Taiwan’s AI chip dominance could deter conflict with China—if leveraged wisely to secure defense commitments and maintain its strategic value amid global semiconductor shifts and rapid technological change.
A decade ago, John Mearsheimer’s “Say Goodbye to Taiwan” posited that Taiwan would be forced to cede its sovereignty in the face of China’s powerful rise. He stated that neither nuclear weapons nor a conventional defense would be practical options against China’s vast military and economic rise. At the time, Mearsheimer also argued that the United States, Taiwan’s largest defense supplier, would likely find an effective defense highly costly. While he has since revised his assessment, a fundamental question remains valid today: beyond its strategic position, what does Taiwan offer that would compel the United States and its allies to intervene in a crisis across the Strait?
One potential answer lies in Taiwan’s core position in the semiconductor industry, specifically its ability to fabricate artificial intelligence (AI) chips—drivers of economic growth and military strength. Although much has already been written on the importance of Taiwan’s chip industry, its importance may be understated. Today, Taiwan should not give up its most advanced chip fabrication facilities for the same reason Mearsheimer argued it should not have stopped developing a nuclear weapons program despite US pressure: chip fabrication plants (fabs) located in Taiwan or foreign soil have a deterrent effect.
This article assesses how Taiwan may be able to maintain its security through its leadership position in the AI chip supply chain. It proposes a diplomatic strategy that leverages its top spot in AI chip fabrication to gain defense commitments from allies that would dissuade conflict. Two key assumptions underlie this approach: (1) China will continue to pressure Taiwan for unification, and (2) international efforts to diversify semiconductor facilities away from the island will proceed. Most of this essay focuses on the second point, while the first is mentioned briefly.
AI Chips
Semiconductors for AI are probably the most critical component for developing AI systems. According to scaling laws, training and deploying more powerful frontier AI models requires increasing amounts of computational power, or compute for short. With enough compute, new AI models are becoming more sophisticated and are improving at scientific, mathematical, and coding tasks. Further, who has the most compute may impact the balance of international power. Paraphrasing Dario Amodei, the founder of Anthropic, an AI data center full of geniuses vastly superior to a country’s entire human population and working 24/7, would have profound consequences for global power. That includes security dynamics for the Taiwan Strait.
Some experts forecast that future AI systems could produce a decisive strategic advantage on the battlefield, potentially unseen since the atomic bomb. Such systems might enable missile defenses that could neutralize a first or second strike, or cyber operations capable of crippling an adversary’s national infrastructure. If national leaders believe these uncertain but plausible futures have a chance of coming true, then retaining the world’s most advanced chip plants on Taiwan makes strategic sense, as it binds the trajectory of AI development to the island’s security.
Is the Silicon Shield Dissipating?
Even still, geostrategic pressures to localize fabs are challenging Taiwan’s ability to keep its fabs on its soil. Seeking supply chain resilience, the United States, the European Union (EU), Japan, and others have implemented industrial policies to attract chip facilities. Consequently, Taiwan’s leading firm, Taiwan Semiconductor Manufacturing Company (TSMC), has been building fabs in the United States, Japan, and Germany, while other Taiwanese firms are investing in Singapore and India. Some have raised concerns that these foreign investments are weakening the Silicon Shield, a concept articulated by President Tsai Ing-wen in 2021. It essentially suggests that Taiwan’s position in the chip supply chain prevents China from attacking primarily because of its dependence on Taiwan’s chips.
The shield is unlikely to lose its strength in the fabrication of AI chips globally in the next decade for two reasons. First, fully operationalizing new fabs abroad is a multi-year process. For example, even on an accelerated timetable, TSMC’s Arizona plant will still need to ship chips back to Taiwan for advanced packaging and testing for years to come. Second, despite ongoing supply chain shifts, Taiwan is set to maintain its position at the most advanced chip level through regulations like N-1, which are intended to prohibit the export of chip firms’ most advanced process technologies.
