Silicon and statecraft
Semiconductors as security: great-power competition and the politics of AI chip export controls
Server racks at NOIRLab’s headquarters in Tucson, Arizona. Photo: NOIRLab/NSF/AURA/T. Slovinský, CC BY 4.0
The world is going through a technological revolution, the integration of AI into our everyday lives is staggering with OpenAI alone being the fastest growing technology company in history to reach 900 million weekly users (OpenAI, 2026). Yet beneath the promise of AI-driven efficiency lies a more consequential dynamic, the race toward artificial general intelligence (AGI) is inseparable from a fierce geopolitical competition over the hardware that makes it possible. This competition is most visibly expressed through a succession of export controls on high-end semiconductors, from the Biden administration's October 2022 restrictions on Nvidia H100 chips to the Trump administration's subsequent ban on H200 exports to China (Stone, 2025, p.454). Yet despite these bans, companies both American and Chinese plough on relentlessly towards this Pandora’s box, with the logic of if we do not act, someone else will (The Economist, 2025).
This essay argues that US-China competition over AI chips is a security issue best understood through Waltz’s lens of structural realism, which argues that the anarchic structure of the international system compels states into competitive self-help behaviour regardless of the costs (Waltz, 2000, p.5). However, realism alone is not enough and produces an incomplete and potentially dangerous response. As Keohane and Nye (2012, p.31) argue, the goals of actors in world politics cannot be understood purely through a military lens, transnational actors pursue varied interests across multiple channels, and complex interdependence has fundamentally reshaped how power operates. The essay proceeds in three parts, firstly establishing how chip competition constitutes a security matter, secondly deepening our analysis through competing theoretical frameworks and finally evaluating what a more adequate policy response might look like.
In order to understand why AI chip competition constitutes a security issue, it is first necessary to establish what security means. As Waltz (1979, p.105) argues, international politics is carried out in a self-help system, where states must exert effort not only towards their own growth but also in guaranteeing their own protection against rivals whose intentions can never be fully known. Applied to the semiconductor competition, this framework reveals why advanced AI chips have become a matter of urgent national security concern. Eric Schmidt, former Chairman of the National Security Commission on AI, identifies several critical ways in which AI capability reshapes the security environment. Most significantly, he argues that AI-enabled military and intelligence capabilities may disrupt the fundamental premises of nuclear deterrence in ways that undermine strategic stability (Schmidt, 2022, p.289). He further warns that within the next decade, China could surpass the US as the world's preeminent AI power (Schmidt, 2022, p.296) a development that would represent precisely the kind of shift in capabilities that Waltzian logic predicts states will take extreme measures to prevent. This concern is further exacerbated by supply chain issues as the US relies on importing cutting-edge semiconductors, especially from Taiwanese factories creating a critical infrastructure chokepoint (Schmidt, 2022, p.297). The 2022 National Defence Strategy makes this framing official, identifying China as the overall pacing challenge for the US military and highlighting the importance of technological superiority to American security strategy (US Department of Defence, 2022, p.4).
Mearsheimer's offensive realism sharpens this argument considerably. Where Waltz argues states seek survival, Mearsheimer contends that the international system creates powerful incentives for states to look for opportunities to gain power at the expense of rivals (Mearsheimer, 2014, p.35). From this perspective, American chip controls represent not a defensive measure but an offensive act, an attempt to entrench technological superiority before China can close the capability gap. Critically, Mearsheimer's analysis of US-China competition also reveals why broad economic decoupling is not a viable instrument of containment. He argues that there is no practical way of slowing the Chinese economy without also damaging the American economy, as countries around the world would eagerly fill the vacuum created by Washington's withdrawal (Mearsheimer, 2014, p.401). This logic explains why, despite the disruption caused by the Liberation Day tariffs, the US has not pursued wholesale separation from China, but has instead focused on specific areas, sanctioning Chinese firms such as SMIC over concerns about civilian-military technology fusion, and restricting the advanced computing components required for leading AI models (Steinfeld, 2025, pp.462-463).
However, while structural realism provides a compelling reason as to why states treat chip dominance as a security risk, it faces empirical limitations due to its core assumptions. For Waltz, the international realm is one defined by ‘relations of strength’ (Waltz, 1979, p.113) a framework that treats military capability as the enforcer behind all international relations. Yet the semiconductor competition fundamentally challenges this state-centric notion. Examining the structure of this industry reveals the highly complex nature of it as no single country can manufacture and design the whole chip. Instead, we see a web of interdependent companies from different parts of the process with companies like ARM dominating the architecture, Taiwan Semiconductor Manufacturing Company dominating fabrication and ASML building the lithography equipment, an essential factory component (Jia et al., 2025, p. 7). These companies are not simply instruments of state power, but instead independent actors whose allegiance to shareholders often outweighs their loyalty to national security, as shown by the recent case of Super Micro Computer Inc illegally smuggling H100 and H200 chips to China (Bloomberg, 2026).
