Did U.S. Semiconductor Export Controls Harm Innovation

Commentary by Andreas Schumacher

Published November 5, 2024

Introduction

In recent years, the United States has enacted a series of targeted sanctions and export controls aimed at curtailing China's progress in advanced semiconductor technologies. These measures, implemented over the course of the last two administrations, have sparked a debate regarding their effectiveness and potential impact on innovation. Critics argue that such controls could stifle technological progress.

This report analyzes research and development (R&D) expenditure, patent filings, and earnings call transcripts from 30 leading semiconductor companies based in the United States and allied nations since 2017. The findings indicate that there is no substantial evidence to date to support the notion that recent U.S. semiconductor controls have hindered these companies' innovative capabilities.

Moreover, the data suggests that by some metrics, companies impacted by the October 2022 controls, on average, even outperformed a comparable group not affected by these controls: the impacted companies exhibited a higher increase in R&D spending and patent filings. This seemingly counterintuitive outcome might be attributable to the companies' advantageous exposure to the rapidly growing market for artificial intelligence (AI) semiconductors, which has fueled their growth and innovation efforts.

These findings offer three recommendations to policymakers:

1. The government should continue to focus on clear, narrowly defined national security rationales and transparently communicate those to market participants.
2. Understanding the market dynamics and competitive situation of affected industries is paramount, and the U.S. government should conduct regular economic impact analysis of its export controls.
3. The U.S. and allied nations should develop new technology export controls multilaterally.

U.S. Export Controls for Semiconductors Date Back to the 1950s

The Coordinating Committee on Multilateral Export Controls established export controls for semiconductor transistors as early as the mid-1950s. During this period, early integrated circuits were primarily manufactured for the U.S. military and NASA, thus placing the early semiconductor industry under stringent export controls.

The perception of semiconductors as dual-use goods evolved with the advent of the consumer electronics industry and the development of a global industry. Semiconductors began to be increasingly regarded as commercial goods. Following the dissolution of the Soviet Union, export controls on telecom items and computers were maintained, yet a priority was blocking the proliferation of weapons of mass destruction. While some semiconductor export controls were maintained on a narrow band of military-spec semiconductors, the prevailing industry sentiment was driven by the exponential growth of consumer microelectronics.

In September 2022, U.S. national security advisor Jake Sullivan reaffirmed the national security imperative of maintaining technological superiority for advanced logic and memory chips. His speech, however, marked one significant policy shift by declaring the need to maintain "as large of a lead as possible" in these critical technologies, rather than a sliding-scale approach. In line with previous policy, those controls were to protect only a "small yard with a high fence."

As a result, in October 2022, and with further refinements in October 2023, the United States implemented new export controls aimed at restricting China's access to advanced semiconductor technology. These controls included limitations on the most advanced AI chip sales and the export of some chip-making equipment.

For industry observers in the United States, allied nations, and China, the onset of this new period had arguably begun three years earlier, in May 2019, when, under President Trump's administration, the U.S. Department of Commerce's Bureau of Industry and Security (BIS) placed severe restrictions on technology exports to Huawei and its affiliates. These moves signaled a more aggressive stance on limiting technological transfers to potential adversaries, setting the stage for the subsequent policies implemented under the Biden administration.

The debate over just how large a yard and high a fence national security requires, and whether controls risk harming the U.S. semiconductor ecosystem without advancing national security remains active and largely anecdotal.

Study Found No Impact of Recent U.S. Semiconductor Export Controls on Innovation

Evaluating a company's capacity for innovation from the outside is more difficult than assessing its revenues or market capitalization[1]. Nonetheless, both quantitative and qualitative indicators are available: R&D spending lays the groundwork for innovation, while patent filings correlate with the development of innovative products. Moreover, significant changes in a company's innovative capacity would also need to be communicated to investors.

This study focused on the 30 semiconductor companies that, as of September 30, 2024, constituted the Philadelphia Semiconductor Index. The analysis relied on public financial data and a detailed, manual review of the company's earnings call transcripts.

Based on statements by the company in earnings calls, either proactively or in response to analysts' questions, companies were classified into one of three groups:

1. Those impacted by the May 2019 entity listing of Huawei technologies (2019 Listing).
2. Companies impacted by the October 7, 2022, U.S. Department of Commerce BIS controls of advanced computing and semiconductor manufacturing items (2022 Controls).
3. Non-impacted semiconductor companies, which constitute the control group.

Two companies were affected by both the 2019 Listing and the 2022 Controls and were included in both groups. Figure 1 provides an overview of the population.

The study then compared cumulative absolute R&D spending (in reporting currency) and the number of patent filings in the two years before and after the implementation of the respective controls. A normalized, non-weighted average was calculated for each group to mitigate quarterly fluctuations and prevent the largest companies from skewing results.

The outcome is summarized in Figure 2:

In the two years following the 2019 Listing, both impacted and non-impacted companies increased their R&D spending by approximately 30 percent compared to the two years prior to the controls. Similarly, patent filings rose by 21 percent and 12 percent, respectively.

Accepting R&D spending and patent filings as proxies for a company's innovative capacity, the data for the 2019 Listing therefore does not indicate a decline in the impacted companies' innovativeness-either in absolute terms or relative to non-impacted peers.

This qualitative finding is further supported by a manual analysis of the impacted companies' quarterly earnings call transcripts over the post-2019 Listing period: the vast majority of company management remained optimistic about innovation, frequently announcing expanded R&D spending. Not a single instance was found when company management reported declining innovation as a result of the controls.

