By Michael Frank
America’s contest with China over advanced technologies features a compelling irony: the more Washington tightens export controls and induces friction in supply chains, the more it eliminates the leverage it might brandish in a real crisis.[1] American policymakers have targeted the semiconductor and AI hardware pipeline with sweeping embargoes on Chinese entities. The logic is that any advanced chip in Chinese hands represents a threat to American national security.[2] It is a compelling theory. A “small hard, high fence” that denies China cutting-edge AI chips can widen America’s power advantage without undermining regional stability in Asia.[3]
The practical reality has played out differently. China is plowing money into local alternatives, forging its own supply chains, and acquiring the hardware it needs from a global web of illicit trade. While Chinese companies in the AI value chain previously preferred to work with international partners over domestic peers, they now have no alternative but to work together to advance the Chinese Communist Party’s goal of technological self-sufficiency. In the near future, America will lose an invaluable pressure point in a confrontation over Taiwan or any other flashpoint. Nothing about that scenario works in America’s favor.
In contrast, weaponized asymmetric interdependence—a strategy in which one side fosters the adversary’s reliance on certain choke points—would better serve American national security interests. America has routinely criticized China’s efforts to cultivate weaponized asymmetric interdependence and acknowledged their efficacy.[4] And yet, policymakers in Washington have made no efforts to give China a taste of its own medicine. Rather than wish for a world that does not exist—a world in which China abides by liberal international rules—policymakers should embrace the world as it is.
Weaponized asymmetric interdependence refers to an intentional policy of keeping a rival reliant on certain goods or services that can be cut off during a crisis, while reducing one’s own vulnerabilities in that relationship. This approach retains the option of economic coercion or punishment at a moment of maximum strategic advantage. Interdependence is not some academic novelty: Robert Keohane and Joseph Nye wrote decades ago about how economic ties can become sources of coercion.[5] Henry Farrell and Abraham Newman labeled the modern version “weaponized interdependence,” where a state uses global supply chains to degrade an adversary’s strategic options.[6] The logic is straightforward. If America retreats behind an embargo wall, it can no longer deny China advanced chips at a decisive moment. America’s policy should do two things: secure national power and maintain strategic stability.
However, China’s own weaponization of interdependence is a cautionary tale in wielding it too soon. China’s partial oil embargo against Japan in 2010 over the Senkaku (Diaoyu) maritime dispute backfired once Tokyo diversified rare earth suppliers.[7] Xi’s government later tried to coerce Australia, Norway, and South Korea on varied political and trade matters, forcing them to pivot to alternate markets or double down on existing alliances.[8] Chinese policymakers learned that their prey rarely folds under such pressure. America should draw a similar lesson: brandishing these tools too early spurs resilience and undermines conflict deterrence.
Since October 7, 2022, export control policies that intend to leverage allied dominance of particular of the AI hardware value chain— such as electronic design automation (EDA) software, extreme ultraviolet lithography, and deposition—have leveraged asymmetric capabilities in exactly the wrong way. The mission should have been to expand those asymmetric advantages to other parts of the value chain, using coordinated industrial and trade policy among allied countries to reverse Chinese firms’ market share gains in assembly, test and packaging. Instead, the export control regime did nothing to reverse those gains and overreached, assuming it would kill China’s AI ecosystem if it did not have access to allied chips and manufacturing tools. (It did not.)
Huawei now has a captive market in China and other sanctioned markets under the ill-conceived “Framework for Artificial Intelligence Diffusion,” published with five days remaining in the Biden administration.[9] The rule is misguided in its application of chip quotas which are functionally meaningless in AI development but kneecap American firms in global market competition. It also antagonizes law enforcement partners in allied countries, such as Singapore (which, despite its exclusion from a list of trusted partners, nonetheless arrested two of its nationals and a Chinese co-conspirator on February 27 for defrauding American server companies in violation of export control polices).[10]
Cloud computing dominance is another source of weaponized asymmetric interdependence that the US government has failed to exploit. Amazon Web Services, Microsoft Azure, and Google Cloud lead in large-scale orchestration and specialized hardware integration. They house the computing muscle on which AI developers store data, train models and run inference, but they also are the backbone of the modern economy. If Chinese companies—all sectors, not just AI developers—rely on these platforms, it gives America a devastating weapon to deploy if China attempts a blockade of Taiwan. American policy should focus on positioning a sword of Damocles over China’s AI ecosystem and broader economy that only falls if they challenge our interests.
