called wafer fabs in the industry. In August, President Donald Trump announced that the government would take a 9.9% equity stake in the semiconductor producer Intel – an unprecedented demand by Washington – in exchange for releasing funds promised to the firm by his predecessor.  

The track record of government policies aimed at this industry provides clear takeaways on the uses and limits of industrial policy more generally. Most government-sponsored attempts to promote innovation and job creation have failed, in the United States and other countries. Other initiatives – including the CHIPS Act – have partly achieved their goals, but at a higher cost than advertised. Whether the CHIPS Act represents a true success, therefore, depends on the hard question of whether the benefits are worth the costs. 

Why governments try 

Economists define industrial policies as attempts by a government to shift the structure of an economy in ways that would not occur absent public-sector intervention in pursuit of strategic goals. Past industrial policies in the United States have typically aimed at strengthening defense-related industries or raising economywide productivity levels. More recently, reducing the economy’s dependence on fossil fuels has come to the fore as a key objective. 

Industrial policies and tariffs have overlapping goals, but they differ in their methods. The former direct subsidies to specific firms and therefore cost the government money, while the latter protect entire industries and raise revenues, mostly from U.S. consumers. Both types of policies aim to increase a country’s global market share in specific industries, but doing so often comes at a cost. The resources – money, workers, and investment – flowing into favored sectors have to come from somewhere, which can mean that other industries get squeezed. 

If the government’s goal is to raise productivity, then the best industries to target are those that enjoy growing demand, large potential for productivity gains, and significant spillovers to other industries – meaning that productivity improvements in the first industry will help other industries become more competitive, too. The semiconductor industry fits these criteria well.  

Critics offer three arguments against industrial policy. First, would-be central planners generally lack enough expertise to accurately pick which firms to back and to properly design and target subsidies in a way that will effectively help them. Second, industrial policies run a high risk of degenerating into a form of crony capitalism. This happens when the targeted industries “capture” – that is, find ways to wield disproportionate influence over – the policymaking process, and become addicted to subsidies without becoming more competitive. Third, capture can work in the other direction as well, if policymakers use subsidies not to strengthen the competitiveness of targeted firms but to force corporate management to accept their political agenda.

Sort-of successful 

The U.S. government’s most successful effort to boost semiconductor makers came in the early days of the industry. In 1958, engineers at Texas Instruments and Fairchild Semiconductor invented the integrated circuit. A few years later, the U.S. Department of Defense created a market for Texas Instruments’ products by incorporating them into the guidance system of the Minuteman missile. NASA, meanwhile, became Fairchild’s dominant customer by using the company’s components in the Apollo moon landing program. Defense and space procurement jump-started America’s industry and helped it dominate the worldwide electronics market for two decades. It’s not obvious whether government procurement from domestic manufacturers should count as industrial policy if government buys products at a fair price with no embedded subsidy. But in any case, the federal government helped the nascent semiconductor industry get off the ground. 

In the 1980s, the Reagan Administration, concerned about the rise of Japanese government-backed producers of semiconductors, singled out that industry for more ambitious industrial policies. The administration applied tremendous pressure on Tokyo to agree to a fixed market share for U.S. producers in Japan. It also created the Sematech consortium to develop next-generation semiconductor production equipment.  

Neither initiative achieved lasting success. While U.S. producers briefly attained their targeted market shares in Japan, most of them didn’t become sufficiently competitive to sustain their position when Washington lost interest. The main result was the rise of Korean and Taiwanese producers at the expense of Japanese firms. As for Sematech, some of the U.S. equipment production firms backed by the consortium went out of business within a few years of receiving federal R&D subsidies. 

Japan, meanwhile, demonstrated that sufficiently large subsidies under the right circumstances could produce substantial gains in market share for older products like memory chips. But Japan’s Ministry of International Trade and Industry – celebrated by Western industrial policy enthusiasts in the 1980s – largely missed the fastest-growing segments of the market: the digital signal processors that enable mobile data transmission and the graphics processing units (GPUs) that have proved essential to advances in AI. As a result, Japan now lags far behind other advanced countries in the production of such products.  

Two technicians hold a reflective wafer at a semiconductor plant. (Aslysun via Shutterstock)

The semiconductor industry has evolved differently in the 21st century than most observers expected during the debates of the 1980s. Western companies have broken down their production processes, outsourcing capital-intensive steps to “foundry” producers – above all, Taiwan Semiconductor Manufacturing Company (TSMC) – and testing and packaging steps to companies in Southeast Asia. U.S.-based firms, including Intel and Texas Instruments as well as newer “fabless” companies like the GPU giant Nvidia and the mobile data innovator Qualcomm, now control almost half the industry’s revenues and perform more than half its R&D. But just 10% of semiconductor manufacturing takes place in the United States. Semiconductor jobs in the United States – almost a third of which are in R&D – pay an average of about $84,000, while those in Taiwan pay an average of about $46,000. 

