Strategic Misalignments: Rethinking US Tactics in the Semiconductor Arena

Mr. Michael Frank
CEO and Founder, Seldon Strategies
Non-resident Fellow, 2430 Group

Download this publication (PDF)

Publications Strategic Misalignments: Rethinking US Tactics in the Semiconductor Arena

Download Strategic Misalignments: Rethinking US Tactics in the Semiconductor Arena

*This paper (published May 8, 2024) is derived from a presentation the author gave at a NEXT Alliance Conference workshop on March 1-4, 2024, in Tokyo. Mr. Michael Frank wrote in his personal capacity. The view and interpretations expressed by the author are solely his own.


One of the great ironies of international politics in recent years is that, of all the actions the American and Chinese governments have taken relevant to artificial intelligence (AI), US semiconductor export controls may turn out to be the single greatest policy tailwind for China’s AI capability, while the Chinese government’s actions have the opposite effect. In this piece, I argue that the United States should scale back the semiconductor manufacturing export control regime to align with Japan and the Netherlands, and allow the export of all but the industry state-of-the-art to China. Such a move—taken in conjunction with more effective measures—would enhance America’s position in its competition with China.

Semiconductor export controls were a wise plan, in theory

National Security Advisor Jake Sullivan’s “small yard, high fence” philosophy has become a matter of faith among the latest generation of Washington China hawks. It appears theoretically sound. The top priority for US policymakers is to avoid losing a great power war that alters the global balance of power to the United States’ detriment. Their greatest fear is that US technology would be deployed against US and allied forces in such a conflict. Speaking at the Brookings Institution in April 2023, Sullivan reiterated that the underlying goal behind many of the Biden administration’s China policies reflect “straightforward national security concerns… We are simply ensuring that US and allied technology is not used against us.”[1] Secretary of State Antony Blinken made a similar point two years earlier, stating that forthcoming policies on export controls, investment screening and visa rules were designed so that “strategic competitors are not exploiting our own innovative ecosystems to gain military or national security advantage.”[2]

The dominant policy in service of the “small yard, high fence” is the semiconductor export control regime, first announced on October 7, 2022 and expanded in October 2023.[3] [4] The export controls represented a fundamental change in how export controls are applied, shifting from restrictions on specific end-users to restrictions on entire countries. Their intended goal: to preclude China from producing or acquiring leading-edge semiconductors that can be deployed to train and run AI models. However, that fundamental change revealed an unfortunate truth about the efficacy of export controls. The justification for a geographic description, rather than end-user restrictions on firms affiliated with China’s military, is based on the Bureau of Industry and Security’s (BIS) own admission that it cannot verify whether AI hardware is going to civil or military use: “China’s military-civil fusion effort makes it more difficult to tell which items are made for restricted end uses, thereby diminishing the effect of these existing controls.”[5] That admission highlights how BIS has been given a near-impossible monitoring and enforcement mission, despite a relatively modest $223m enforcement budget and fewer than a dozen law enforcement personnel stationed overseas.[6]

Export controls are porous by nature. While companies domiciled in allied countries dominate the global semiconductor value chain, China is able to acquire chips, equipment, and know-how through both licit and illicit means. Chinese AI researchers affirm they can overcome the most detrimental effects of the export controls thanks to stockpiles of advanced chips before the controls went into effect; accessing chips through the cloud in uncontrolled jurisdictions; and efficiency improvements elsewhere in the AI value chain (such as training algorithms). Chinese AI developers are global firms with offices all over the world and robust relationships with international cloud providers like AWS, Microsoft, and Oracle. Huawei and SMIC, two leading companies in China’s semiconductor manufacturing ecosystem, now have a captive market from which they can draw large and growing revenues to finance advances in their own design and manufacturing capabilities.

The equipment export control policy depends on the coordinated unilateral action of the United States, Japan and the Netherlands, three countries with collectively dominant market shares in semiconductor manufacturing categories such as advanced lithography, deposition tools, and design software. However, there are gaps in that coordination, and dominance is not the same thing as monopoly control. The Japanese controls did not go into effect until nine months after the October 7th US policy was announced; the Dutch controls lagged an additional six weeks. Other important players in semiconductor value chains, such as South Korea and Germany, have not joined the policy. Key American allies are opposed to further steps to strangle China’s semiconductor ecosystem through export controls.[7] Many of their firms continue to sell into China while their American competitors are barred.[8] A recent report from the Federal Reserve Bank of New York found that the export controls have not induced reshoring or friend-shoring, while wiping out $130bn in the affected American firms’ market capitalization.[9] Finally, nothing is stopping Chinese entities from tapping enormous state support to hire away talent from foreign firms.

In practice, the export controls are harming US national security

These points all underscore the export controls’ limited effectiveness. Reality is actually worse: not only are the export controls ineffective, but they are more likely to advance China’s AI capabilities relative to the United States, rather than hinder them.