There are other benefits to weaponized asymmetric interdependence that compound America’s AI advantages. Tech executives have been clear that they need to grow revenues to fund advanced research and offset ballooning capital expenditures in new American fabrication plants and cloud computing facilities. Forcing chip and cloud companies to only sell into a small list of markets starves industry of an important revenue stream to invest in American manufacturing and technological infrastructure. Unrestricted chip and cloud sales would free up more money for investment into R&D and new manufacturing under development in Arizona, New York and Texas.
A more aggressive trade policy can also reinforce asymmetric interdependence. The Office of the United States Trade Representative should impose a rising tariff schedule on strategic technology goods from China, boosting it by ten percentage points every year for five years. Allied manufacturers would have new incentives to ramp up output in the US and reduce dependency on China while still empowering them to mount a competitive assault on Huawei, SMIC, and the broader Chinese AI ecosystem. That arrangement introduces the asymmetry America needs: no total ban, but a policy environment in which alternative suppliers proliferate, while China’s allure dims. Many allied companies are also tired of forced technology transfers in China and would be open to a comprehensive arrangement that offers them better market access in America in exchange for tariff compliance.
Critics may charge that limiting broad decoupling and controlling advanced tech exports hands China an opening to outcompete the West. No one denies that Xi Jinping is trying to wean his country off foreign technology. “Dual Circulation,” “Made in China 2025,” and associated campaigns exist precisely for that reason.[11] Herman Kahn’s arguments about escalation show that overly cautious measures can encourage the other side to push its luck.[12] True enough. But a bigger risk comes from squandering the single most powerful economic weapon: the capacity to cut off AI hardware at the moment of maximum leverage. Robert Jervis wrote at length about misperception and how states misjudge each other’s intentions.[13] If America maintains the current policy, it will play a role in China’s growing insulation from the pain of a future embargo—undermining an American deterrent in a crisis over Taiwan.
Reserving that cutoff as a clear—even if quietly signaled—response to a major violation, such as a blockade or invasion of Taiwan, fosters a more stable environment in East Asia. Both sides know the consequences of crossing the line, but neither triggers the punishment prematurely.[14] Clear communication to China’s leaders is essential. Mearsheimer’s studies of deterrence suggest that an uncertain or muddled threat can fail to alter an adversary’s calculus.[15] American officials need to convey to Chinese counterparts that forcibly integrating Taiwan means an immediate embargo on advanced semiconductors and cloud services. China’s leaders might dismiss that threat, but they will have to account for the catastrophic effect on their economy if America delivers on the promise. That might not stop an invasion plan, but it shapes the policy discussion in China. Managing that conversation is better than the alternative, which sees no commercial ties left to sever once war clouds appear.
Grand strategy only works if industry has a seat at the table. Executives at America’s leading companies cannot be treated like rogues given the deep expertise they have about China’s AI ecosystem, supply chains and vulnerabilities. Their companies are critical to building up new fabrication plants on American soil and verifying suspicious sales. The Bureau of Industry and Security’s enforcement division can draft more rigorous guidelines for how firms report suspicious end-uses. If a Chinese customer with potential PLA links tries to buy top-tier chips or GPU clusters, the provider can escalate the request for official review. That arrangement mirrors a “trusted partnership” model. It is not bulletproof, but it is better than implementing rules that freeze American firms out of huge parts of the global semiconductor market and prematurely cede leverage that could forestall a crisis.
Relying on a phased approach to tariffs and partial restrictions carries the risk that China might practice salami tactics in the Taiwan Strait—small provocations that do not cross any bright line.[16] Barry Posen’s work notes that incremental aggression can undermine deterrence.[17] No plan can eliminate that risk. The real question is whether preserving a doomsday sanction on advanced technology gives Chinese leaders pause before opting for an outright blockade or full invasion. If the White House is consistent about threatening a total cutoff at that moment, there is a better shot that Xi Jinping and his advisers see the cost as unacceptable. Weighted against that chance is the near certainty that a maximal embargo right now spurs further Chinese self-reliance.