Contrasting with realist vision in which military security outweighs all other goals, Keohane and Nye (2012, p.31) present a view where the state’s goals are not a singular fixed idea, but instead fluid with transnational actors pursuing their agendas. Crucially, they take a more nuanced stance with the idea of power resources, stating that they are not easily translated across effective fields as military capabilities are not effectively translated to economic issues (Ibid, p.43). Washington faces precisely this struggle, as while being capable of bringing superior military might to bear, this cannot be translated into stopping China from competing in the AI arms race. This is further complicated by the conflicting role that multinationals play, simultaneously increasing the appearance of US influence while multiplying the social channels between societies in ways that resist state direction (Keohane and Nye, 2012, p.42). Indeed, as Ibrahim (2025, p.287) observes, despite geopolitical tensions, the American and Chinese AI ecosystems remain deeply interconnected through supply chains and academic collaboration, a factor that is heavily overlooked by neorealism.
This tension is sharpened by an ideological contradiction at the heart of the US position. As Wu and Davidson observe, in the US the foundational assumption is that businesses operate independently in the free market system, whereas historically in China, Confucian tradition emphasises a closer relationship between state and business in service of the greater society (Wu and Davidson, 2011, p.113). This divergence is further illustrated by contrasting governance models, while the American approach emphasises privacy, civil liberties, and open markets, China's model reflects collectivism and social harmony (Ibrahim, 2025, p.290). This creates a profound legitimacy problem for Washington, which is attempting to dictate to companies such as Nvidia and ASML which markets they may serve and which technologies they may sell, while simultaneously positioning itself as the defender of free market principles against Chinese state capitalism. Furthermore, Ibrahim (2025, p.286) identifies a structural security dilemma at the heart of US restrictions intended as defensive measures are perceived by China as offensive containment, prompting their own technological militarization as a means to securing long-term security. There is a further economic contradiction: AI contributed 2.51% to US GDP growth in the first nine months of 2025 (Federal Reserve Bank of St. Louis, 2026), meaning the key driver of American economic growth is very reliant on the very AI spending that export controls and tariffs disrupt. If one follows Mearsheimer's own logic that economic strength is a precursor to military strength (Mearsheimer, 2014, p.17) then weakening the US economically will only embolden China militarily.
Instead, a more prudent policy would be to offer older chips such as the H200 and H100 as a bargaining tool with China in order to gain concessions for how the development of AI is carried out. This again is an idea Schmidt would endorse, as if AI is that much better at identifying where second-strike nuclear capabilities are then it removes the mutual vulnerability creating a dangerous situation (Schmidt, 2022, p.294). Thus, if a more liberal approach is used to gain concessions from both parties about the implementation of AI into the military it takes advantage of this mutual interdependence, while still being acceptable to the US, by not allowing China access to its next generation of high-tech chips.
The US must retain the competitive logic of structural realism while channelling it through a more sophisticated diplomatic framework than just unilateral export controls. Schmidt correctly identifies that AI-enabled cyber weapons have the capacity to learn and adapt, extending far beyond their intended targets, making uncontrolled AI proliferation a genuine threat to strategic stability (Schmidt, 2022, p.294). However, the record undermines the case for escalatory blanket controls. As Ibrahim (2025, p.289) demonstrates, China has already shown the capacity to adapt around export restrictions, developing the DeepSeek model on stockpiled chips from before the ban, a more cost-efficient equivalent to many US models that suggests it is only a matter of time for the models to catch up (Ibrahim, 2025, p.291). The Super Micro Computer incident further illustrates that a complete crackdown presents an extraordinary enforcement challenge in a system characterised by the very corporate non-compliance that Keohane and Nye's framework anticipates (Bloomberg, 2026). A more theoretically informed response would adopt a middle ground approach, offering older generation chips such as the H100 as a diplomatic instrument, using market access as leverage to secure meaningful concessions on the integration of AI into military systems. Schmidt's own analysis supports this logic, if advanced AI reaches the point of identifying and targeting second-strike nuclear capabilities, it removes the mutual vulnerability underpinning strategic stability, creating incentives for preventive action that both sides have a structural interest in preventing (Schmidt, 2022, p.294). A liberal institutionalist framework, using interdependence as leverage rather than treating it as a vulnerability to be severed offers a more durable path to managing this risk.