To start, recent US export controls undermine implementation of the CHIPS Act, which is helping to reshore advanced semiconductor manufacturing in the United States. The vast majority of CHIPS Act-related spending is actually private money. The Commerce Department is, wisely, using its $52.7bn in direct incentives to augment private spending on the margins, helping to internalize the externality of vulnerable supply chains that have over-optimized for efficiency versus resilience over the past few decades. Industry insiders affirm that the CHIPS Act’s 25 percent investment tax credit will ultimately amount to a far greater sum than the grants and subsidies over the life of the program, given how much private companies are investing. Any firms that take incentives are prohibited from expanding operations in China. As of April 2024, that list includes industry giants TSMC, Intel, GlobalFoundries, Samsung, and BAE Systems.[10]

The semiconductor industry has argued that they need revenues from their China business to finance R&D and capital investment. With the requirements of the CHIPS Act, that is not an esoteric argument. Any revenue these firms make in China cannot go towards expanding their operations. Every dollar of profit in China is either returned to shareholders or invested anywhere but China. The CHIPS Act has helped to rectify misaligned incentives for firms’ China business. That means the controls are actually shutting off a vital source of industry capital to expand on the already-announced investments associated with the CHIPS Act. If the controls were effective, this argument would be moot: the national security trade-off would clearly justify forgoing revenue from the China semiconductor market. Alas, that is not the current situation.

Making matters worse, the export controls have delivered perennial nuisances Huawei and SMIC a captive market. Until recently, both firms were near death, but now they are quite alive. Huawei’s profit more than doubled in 2023. Huawei’s current leading data center chip (which SMIC manufactures), the Ascend 910B, is superior to Nvidia’s A100, the generation immediately preceding the H100, which is arguably the current industry leader. US export controls apply to the A100 (and at least one generation prior), which means Nvidia is not competitive in China and yet AI researchers have access to Nvidia-level AI hardware. Without the export controls, it is conceivable Nvidia could compete with Huawei in their own backyard without meaningfully augmenting the technical capabilities of Chinese AI researchers. Instead, Huawei is racking up the profit from Chinese hyperscalers and investing into ever more rapid R&D advances.

This is an urgent national security challenge. The export controls are on the verge of costing the US advanced semiconductor ecosystem huge sums of money that can only be re-invested outside of China in exchange for a potential degradation of Chinese AI researchers’ capabilities that does not appear to be the practical reality.

Strategic competition calls for a more outcome-oriented approach

The export control policy does not exist as a goal unto itself. The “small yard, high fence” is intended to buoy the US position in a peer competition with China. Even that goal remains vaguely defined. What is the specific outcome we are seeking to effect or prevent?

The overarching goal of US foreign policy should be the preservation of the liberal international order, which is responsible for the greatest period of great power peace and prosperity in modern history. However, it is clear that the United States must remain the preeminent power in the system for the order to endure. Only China has the potential to supplant the United States, and its leadership has been clear in its intention to do exactly that.

With this in mind, the top priority for the United States for the foreseeable future should be preventing a military takeover of Taiwan, which would violate the entire security architecture in Asia and end the post-World War II status quo between regional powers. The United States and its allies would then be forced to choose to fight a full-scale, intercontinental war or let the entire regional (and possibly global) order collapse. Given that choice, the United States will almost assuredly choose the former. However, the costs would be horrific. That is why the United States must build deterrence through overwhelming evidence that China is incapable of winning a war over Taiwan, and therefore should not attempt it.

Few people in the US foreign policy establishment are willing to publicly articulate this view. They are rightly concerned that China will see it as a provocation and America’s interest in Taiwan as a violation of its core interests. In that sense, the export controls actually prioritize risk aversion and a misunderstanding of strategic stability over advancing strategic objectives—rather than saying the quiet part out loud, we will build deterrence by handicapping China’s AI development, therefore undermining its power projection capability and underlying economic base.

Unfortunately, the idea that refraining from articulating the Taiwan message enhances stability is completely backwards. A hostile, expansionist China that has deconstructed the Asian security architecture and controls the most strategically important island in the first island chain is unacceptable to the United States. It is unacceptable to Japan and Australia. There will be a great conflagration if China attempts to subjugate or conquer Taiwan. The United States can acknowledge China’s position while also asserting that Taiwan represents a core American and allied interest as well. Strategic ambiguity is dead. It died when the Chinese government removed the word ‘peaceful’ from its references to reunification with Taiwan.[11]

Working backwards from that goal helps to craft a more effective approach to the semiconductor industry. US policy should retain a ruthless commitment to enhancing the allied balance of power in East Asia, which reflects not just China’s AI today but also its future capability, as well as the allies’ relative technological, economic, and military advantage.  The complexity of semiconductor manufacturing and dominance of firms domiciled in allied countries has made control an alluring priority. Alas, US appeals to allies to take restrictions further have fallen on deaf ears. The United States should accept this reality and roll back export controls on semiconductor manufacturing equipment to align with the Netherlands and Japan so that US firms are not disadvantaged relative to their Dutch and Japanese competitors.

