How America Can Stay Ahead of China in the AI Race

In October 2022, the United States imposed sweeping restrictions on the export of advanced chips and chip-making equipment. The move, which followed steps taken in the first Trump administration to curtail China’s access to cutting-edge semiconductors, jump-started a comprehensive effort to cut China off from the world’s most powerful computer processors—the inputs required to develop and run sophisticated artificial intelligence systems that could be used to power autonomous weapons, conduct mass cyberattacks, and augment intelligence collection.

Officials knew that U.S. export controls could not succeed on their own. The semiconductor supply chain is global, and countries such as Japan, the Netherlands, and South Korea produce equipment, materials, and components that are key to chip production. Truly hindering China’s progress would require persuading those countries to act in concert with the United States. Officials in the first Trump administration successfully pressured the Dutch to stop selling its most advanced chip-making equipment to China. Building on this effort, the Biden administration began a systematic diplomatic push to curtail China’s access to advanced technologies.

This approach has obvious upsides. Other countries produce semiconductor technologies that would greatly enhance China’s ability to manufacture advanced chips. The Dutch firm ASML, for example, is the only company in the world currently able to build the extreme-ultraviolet lithography machines needed to print intricate designs on the most advanced chips. Plurilateral controls also prevent Chinese firms from simply replacing U.S. equipment with, say, Dutch or Japanese alternatives. And working with allies and partners fit with the Biden administration’s overall preference for diplomacy over unilateralism.

The strategy achieved notable successes: the Dutch and the Japanese announced their own semiconductor export controls in 2023 and again early this year. But there were downsides, too. Coordinating controls takes time and requires difficult compromises, and in the meantime China was able to stockpile large quantities of advanced chips and equipment.

It can be tempting to jettison diplomacy in favor of a more aggressive, unilateral approach, especially amid concerns about a shrinking U.S. lead in the AI competition with China. That President Donald Trump’s second administration has generally favored unilateral action makes such a move even more likely. But the United States should be realistic about the benefits and the substantial risks of going it alone. Expanding extraterritorial U.S. restrictions on selling materials and equipment to China may help Washington maintain a technological edge over Beijing in the short term. It may also help level the playing field between U.S. technology firms and their foreign competitors. But the efficacy of these actions is contingent on the United States’ technological superiority, which may decline over time, and on its diplomatic leverage, which may well be eroding. Working with foreign partners to restrict technology exports has made a real dent in China’s capabilities. For the United States and its partners to successfully limit Beijing’s access to advanced technologies, Washington must find a way to keep some measure of diplomacy alive.

OFF LIMITS

Export control diplomacy is not easy. Foreign countries, including close U.S. partners, can be reluctant to restrict their companies’ access to Chinese markets, alienate Beijing, or craft novel regulations at the request of U.S. diplomats. At the best of times, coordinating controls is a lengthy process, requiring consensus on both the overall strategic approach and minute technical details. This makes it harder for policymakers to adapt nimbly to technological advances and provides China with lead time to stockpile key chips and equipment.

These concerns are not just academic. Anticipating tighter restrictions, China quadrupled its imports of Dutch lithography machines in 2023, before new controls went fully into effect. In the first seven months of 2024, as the United States and foreign partners considered whether and how to tighten controls, China imported roughly $26 billion worth of chip-making equipment, a record amount. This was a boon for firms such as ASML, Tokyo Electron, and the American company Applied Materials, but a significant blow to the efficacy of U.S. export control policy.

What is more, when it comes to China, other countries tend to be more risk-averse than the United States, especially because of Beijing’s willingness to retaliate in response to semiconductor export controls—for example, by threatening to cut off Japan’s access to minerals used in the production of Japanese automobiles. In addition, other countries often wish to satisfy a higher evidentiary burden before acting. The United States may thus be forced to choose between doing more on its own—which disadvantages U.S. firms by denying them the kind of access to Chinese markets afforded to foreign competitors—or “leveling down” controls to whatever the most conservative country is willing to do. In practice, this might mean blacklisting fewer Chinese chip-manufacturing facilities, restricting a smaller subset of advanced equipment, or sidestepping whole categories of controls.

In recent years, U.S. policymakers, frustrated with the drawbacks to diplomacy but nonetheless cognizant of the need to limit access to foreign-made semiconductor technologies, have landed on a little-known export control authority called the foreign direct product rule, which allows Washington to restrict the sale of any product made with U.S. technology. The rule was first introduced in 1959, and for decades was used sparingly. Starting in 2020, however, the U.S. government began exercising it to more aggressively control foreign-made items, including products destined for Russia, Belarus, and the Chinese telecommunications firm Huawei, as well as chips and semiconductor manufacturing equipment.

In December 2024, the Department of Commerce dramatically expanded the applicability of the rule to encompass foreign-made equipment that contains even one chip manufactured with a U.S. tool—a criterion met by virtually all advanced chip-making equipment in the world. Now the United States can unilaterally impose extraterritorial controls on foreign-made chips and equipment in order to block China from alternative markets and promote a level playing field for U.S. firms, all while bypassing lengthy and potentially unsuccessful attempts at diplomacy with third countries. Notably, however, the 2024 rule exempted tools made in a list of countries that participate in multilateral export control regimes and have the authority to control semiconductor technology, among them Japan and the Netherlands. The theory was that these countries could and would impose their own equivalent controls. Indeed, Japan and the Netherlands published complementary controls in early 2025.