Today, policymakers of both parties worry about China’s growing clout in the industry and the Chinese threat to Taiwan, where 90% of the world’s most advanced chips are fabricated. Measures restricting Chinese access to U.S. technologies have had modest short-term effects, but also have led China to redouble its efforts to secure independence from Western technology and increase China’s market share. In the United States, both left- and right-leaning populists also view the semiconductor industry not only as a national security priority but as a central focus of efforts to increase domestic manufacturing jobs.  

The CHIPS Act and Intel 

The CHIPS Act authorized $39 billion in grants and tax credits for new wafer fabs, plus $13 billion for three new university-operated R&D centers. The Biden White House committed $33 billion to fab projects and $7 billion for a nonprofit entity to oversee R&D grants.  

These commitments included about $8 billion for new Intel sites in Ohio and Oregon, $7 billion for TSMC manufacturing and R&D facilities in Arizona, $5 billion for a Samsung fab plant in Texas, $1.6 billion for additions to a Texas Instruments site in Texas, and other grants, with billions in additional tax credits available to these firms if they hit defined milestones. Yet for all this largesse, the administration released only about $5 billion of promised funds before leaving office. 

This February, President Trump called the CHIPS Act a “horrible, horrible thing” and stopped payment on outstanding commitments. No grantees received money between Jan. 21 and late August, according to company financial filings. President Trump also disbanded the Biden Administration’s R&D entity, noting that it was staffed by Biden White House veterans, raising questions about whether the R&D program will continue.  

President Trump has demanded that, as a condition for receiving funds, grantee firms boost their planned investments. This has led to increases from TSMC and Texas Instruments. In August, the president announced the federal government’s plan to take a stake in Intel, which it has since done. It’s worth noting that Intel made no new investment commitments and has scaled back its investment plans in view of soft end-markets for its products.  

The results so far 

Since the CHIPS Act was launched, producers have announced U.S. investments amounting to about $300 billion more than they likely would have spent absent the legislation. (This estimate is based on trends in place before Congress took up the bill.) But every grantee has indicated it will only invest its announced amount if business growth justifies it. I estimate that grantee firms are on track to invest around $120 billion more than they otherwise would have between 2023 and 2026, plus another $60 billion over the subsequent six years if they keep investing in their expanded U.S. facilities at historically normal levels. CHIPS Act grants and tax credits will cover a majority, though not all, of the extra cost of building facilities in the United States as opposed to Taiwan. 

Factories made possible by the CHIPS Act will employ about 30,000 U.S. manufacturing workers, according to company announcements. (They’ll employ considerably more construction workers to build them, but it doesn’t make sense to count these as new jobs, since virtually all U.S. construction workers are already fully employed, so these projects will just cause other projects to be delayed.) According to industry estimates, these facilities will raise America’s market share in chip manufacturing to around 14% from 10%. For the most advanced semiconductors, the U.S. manufacturing share will rise to just over 25% (from zero today). 

The CHIPS Act will have enduring effects. Based on announcements to date, it will reallocate around $150 billion of industry investment – expected gross investment less subsidies and tax credits – to the United States from other countries over a decade, amounting to about a tenth of worldwide investment. It will create 30,000 jobs that pay moderately better than U.S. averages. The share of America’s chip purchases that come from China and Taiwan will fall by around 2 to 3 percentage points, but will remain high.  

These benefits, however, will come at an expense of at least $32 billion in taxpayer dollars, as well as reduced construction in other industries. In other words, for each job created, the subsidies will total more than $1 million – an expensive way to increase manufacturing employment, even compared with past industrial policies. These policies will advance national security by marginally reducing America’s vulnerability to Chinese coercion, but it’s difficult to quantify this improvement. And the industry will likely request more money in the future. 

Intel processor in a laptop. (Jochen Gittel via Shutterstock)

Other current policies may reduce the impact of the CHIPS Act. Because production processes are distributed across the world, higher tariffs on imported components will likely add to the costs of fabricating chips in the United States, reducing production. Labor shortages associated with immigration restrictions, meanwhile, will likely increase construction costs, reducing the ability of producers to add manufacturing capacity. 

It is too early to predict how the Trump White House or future administrations will use the leverage that comes from equity stakes in Intel and perhaps other large U.S. enterprises. History suggests the possibility of politically driven demands on management – which would not contribute to American prosperity in the long run.

What does this all indicate about the wisdom of industrial policies today? In certain sectors like semiconductors – with growing demand and large room for productivity gains from increasing production scale – subsidizing domestic manufacturing can produce some of the desired effects. The cost to taxpayers of creating manufacturing jobs in the United States through industrial policies is typically high, but it may be a cost worth paying if the policy accomplishes other national goals. In semiconductors as in certain other well-targeted industries – essential pharmaceutical ingredients, for instance – increasing the resilience of vital supply chains and reducing vulnerability to foreign coercion may be worth imposing even relatively high costs on taxpayers. 

But the costs of industrial policy are real, as are the risks it raises by increasing political control of industry. So we shouldn’t get too excited about its return.