The Commerce Department should continue to monitor the state-of-the-art for AI chips and semiconductor manufacturing equipment, operating under a presumption of denial to China for only the latest generation as long as Chinese alternatives are incapable of providing a similar level of performance. For their part, the semiconductor industry should still take appropriate know-your-customer (KYC) steps and prioritize filling orders outside of China when reasonable. These steps would undo the damage to American firms’ competitive position with respect to their international peers and bring more capital into CHIPS Act-related investments.

That doesn’t change the fact that improving our semiconductor competitiveness only indirectly supports the Taiwan goal. As Jim Lewis points out in his recent piece, America is Losing the Shoe Race With China, American predictions of technological competition typically have not “place[d] their analyses in the context of larger national economies.[12] Instead, they relied on picking illustrative metrics, usually proxy indicators that provide an indirect measurement of technological success.” Measuring semiconductors is a means to an end. People can disagree with the assertion that preventing the hostile takeover of Taiwan should be that end goal, but obsessing over indirect measures without addressing the “context of larger national economies” is a recipe for ineffective or counterproductive policy.

As I alluded to in the first paragraph, more effective policies would be those that reinforce the Chinese government’s proclivity for killing the golden goose: scaring off foreign investment; subjugating its most innovative firms; reclaiming the state’s role in resource allocation at the expense of the market; imposing burdensome AI regulation; and instilling paranoia that a foreign agent lurks behind every corner. Designing those policies requires re-engaging some of the China scholars who have been disregarded as doves. They understand the Chinese government and what buttons to press to elicit those kinds of self-destructive reactions.

Perhaps China will surprise the world and conclude it is waging a quixotic struggle for an outdated definition of national glory and return to the trend of liberal international engagement that served it so well from 1978 to 2012. Regardless, Washington’s China hawks should consider an evolution of the semiconductor export controls into policies that more directly and effectively addresses the balance of power.

The US-Japan NEXT Alliance Initiative is a forum for bilateral dialogue, networking, and the development of joint recommendations involving a wide range of policy and technical specialists (in and out of government) to stimulate new alliance connections across foreign, security, and technology policy areas. Established by Sasakawa Peace Foundation USA with support from the Nippon Foundation, the goal is to help improve the alliance and how it serves shared interests, preparing it for emerging challenges within an increasingly complex and dynamic geostrategic environment. Launched in 2021, the Initiative includes two overlapping lines of effort: 1) Foreign & Security Policy, and 2) Technology & Innovation Connections. The Initiative is led by Sr. Director Jim Schoff.

[1] “Remarks by National Security Advisor Jake Sullivan on Renewing American Economic Leadership at the Brookings Institution.” The White House, The United States Government, 27 Apr. 2023,

[2] Secretary Antony J. Blinken at the National Security Commission on Artificial Intelligence’s (NSCAI) Global Emerging Technology Summit – United States Department of State, Accessed 7 May 2024.

[3] “U.S. Imposes Additional Export Controls Restrictions on Advanced Computing and Semiconductor Manufacturing Items.” Covington & Burling LLP, Accessed 7 May 2024.

[4] Reinsch, William Alan, et al. “Insight into the U.S. Semiconductor Export Controls Update.” CSIS, Accessed 7 May 2024.

[5] “The Federal Register.” Federal Register:: Request Access, Accessed 7 May 2024.

[6] “BIS Seeks 17% Budget Hike in FY 2025.” Export Compliance Daily, 12 Mar. 2024,,million%2C%20the%20Commerce%20Department%20said.

[7] Hawkins, Mackenzie, et al. “US Faces Pushback by Netherlands, Japan on More China Chip Export Bans.” Bloomberg.Com, Bloomberg, 12 Apr. 2024,

[8] Satoh, Ryohtaroh. “U.S. Trade Curbs Spur China Business for Japan Chip Industry.” Nikkei Asia, Nikkei Asia, 20 Feb. 2024,

[9] Crosignani, Matteo, et al. “Geopolitical Risk and Decoupling Evidence from US Export Controls.” New York Fed, Federal Reserve Bank of New York, Apr. 2024,

[10] “Funding Updates.” NIST, 26 Apr. 2024,

[11] “Chinese Government Drops References to ‘peaceful’ Taiwan Reunification.” South China Morning Post, 25 May 2020,

[12] “America Is Losing the Shoe Race with China: Working Papers.” CSIS, Accessed 7 May 2024.

Download Strategic Misalignments: Rethinking US Tactics in the Semiconductor Arena

2024 Sasakawa USA | Privacy Policy | Sitemap

Custom WordPress Design, Development & Digital Marketing by time4design