A 12-inch wafer is displayed at a semiconductor manufacturing company.

Blog Name

The US and Korean CHIPS Acts are spurring investment but at a high cost

Date

Photo Credit: REUTERS/Richard Chung

Note: This research is part of a 2024 series on the Korean economy. PIIE gratefully acknowledges funding from the Ministry of Economy and Finance, the Republic of Korea, for the research presented in this blog. The research was conducted independently. Funders are never given the right to final review of a publication before its release.

Body

Soon after taking office in 2021, President Joseph R. Biden, Jr. threw his weight behind efforts to use federal government resources to strengthen US capabilities to produce cutting edge semiconductors to keep up with global competitors, especially China. The resulting Creating Helpful Incentives to Produce Semiconductors (CHIPS) and Science Act of 2022 is one of the world’s most ambitious industrial policies today, with subsidies, loans, tax credits, and support for research and development (R&D) estimated to total $79.3 billion over 2022–31. In fact, the CHIPS Act has been much more successful in attracting investment in the United States than anticipated, luring leading-edge chip producers who might otherwise have been tempted to invest in other countries. South Korean firms are major beneficiaries of CHIPS Act funds, but the pull of US subsidies is also a challenge to Seoul and governments in other allied countries, compelling them to redouble support to semiconductor firms in their own countries just to defend their current positions in the global chip value chain, let alone to build new capabilities.

Last year, Korea responded to the US CHIPS Act and other countries' increased subsidies with its own Korean Chips Act, or K-Chips Act, increasing existing tax credits and continued assistance in kind, such as with infrastructure for chipmaking facilities. The architect of the K-Chips Act, a former Samsung executive, has said: “The winner of the global chip battle will control the economic security order, while the loser will end up becoming a technological colony.” President Yoon Suk Yeol has concurred, likening the semiconductor competition to “an industrial war.”

The United States is on track to add more investment in construction for computer and electronics manufacturing in 2024 alone than it did over the 20 years before the CHIPS Act first passed, according to US Census Bureau data.[1] A recent report by the Boston Consulting Group and the Semiconductor Industry Association (BCG/SIA) projects that the United States, starting from zero today, will have achieved its main goal of attracting a significant share of advanced logic chip fabrication onshore by 2032 at the latest.

The BCG/SIA report also projects that Korea will attract $300 billion in capital expenditures for chip fabrication in the next eight years, or 18 percent of its 2022 GDP.[2] According to the Wall Street Journal, the government in Seoul plans to help create a cluster of 37 chipmaking factories in eight cities, creating more than 3 million jobs.

But a striking feature of the US CHIPS Act, one that no doubt raises concerns in Seoul, is that it has enticed leading firms not just in the United States, but also from Korea and Taiwan, to make major investments, while Korea’s policies have largely not yet attracted firms headquartered outside Korea. Both policies may be necessary to achieve a variety of economic and non-economic goals, but they are expensive, generating an effective bidding war for investors that ramps up budget outlays while risking overproduction in a notoriously volatile marketplace, now also filled with uncertainty about future demand for chips for artificial intelligence (AI).

What has become an effective subsidy race should have sobering implications for policymakers. To avoid such a race, the United States, Korea, Japan, and the European Union should enhance coordination to monitor the global expansion of semiconductor production with a focus on avoiding overcapacity and reducing the need for continued renewal or increases of subsidies when the current sets of policies expire.

The United States and Korea in Semiconductor Manufacturing

A reliable supply of chips used in everything from dishwashers and cars to AI datacenters is increasingly seen as essential for national security and economic prospects. But unlike oil, chip production is based not on natural endowments but years of strategic investments and competition by firms and governments. Government industrial policies to entice leading semiconductor firms to invest in production in national jurisdictions have long been widespread, but since the end of the pandemic they have spread to more countries with larger programs, from the United States and Europe to Korea, Japan, and China.