Building on this liberalist edge to a realist approach, the US must work with its allies if it wants to control the flow of high-tech components into China. As Jia et al (2025, p.7) point out there is no one country with a complete monopoly over the semiconductor supply chain and companies like TSMC, ASML and SK Hynix are leading edge from countries that are American allies such as Taiwan, Netherlands and South Korea. Rather than threatening allied partners with tariffs, the United States should leverage its two most significant instruments its huge consumer market and its security architecture offering formal defence commitments in exchange for coordinated export limitations on cutting-edge components to China. This does not require abandoning realist logic it requires recognising, as this essay has argued throughout, that realism explains why the competition is unavoidable and that liberalism explains how it must be managed.
The US-China competition over AI chips is a genuine and urgent security issue, rooted in the great power rivalry emphasised by Waltz and Mearsheimer. In the anarchic self-help system, the notion that the US simply lies back and allows China to take advantage of their cutting-edge tech to improve their military capabilities is unthinkable. Yet this two-dimensional approach belies a more complex reality, that being the complex interdependence identified by Keohane and Nye. The semiconductor supply chain does not map onto the state-centric world that Waltz describes, it is transnational, corporate, and deeply entangled in ways that make unilateral containment impossible. This essay has demonstrated that these two frameworks are not competing explanations but operating at different levels; realism reveals why the competition is structurally inevitable, while liberalism reveals why the response must be multilateral, institutionalised, and attentive to the interests of the corporate and allied actors on whom effective strategy ultimately depends. If the US wants to remain the hegemon in this new era, it must recognise that the world is defined by these supply chains, corporate actors and allied partners. As Schmidt (2022, p.298) warns how the US deals with these challenges will shape its place in the world by the end of the decade.
Bibliography
Bloomberg (2026) ‘China AI firm discloses $92 million of banned Nvidia chip servers to Beijing’, Bloomberg, 10 April. Available at: https://www.bloomberg.com/news/articles/2026-04-10/china-ai-firm-discloses-92-million-of-banned-nvidia-chip-servers-to-beijing (Accessed: 12 April 2026).
The Economist (2025) ‘Artificially incautious: AI labs’ all-or-nothing race leaves no time to fuss about safety’, The Economist, 24 July. Available at: https://www.economist.com/briefing/2025/07/24/ai-labs-all-or-nothing-race-leaves-no-time-to-fuss-about-safety (Accessed: Tuesday 7 April 2026).
Federal Reserve Bank of St. Louis (2026) ‘Tracking AI’s contribution to GDP growth’, On the Economy Blog, 12 January. Available at: https://www.stlouisfed.org/on-the-economy/2026/jan/tracking-ai-contribution-gdp-growth (Accessed: 13 April 2026).
Ibrahim, N.T., 2025. The Impact of AI on Global Power Shifts: A Case Study of US-China Rivalry. China Quarterly of International Strategic Studies, 11(03), pp.281-309.
Jia, F. et al. (2025) 'Global semiconductor supply chain: a national competitiveness analysis', International Journal of Production Research [Online first]. Available at: https://doi.org/10.1080/00207543.2025.2568987 (Accessed: 14 April 2026).
Keohane, R.O. and Nye, J.S. (2012) Power and interdependence. 4th edn. Boston: Longman.
Mearsheimer, J.J. (2014) The tragedy of great power politics. Updated edn. New York: W. W. Norton & Company. Available at: ProQuest Ebook Central (Accessed: 8 April 2026).
OpenAI (2026) Accelerating the next phase of AI. Available at: https://openai.com/index/accelerating-the-next-phase-ai/ (Accessed: 7 April 2026)
Schmidt, E. (2022) 'AI, Great Power Competition & National Security', Daedalus, 151(2), pp. 288–298.
Steinfeld, D. (2025) 'Chips, china and choke points: the economic and national security consequences of u.s. semiconductor export control policy', New York University Journal of International Law and Politics, 57(2), pp. 456-484.
Stone, R. (2025) ‘Tariff war takes toll on labs in China’, Science, 388(6746), p. 454.
US Department of Defense (2022) 2022 National Defense Strategy of the United States of America. Available at: https://media.defense.gov/2022/Oct/27/2003103845/-1/-1/1/2022-NATIONAL-DEFENSE-STRATEGY-NPR-MDR.PDF(Accessed: 9 April 2026).
Waltz, K.N. (1979) Theory of International Politics. Reading, MA: Addison-Wesley.
Waltz, K. N. (2000) ‘Structural realism after the Cold War’, International Security, 25(1), pp. 5–41.
Wu, J. and Kirk Davidson, D., 2011. The business‐government‐society relationship: a comparison between China and the US. Journal of Management Development, 30(1), pp.112-125.