Furthermore, the Semiconductor Industry Association projects that by 2032, Taiwan will possess nearly half of the world’s leading-edge fabrication capacity for chips below the 10-nanometer (nm) process node level—the very chips powering AI. Even though non-AI chips (those above 10 nm) are not the focus of this paper, it is worth mentioning that Taiwan will lose some of its global share, likely to China. That will negatively impact its total share of global chip fabrication as a whole. Regardless, barring major additional government intervention, it is safe to assume that Taiwan will likely maintain its leadership position in AI chip fabs well into the third decade of the twenty-first century.
The AGI Wildcard
While Taiwan’s global share of production capacity for all chips may slowly decrease over time, the intensifying AI racebetween the United States and China introduces a new dynamic. A recent argument suggests that Taiwan’s shield may be turning into a direct target for China. Their hypothesis centers on the race to achieve artificial general intelligence (AGI), often defined as machine intelligence superior to humans at most cognitive tasks. This hypothesis is not new, but the authors’ focus on who gets to AGI first is.
Assuming AGI is only years away, as one AI futurist predicts, the first nation to achieve it could gain a decisive military advantage. Consider, for example, an ability to disable an enemy’s cyber defenses or run highly accurate combat simulations that gives one military a computer network or planning advantage, respectively. As for the race itself, the United States currently holds a lead in total compute. China’s Huawei is projected to produce around 200,000 AI chips in 2025, compared to Nvidia’s ten million or more. However, its lead in frontier model development is estimated at only half a year. AI model progress is increasing in pace, now measured in terms of months. For instance, China had very few open-weight AI models equivalent in quality to the United States in early 2024, but now they are a leading developer of these systems—even if their scaled deployment still remains limited.
So, if the United States were on the verge of an AGI breakthrough around the 2027 timeframe—the year former Indo-Pacific Commander Davidson suggested China’s military would be ready to take Taiwan by—then China may have an incentive to strike Taiwan in order to catch up in the race for AI superiority. According to Bresnick and Powers-Riggs, an invasion could accomplish two goals at once: unifying the island and leveling the AI playing field by cutting off America’s access to AI chips made in Taiwan.
As a result, it should be noted that Beijing’s motivation for unification is fundamentally about politics, and technologists have correctly recognized this. There is plenty of authoritative evidence to support that the Chinese Communist Party’s motivation for unification is related to national rejuvenation. One big unknown, however, is whether party leadership has actually prioritized the importance of AGI, especially as part of its war aims. But the added incentive of impeding the American AI ecosystem cannot be dismissed, given the levels of uncertainty surrounding US-China cooperation on AI demonstrated to date. At a minimum, AGI dynamics contribute another layer of complexity that Taiwan authorities must contend with.
Taiwan’s Chip Diplomacy
Mearsheimer’s 2014 piece did not account for how China’s persistent threat against Taiwan would mix with these economic and technological forces now at play. Indeed, Taiwan faces a trilemma: Beijing’s persistent military threat, a gradually dissipating Silicon Shield due to global economic incentives, and the strategic risks posed by transformational AI technologies. To navigate this, Taipei must implement a coordinated chip diplomacy strategy with its Indo-Pacific partners. The logical partners are the countries best positioned to respond to a Strait crisis: the United States, Japan, Australia, and the Philippines. US bilateral defense ties are already strong, and there is more room for growth between these countries.
The goal of a diplomatic chips strategy is straightforward: Taiwan should offer access to its chip fabrication capabilities in exchange for discreet diplomatic and military concessions. This would involve Taiwanese firms setting up fabs (for five nm to seven nm process nodes) or related research and development (R&D) and training activities in partner territories. To avoid unnecessarily changing the status quo, concessions involving foreign military systems would not have to be located on Taiwan itself during peacetime.
This strategy could be tailored to each partner’s needs and capabilities based on two frameworks: the AI infrastructure (compute, data, AI talent, policy) and its geographical location relative to Taiwan. For the Philippines, given its proximity, it is a key staging area for contingency forces. In exchange for closer defense cooperation on forward-basing equipment and autonomous systems, Taiwan could offer semiconductor workforce training and R&D collaboration.