The United States and Korea are among a handful of leading countries in the global semiconductor supply chain, with different and complementary strengths.[3] The United States is strongest in electronic design tools, semiconductor manufacturing equipment, and logic chip design, while Korea is strongest in memory chip design and production, as well as advanced logic chip production. The US CHIPS act was primarily driven by concerns that no advanced logic or memory chips are produced within the United States.

Before the CHIPS Act, a previous BCG/SIA report estimated only 10 percent of total global wafer fabrication[4] capacity was located in the United States, which was projected to decline further. Measured in sales volume, though, Gary Clyde Hufbauer and Megan Hogan find that US production was at around 20 percent in part due to the United States specializing in more advanced chips (though not the most advanced ones). Korea has a higher share of 17 percent of semiconductor fabrication capacity, including both cutting-edge memory and logic chips. The United States and Korea are both concerned about the impact that China’s rapid, heavily subsidized advancement in many segments of semiconductors could have on their competitiveness and potential dependence on future Chinese production.

Existing Industrial Policy was more Generous in Korea

“The semiconductor industry is one where companies demand incentives and will not invest without them,” according to subsidies expert Kenneth P. Thomas. Estimating such incentives is a major challenge because support comes in many forms, from both national and subnational governments, and is often opaque. The Organization for Economic Cooperation and Development (2019) compared subsidies provided to major chipmaking firms and found that Chinese firms received the largest subsidies as a share of revenue. Korea's Samsung was the largest subsidy recipient, but only because of its large size (Samsung only received 1 percent of its revenue in government support and SK Group even less). US semiconductor firms received some funding from the United States but also from other jurisdictions in which they have located production or research, such as Singapore and Taiwan.

An earlier BCG/SIA report compared the cost of building and operating a semiconductor fabrication facility for 10 years, which they found to be much higher in the United States than in Korea and other Asian countries. The report estimated Korea was similar to Taiwan in the scale of subsidies, around double that of the United States before the CHIPS Act, and found China provided the most generous subsidies. The report attributed 65 percent of the US-Korea cost gap in advanced logic chip manufacturing to Korea’s comparatively more generous government incentives. These have included R&D support, tax incentives, and infrastructure help such as support for power and water supplies, which semiconductor producers use in immense quantities.

What do the respective Chips Acts do?

The US CHIPS and Science Act combines $38.2 billion in direct funding for semiconductor manufacturing, such as grants, including up to $6 billion to provide loans and loan guarantees, and $13.2 billion for workforce development and research and development. The CHIPS act also provides a special tax credit of 25 percent for investments made in buildings and certain equipment used for semiconductor manufacturing in the United States that begin construction by the end of December 2026, which the Congressional Budget Office initially projected would cost another $24.25 billion in foregone revenue.[5] The tax credit is available to any facility whose primary purpose is the manufacturing of semiconductors and semiconductor manufacturing equipment. The Department of Commerce, which oversees most of the CHIPS Act implementation, has rapidly built industry expertise within a specialized CHIPS Program Office that evaluates applications for grant funds.

CHIPS Act recipients are limited through so-called “guardrails” from expanding their production capacity in China to avoid the concern that fungible CHIPS Act money could free up money used to finance production there. This limitation is a major concern for Korean businesses that have large operations in China but that hope to benefit from CHIPS Act funding to produce facilities in the United States.

The Korean K-Chips Act started small but has grown into a much larger program combining tax credits and preferential loans. In December 2022, the Korean government passed an 8 percent tax credit for investment to expand facilities in a set of “National Strategic Industries” including semiconductors. This fell significantly short of what was on offer in other countries, in part due to the political opposition's resistance to “preferential treatment for large corporations.” To more closely match support on offer from other jurisdictions, the government passed an amendment in April 2023 increasing the deduction to 25 percent for small and medium-sized companies and 15 percent for larger firms—still short of the tax credit in the United States. R&D investments and boosts above past patterns of capital spending get even larger credits. Since the K-Chips Act, the Korean government has continued to expand public support for the chips sector, deeming the K-Chips act alone as insufficient. The credit was originally meant to expire at the end of 2023 but has been extended to 2024. In May 2024, the government announced another $19 billion incentive package, with details to come, in addition to a project to form a megacluster of semiconductor production in the Gyeonggi province, with around $470 billion in private investment over the next two decades.