The 2022 Controls present an even more pronounced and surprising picture. Impacted companies increased their R&D spending by 68 percent compared to a 27 percent increase among non-impacted peers.

A patent analysis for this period is not yet possible due to an 18-month lag between patent filing and publication.

As with the 2019 Listing, the quantitative findings for the 2022 Controls are corroborated by earnings call statements: company management consistently conveyed a positive and confident outlook on their R&D efforts to the capital markets.

Finally, while the companies' revenue development in reaction to the two specific events was not a focus of this study - corporate revenues and profits, while a precondition, do not necessarily have to be invested in innovation - the qualitative results are the same.

Unique Markets and Circumstances Could Explain the Results

A financial analyst with a narrow focus on companies' financial performance and earnings calls may not find the results of this analysis surprising. However, it might be unexpected for policymakers and offer new aspects for academic research on export controls. For instance, a comprehensive analysis by the Federal Reserve Bank of New York of U.S. export controls since 1997 across all industries identified negative abnormal stock returns, a reduction in bank lending, and declines in profitability and employment as consequences.

To make sense of this more focused study's findings, a closer look at the semiconductor industry's situation during the relevant time period and the specific nature of the most recent export controls might help. Four explanations are possible:

1. Correlation with Market Dynamics: The target of the October 2022 controls were advanced microprocessors and memory for products (and technologies) needed to build cutting-edge datacenters for AI models. This market, and the companies serving it, continue to experience growth well above the overall semiconductor industry. Because the impacted companies by definition benefited more strongly from the positive market environment than the non-impacted peers, they could also grow R&D over-proportionally.
2. Concurrent Government Stimulus Programs: Historically, the critical role of semiconductors in national and economic security has resulted in both, export controls and-direct and indirect-government support. The U.S. CHIPS Act, along with similar initiatives in the European Union, Japan, and Korea, exemplifies support for U.S. and allied semiconductor companies. These programs provided significant financial incentives that coincided with the implementation of the recent export controls. While the impact of the incentives might occur in the future, they provided a forward-signaling effect U.S. and allied governments would continue to support innovative industries.
3. Intact Innovation Ecosystems: High-tech companies' ability to innovate is supported by collaborations with the most innovative customers and suppliers. These innovation ecosystems remained intact during the study period and for the companies under consideration. The United States is home to the world's most advanced datacenter hyperscalers and AI companies, offering valuable innovation sources for designers and manufacturers of advanced semiconductors. Furthermore, leading semiconductor manufacturers, such as Taiwan Semiconductor Manufacturing Company Limited, remained key customers for semiconductor equipment manufacturers, allowing continued innovation with essential suppliers.
4. Leading semiconductor companies command significant market shares: Leading advanced semiconductor design, manufacturing, and equipment supply companies benefit from decades of successful innovation and enjoy substantial market shares and profits. Even if these companies' revenues had been impacted-a factor outside the scope of this study-the industry's structure would likely have mitigated any potential negative effects on R&D spending.

These factors-and how they will develop over time-provide important hints as to whether future export control restrictions will have material collateral effects or not.

Policy Recommendations and Future Studies

Innovation is the lifeblood of the semiconductor industry. The industry, and by extension national security as well as commercial applications, have benefited like few others from international cooperation and access to global markets. Semiconductor innovation is key to a digitalized and decarbonized future.

The full, long-term impact of limiting market access and diverging technological spheres on semiconductor companies remains to be fully understood. A comprehensive net assessment of technology ecosystems must not only be analytically rigorous but also consider respective government actions to promote domestic industries. Part of this comprehensive assessment should explore secondary effects, such as whether a market share loss ("design-out") of U.S. and allied semiconductor companies is indirectly caused by export controls or simply a result of China's concurrent efforts to establish domestic supply chains for critical technologies.

The findings of this study should not be interpreted as an endorsement for policymakers to indiscriminately expand export controls. The implications of such measures necessitate careful consideration and strategic implementation to avoid unintended consequences.

In invoking export controls, the government should continue to focus on clear, narrowly defined national security rationales and transparently communicate those to market participants. This is especially important since barriers that would affect commercial success and innovation in the semiconductor industry also impact progress towards shared global responsibilities such as electrification and decarbonization: the national security prerogative should be made clear, and the trade-offs need to be at least understood if not accepted.

Understanding the market dynamics and competitive situation of affected industries is paramount to target controls, and also to conceive, if needed, supportive industrial policy. Given the breadth of responsibilities of export control agencies in the U.S. and allied countries, governments need to staff them sufficiently, and ensure direct, regular information exchange with industry participants. For these exchanges to be impactful and fact-based, the governments should conduct regular economic impact analyses of their export controls and build an independent knowledge base of market and competitive dynamics.

Multilateral export controls are more effective than unilateral controls, and the United States should continue to work with allies on an aligned approach, preferably in a suitable overall framework. In addition to being more effective, a multilateral approach also ensures a level playing field among U.S. and allied companies.

_{¹Revenue and market capitalization, while important financial metrics, have a lower correlation with a company's ability to innovate revenue, if profitable, could e.g. be used to finance share buy-back, dividends, executive compensation, or marketing expenses.}

Andreas Schumacher is a visiting technology fellow in the Economic Security and Technology Department and the Scholl Chair in International Business at the Center for Strategic and International Studies in Washington, D.C.

The author is grateful to William A. Reinsch, senior fellow and Scholl Chair in International Business, and David Peng, senior fellow for data science, both at the Center for Strategic and International Studies, for valuable discussions and suggestions on this paper. The author also gratefully acknowledges support on data analytics by Yun Tan from the National University of Singapore.

Programs & Projects