HELPING HANDS

Washington is now at an inflection point on export controls. In cases in which American companies are the ones most directly regulated—for example, as part of restrictions on U.S.-designed AI chips, including potential new controls on Nvidia’s H20—unilateral action is a relatively simple proposition. But when foreign sales are more directly affected, such as in the case of restrictions on semiconductor tools, the situation is more complex. Here, the new administration must decide whether to go it alone or to recommit to diplomacy and continue to exempt countries such as Japan and the Netherlands from extraterritorial controls. Neither option is perfect, and both entail significant tradeoffs.

Those who favor expanded extraterritorial controls on close partners argue that they are the only way to truly constrain China’s ability to produce advanced chips and maintain a level playing field for U.S. firms. Even countries that have imposed their own export controls have not matched Washington’s comprehensive restrictions, leaving U.S. firms at a distinct disadvantage. Japan and the Netherlands, for example, have prohibited the export of certain listed tools but have not imposed the kind of end-use, end-user, or person-based controls that Washington has. This means that although there are no longer any U.S. items being sent to advanced Chinese chip-manufacturing facilities, or any American workers there, Dutch and Japanese firms can still send unlisted items to those factories, including some key parts and components, and provide some upgrade and maintenance services on the ground. Imposing extraterritorial controls could close this gap.

In addition, proponents of expanded extraterritorial restrictions argue that export control diplomacy has been a failed experiment costing the United States time and effort as it cedes ground to China. Freed from the obligation to align and coordinate export controls with foreign partners, the United States could publish export control updates more regularly, keeping better pace with technological developments in China. It would also be easier for the United States to pursue more aggressive controls, including blacklisting additional Chinese manufacturing facilities or restricting more tools.

Yet the failure of plurilateral controls can be overstated. If one of their primary goals is to degrade China’s ability to manufacture chips to develop and use advanced AI, then the policy has had some real success. While U.S. companies have planned massive data-center buildouts for millions of U.S.-designed chips to create increasingly powerful AI systems and deploy them at scale, China has struggled to produce even a small fraction of U.S. output—and what it has made has cost more and been of lower quality. China does not appear to have indigenously manufactured any of Huawei’s AI chips, and even the Chinese AI firm DeepSeek, the country’s biggest AI success story, has freely admitted that access to computing power is its greatest constraint.

Expanded unilateral extraterritorial controls, meanwhile, are not a panacea. For one thing, the foreign direct product rule can extend Washington’s jurisdiction only to where there is a U.S. technology hook—for example, the presence of a U.S. chip or a foreign chip made with U.S. technology. It will be less effective where the U.S. input is generic and therefore easily replaceable, as is the case for some semiconductor materials and components. In the longer term, expanded use of the rule could further incentivize foreign companies to reduce their reliance on American parts throughout the semiconductor supply chain. Ultimately, this could leave the United States with less leverage as AI becomes more important and more powerful and make it harder for the United States to cut Beijing off from advanced technologies in the event of a Chinese invasion of Taiwan.

U.S. leverage is unlikely to last forever.

Extraterritorial controls can also be a diplomatic nightmare. Some countries could be persuaded to accept U.S. jurisdiction to deflect blame from China, and where that is the case Washington can seek buy-in and move forward with restrictions. But many understandably balk at the perceived loss of sovereignty that comes with U.S. control over their firms. Countries such as Japan and the Netherlands are also sensitive to U.S. actions that could harm companies that play an outsize role in their economies and that they are wary of alienating. And with an underresourced enforcement team, the Department of Commerce cannot truly compel foreign companies to comply, especially if their governments direct them to ignore U.S. diktats.

The full diplomatic costs of extraterritorial action are hard to predict. In addition to choosing not to comply, countries could decide to retaliate, roll back existing controls, or reduce cooperation with the United States on other issues, including those outside the technology domain. Whether countries would risk provoking Trump with an aggressive response is an open question—and likely depends on how much leverage the United States has, both diplomatically and technologically, and for how long.

What is clear, however, is that this leverage is unlikely to last forever. In the long run, foreign companies may find ways to reduce or eliminate their reliance on the United States, which means that tools such as the foreign direct product rule could eventually lose their bite. The United States may therefore wish to exhaust diplomatic options before resorting to unilateral action. Unfortunately, if Trump goes through with even a fraction of the global tariffs he announced in early April, it will likely decrease the administration’s leverage to extract concessions on technology controls and could undermine ongoing cooperation. In the wake of the announcement, for example, ASML’s shares dropped to their lowest level in more than a year, a development that is unlikely to make the company or the Dutch government sympathetic to calls for further restrictions. The Trump administration’s more confrontational approach to traditional allies, particularly those in Europe, will undoubtedly color attempts at export control diplomacy, and some European states may be reluctant to concede too much, given the risks of domestic political blowback.

Still, it is worth trying to find common ground, and sometimes the United States will need to make the first move to apply sufficient pressure on partners to “level up” their own controls. Allies, meanwhile, can lower the perceived costs of diplomacy by signaling their willingness to work fast to close gaps in current controls. The ideal terms of such an offer may change depending on the circumstances but could include commitments to regularly blacklist Chinese manufacturing facilities, restrict more tools, ban onsite maintenance activities, and share intelligence on equipment locations and production capacity in China.

Even if Washington ultimately decides to expand its use of the foreign direct product rule, there will be long-term value in continuing to engage in some form of export control diplomacy: doing the long, arduous, sometimes frustrating, but strategically rewarding work of persuading countries that it is in the collective interest of the United States and its allies to act together to maintain a technological lead over China.

Loading...