The two acts are difficult to compare directly because there is not yet available data on how much the Korean government has given in tax credits. Samsung reported a 5.3 trillion Korean won (KRW) increase in its tax credits in 2023 (around $3.85 billion), and SK’s tax credits carried forward increased by around KRW 630 billion (around $458 million), adding up to around $4.3 billion; both companies are active outside the chips industry, though, so other tax credits unrelated to the K-Chips Act are likely included in these upper bound figures. $4.3 billion is nearly identical to the $4.5 billion in annual tax credits implied if the BCG/SIA prediction of $300 billion in capital spending for the next decade is divided, $30 billion per year, and multiplied by the 15 percent tax credit that large firms like SK and Samsung can take.

Impact on Production

Anticipating passage of the CHIPS Act, the United States has seen a boom in manufacturing construction spending, which is driven largely by computer, electronics, and electrical manufacturing. Prior to the CHIPS Act (and the Inflation Reduction Act of 2022), such spending averaged around $6 billion per year (figure 1) but is up more than 20-fold to an annual pace of $128 billion as of March 2024.[6]

A US Treasury report examined the most analogous data series on construction manufacturing in other countries and did not see similar rises, including in Japan. The report did not include Korea, but the  Korean government’s statistical agency produces a data series on construction starts for factories: That category shows a pronounced decline starting in 2022 , which has stayed significantly below the last decade’s average pace despite Korea’s additional tax incentive for additional investment put into place in 2023 (figure 2). It may be that the credits target a set of investments narrow enough not to show up in the aggregate data, but considering that semiconductors were over 20 percent of Korean exports in 2023, they should be large enough to show up in these data if Korean industrial policy were leading firms to expand their production capacity significantly. Chip production has grown since late 2023, however, which may be due to higher utilization of existing capacity amid a cyclical recovery in chip demand.

Before the CHIPS Act was passed, the BCG/SIA projected both the United States and Korea to lose 2 percentage points of their expected 2020 global semiconductor fabrication capacity share by 2030. However, the stronger than expected response to the US CHIPS Act led the BCG/SIA authors to revise their US projection for 2032[7] to 14 percent of global capacity, up 4 percentage points, while Korea is expected to gain 2 percentage points to 19 percent of global capacity. Figure 3 shows how each country's capacity share is projected to evolve in each key niche of semiconductor production.

Figure 3 After passage of the US and Korean Chips Acts, Korea is projected to build on its lead in memory chips while losing ground to the United States in advanced logic chips

Supply chain reliance on Taiwan for advanced node logic chips, one of the key reasons Bloomberg Economics projects that potential conflict over Taiwan would be more costly to the global economy than the COVID-19 pandemic or the global financial crisis of 2008, is expected to decline from over two-thirds to less than half of global capacity in 2032. Highlighting the challenge posed by the US CHIPS Act, Korea is projected to decline from 31 percent to 9 percent of global capacity for advanced node logic chip production, most which is projected to move to the United States. The BCG/SIA report projects that the United States, with no production of advanced node logic chips in 2022, will produce 28 percent of the global supply of these chips in 2032, and the European Union and Japan—also starting at zero production—are expected to produce around 5 percent each. The report is even more optimistic than Commerce Secretary Gina Raimondo, who recently said the United States is on track for a 20 percent share of advanced chip production by the end of the decade. A Taiwan conflict would still be disastrous for the global economy, and the most advanced generation of chips will still be entirely produced in Taiwan,[8] but if the projections prove accurate, then a much more diversified base of production would dampen the impact.