China’s government acknowledges that advanced chipmaking remains dependent on foreign lithography and EDA software.[18] America can exploit that fact by offering partial access until the day Beijing crosses the line. If the door is slammed shut immediately, all the incentives align for Xi Jinping’s dream of standing up domestic production at scale. That path undercuts the very leverage America might need in a genuine crisis. The threat of weaponized interdependence must be potent and believable when the stakes are highest.[19] America’s task is to keep that threat potent and believable, not to diffuse it prematurely.
Weaponized interdependence requires a deft touch. The basic game plan—preserve a targeted relationship that China still values, reinforce America’s lead in advanced tech, and let Beijing know the penalty for crossing red lines—stands a better chance of boosting American power than the current “whack-a-mole” policy that gets increasingly indefensible with age. America can maintain the readiness to hit the kill switch if a major confrontation erupts. China’s push for self-reliance will continue, of course, but a slower path to autonomy leaves room for misgivings inside the Chinese system. And rather than leaving China’s leaders with no options, weaponized asymmetric interdependence gives them a choice with respect to the global order: live with the status quo, or suffer economic devastation that destroys the gains of the past few decades and puts CCP rule in immediate jeopardy.
[1] U.S. Department of Commerce, Bureau of Industry and Security, “Export Administration Regulations,” 2021.
[2] Henry Farrell and Abraham L. Newman, “Weaponized Interdependence: How Global Economic Networks Shape State Coercion,” International Security 44, no. 1 (2019): 42–79, https://doi.org/10.1162/isec_a_00351.
[3] Jake Sullivan, Speech at the Brookings Institution, April 2023.
[4] Victor D. Cha, “Collective Resilience: Deterring China’s Weaponization of Economic Interdependence,” International Security 48, no. 1 (July 1, 2023): 91–124, https://doi.org/10.1162/isec_a_00465.
[5] Robert O. Keohane and Joseph S. Nye, Power and Interdependence: World Politics in Transition (TBS The Book Service Ltd, 1977).
[6] Farrell and Newman, “Weaponized Interdependence,” 42–79.
[7] Ministry of Economy, Trade and Industry, Japan, “Rare Earth Import Data,” 2010.
[8] James Reilly, “China’s Economic Statecraft,” Asian Survey 50, no. 5 (2010): 922–944.
[9] “Framework for Artificial Intelligence Diffusion,” United States Department of Commerce, Bureau of Industry and Security, January 15, 2025, https://www.federalregister.gov/documents/2025/01/15/2025-00636/framework-for-artificial-intelligence-diffusion.
[10] “Transcript of Media Conference with Mr K Shanmugam, Minister for Home Affairs and Minister for Law, Regarding the Case Involving the Three Men Who Were Charged on 27 February 2025 for Fraud by False Representation,” Singapore Ministry of Home Affairs, March 3, 2025, https://www.mha.gov.sg/mediaroom/speeches/transcript-of-media-conference-with-mr-k-shanmugam-minister-for-home-affairs-and-minister-for-law-regarding-the-case-involving-the-three-men-who-were-charged-on-27-february-2025-for-fraud-by-false-representation.
[11] Ministry of Industry and Information Technology, “Made in China 2025,” (State Council, 2015).
[12] Herman Kahn, Thinking About the Unthinkable (New York: Avon Books, 1962).
[13] Robert Jervis, Perception and Misperception in International Politics (Princeton, NJ: Princeton University Press, 1976).
[14] Thomas C. Schelling, Arms and Influence (New Haven, CT: Yale University Press, 1967).
[15] John J. Mearsheimer, Conventional Deterrence (Ithaca, NY: Cornell University Press, 1983).
[16] Barry R. Posen, Inadvertent Escalation: Conventional War and Nuclear Risks (Ithaca, NY: Cornell University Press, 1991).
[17] Ibid.
[18] Ministry of Industry and Information Technology, “Semiconductor Industry Annual Report,” (State Council, 2021).
[19] Schelling, Arms and Influence.