The pattern of firms investing in fabrication in the United States and Korea looks very different, however. Investment funding the United States’ projected $646 billion in capital expenditure from 2022 to 2032 is evenly balanced between firms headquartered in the United States, Taiwan, and Korea, while the source of Korea’s expected capital expenditure of $300 billion is almost all coming from Korean firms. While many international semiconductor firms are expanding operations in Korea, they are primarily in only one part of the supply chain: semiconductor equipment makers Tokyo Electron, Applied Materials, LAM Research, and ASML. These investments, however, are for minimal amounts compared to those from domestic firms.

The largest US CHIPS Act recipients so far, in decreasing amounts, respectively, are Intel (US), TSMC (Taiwan), Samsung (Korea), and Micron (US), but SK Hynix (Korea) is expected to receive an award too and already has announced a nearly $4 billion memory chip packaging facility in the United States. Despite initial negative news of delays and cultural issues at TSMC’s Arizona facility, in early 2024 the delays eased, the facility has expanded to include three fabrication plants, and the planned production of 3 nanometer (nm) chips has been upgraded to the more advanced 2 nm chip. It is striking to see such a strong positive response despite challenges like finding skilled workers (including obtaining visas for foreign workers) and criticism for what some see as excessive strings, like requirements to ensure childcare availability, that have been attached to grant funds.

While many of the forms of funding have a statutory cap, the much larger than expected response to the CHIPS Act will also significantly increase its price tag. The US Treasury and Internal Revenue Service proposed rules for the investment tax credit to allow the 25 percent deduction for all but the administrative portions of semiconductor fabrication facilities. There is still a great deal of uncertainty about the ultimate CHIPS Act tax credit price tag, in part because the implementing rules have not yet been finalized. But even under very conservative assumptions based on the current investment trends, the CHIPS Act tax credit alone could cost over $85 billion, more than the original CBO projected cost of the entire CHIPS Act and resulting in a total cost overrun of nearly 80 percent.[9]

Conclusion and Recommendations

The K-Chips act is best thought of as a supplement to existing effective strategic support of Korea's semiconductor sector, while the US CHIPS and Science Act is a much more complex and ambitious attempt to catch up in an area where the United States has had limited capacity. The US CHIPS Act has been able to go above Korea’s support levels with tax credits and grants. Korea's continued expansion of incentives indicates policymakers in Seoul are legitimately concerned about being outbid, both by the United States and other jurisdictions.

While both countries are projected to gain global semiconductor share, the United States’ larger subsidies are attracting larger investment, attracting companies from outside most recent US expertise and from both Korea and Taiwan to get a more complete value chain located inside the United States. The cost is likely to be much higher than policymakers anticipated, but these bets by the United States and other jurisdictions could lead to a much more geographically diversified supply chain that is less dependent on Taiwan, as more regions around the world make more advanced semiconductors.

The most important factor for whether these bets pay off will be the future trend of demand. Chips are a cyclical business, with wide-ranging structural factors—from large AI-driven demand for specialized chips known as GPUs, high bandwidth memory (HBM), cloud computing, electronic vehicles, and the Internet of Things (IoT)—driving expected future demand.

Some argue that the industry could gain useful resilience with somewhat more slack in the supply chain (15–20 percent) than exists today, avoiding shocks caused by the pandemic era chip shortage and other developments. Yet damaging overcapacity could still result. China, which imports many of its chips today and is the main market for South Korean chip exports, may be much less of a market for non-Chinese in the future as it pushes for self-sufficiency. If the AI boom fails to meet expectations, or if the global economy weakens, demand could slump below projections. The risk is that if all the new supply is far more than demand, then utilization rates will be low, semiconductor firms will beg for more subsidies to cover losses, prices will fall, and some of the lavishly subsidized capacity will need to be shut down, a waste of taxpayer resources.

It is also yet to be seen whether the CHIPS Act momentum in both countries can be sustained when the current batch of subsidies end and projects come on line. Hopefully, newly built competencies and experience in advanced chip production, more upstream and downstream links in the supply chain locating in the United States, and other CHIPS Act programs like those to boost the available workforce make it feasible to build future capacity in the United States with less future government support. 

Global coordination, at least among the G7 nations, will be needed to foster resilience, curb the subsidy race, and prepare for shifts in demand while avoiding a waste of US, Korean, and other countries' national resources.

Notes

1. What is now known as the CHIPS Act was first passed as part of the 2021 National Defense Authorization Act in December 2020. Investment began to rise with the expectation that Congress would appropriate funding for it, which it did in July 2022, and President Biden signed the final CHIPS and Science Act into law in August 2022.

2. Calculated using page 13 of the BCG/SIA report Emerging Resilience in the Semiconductor Supply Chain for the investment figures, World Bank open data for GDP.

3. Semiconductors are generally produced on wafers, which are divided into “chips” that are then integrated into devices. Chip production is often measured as the capacity to produce a given number of wafers in a given period. For a detailed overview of the semiconductor supply chain, see https://www.piie.com/publications/working-papers/how-united-states-marched-semiconductor-industry-its-trade-war-china.

4. Semiconductors are generally produced on disc-shaped wafers, which contain many “chips” that are separated from the wafer and incorporated into devices. Output is often measured in wafers rather than chips.

5. For a detailed dive into the CHIPS and Science Act, see Hufbauer and Hogan (2022).

6. Author's calculation based on Census Data, US Census Bureau.

7. Projected a decade from CHIPS Act passage.

8. TSMC’s Arizona facility expects to produce 2 nanometer (nm) chips by 2028, but it expects to produce 1.4nm chips in Taiwan by that time, and 1nm by 2030.

9. The Census Bureau data on construction for computer/electronics/electronical manufacturing facilities presented earlier are a sensible basis for estimating the ultimate cost, with adjustments. Semiconductor manufacturing equipment, which industry estimates suggest makes up at least 50 percent of the cost of a fabrication facility, counts towards the tax credit, but the Census Bureau clarified to the author that the construction data do not include this equipment. Therefore, multiplying this investment by the tax credit would tend to underestimate the budgetary cost substantially.

On the other hand, some of the investment reported by the Census Bureau is related to the Inflation Reduction Act (IRA) incentives passed nearly contemporaneously. The Rhodium Group and Massachusetts Institute of Technology Center for Energy and Environmental Policy Research’s Clean Investment Monitor estimates IRA-related manufacturing investments in areas like batteries, which, according to Census representatives, would be included in the Census manufacturing construction for the computers/electronics/electronical data series. Subtracting the full amount of Clean Investment Monitor estimated investment that is likely in the Census computer/electronics/electronical data from the Census figure yields an upper bound of 28 percent and 32 percent of the manufacturing total in 2022 and 2023, respectively. It is an upper bound because the Rhodium/MIT data refer to manufacturing investment more broadly, not just construction. It thus seems sensible to estimate that the remaining approximately two-thirds of Census reported construction investment is related to the CHIPS Act and could get the tax credit.

The estimate in the text assumes that all the resulting two-thirds of computer/electronics/electronical construction investment is deductible, which should be reasonable, considering the IRS proposed rules to only exclude the least costly parts of a fabrication facility like those for administrative uses without the complicated purification equipment. It also assumes that the pace of investment in computer/electronics/electronical construction investment will continue at the 2024 Q1 pace throughout 2024 and into 2025 and 2026, and that the IRS-related share will stay at one-third. The estimate is biased down by the lack of data on semiconductor equipment in the Census series, the overestimation of IRA investment taken out, and the cutoff in 2026 (considering current proposals are to continue to allow deductions beyond that date as long as the site’s construction started by the end of 2026).

Data Disclosure

The data underlying this analysis are available here.

More From