For Australia, its distance provides relative safety, making it a vital logistical hub for defending sea lanes. To address Australia’s technical talent deficit, Taiwan could offer training and potentially transfer second-generation chip technology.
Lastly, because Japan and the United States are leaders in some segments of the semiconductor supply chain, TSMC’s investments in both countries are a strong start. A key concession could be securing agreements that, in wartime, the output of these new fabs—even those producing chips for automotive or industrial sectors—could be repurposed to power military systems like autonomous vehicles.
Such a strategy creates supply chain redundancy for customers while also gaining concrete military assurances in exchange for fabs. Overall, negotiations should be guided by a partner’s technical capability—across metrics such as compute, talent, data, and policies—in addition to geography, to determine its ability to integrate advanced AI military systems relevant for a Taiwan contingency. As of this writing, Taiwan’s diplomats appear to be taking some initial steps in the right direction by engaging directly with private industry and foreign partners.
Converting Diplomatic Wins to Military Power
Ultimately, successful chip diplomacy must translate into tangible military capabilities designed to counter the People’s Liberation Army (PLA). Specifically, the capabilities must be able to counter the PLA’s anti-access/area-denial (A2/AD) zones. PLA joint exercises around Taiwan have increased significantly over the past several years. Therefore, capabilities should also be oriented to respond to any of the PLA’s joint campaigns—the firepower strike, blockade, or landing. To be effective against PLA operations, allied military capabilities have to feature distributed and autonomous systems that can close the kill chain faster. Using AI chips secured through diplomacy, partners could develop and stage two key capabilities organized around two phases. For simplification, they are pre-conflict (Phase 0) and during-conflict (Phase 1).
Preceding any conflict, allies could host an AI-powered data center dedicated to offensive and defensive cyber operations. Its primary objective would be to disrupt the PLA’s Command, Control, Communications, Computers, Intelligence, Surveillance, and Reconnaissance (C4ISR) networks, which are central to its A2/AD platform. By targeting these nodes and the PLA’s dual-use logistics infrastructure, cyber operations can be employed to slow the PLA’s physical maneuvers, as Joel Wuthnow wrote in chapter ten of Crossing the Strait. Less coercive than kinetic attacks, these operations pose a medium escalation risk and can be integrated across allied territories and run remotely.
During active combat, an autonomous fleet of unmanned aerial, surface, and underwater vehicles (UAVs, USVs, and UUVs) would be deployed from nearby locations like southern Japan or the northern Philippines. The fleet’s mission would be to detect and assault PLA forces involved in a joint campaign. This could involve autonomous missile defense systems to counter a firepower strike or swarms of drones to deny the PLA Navy access to Taiwan’s ports and shores. While these kinetic actions are inherently more escalatory and face logistical hurdles, having the regional capacity to produce and integrate AI chips into these systems is a crucial first step.
Both military capabilities are practical and likely available. American and Taiwanese defense organizations are already taking such actions. Additionally, the latter two are also focusing on the co-production of military systems likely relevant for a crisis scenario. Taiwan needs to take further steps with other countries, such as Japan, Australia, and the Philippines, to enhance allied interoperability in the event of a crisis. If these diplomatic and military exchanges are effectively implemented, Taiwan may be better prepared to counter China’s increasing pressure.
Conclusion
Ten years ago, Mearsheimer wrote that Taiwan had three potential paths for dealing with China’s quest for national rejuvenation. The first path stated Taiwan could decide to obtain nuclear weapons, the second path identified a strategy of conventional defense that would be costly and bloody, and the last pointed to an arrangement similar to Hong Kong. While the first is almost definitely off the table, and the last seems unacceptable to Taiwan citizens based on recent unification polling, defense appears to be the route it has chosen.
First, Taiwan’s fate will be determined not only by its decision to reinforce its defenses but also by the technological and economic crosswinds it is currently experiencing. The Taiwan Trilemma consists of three major risks to Taiwan’s continued existence to different degrees. China’s military threat is the most serious, while a dissipating Silicon Shield will occur over time; uncontrollable dynamics may be produced by the AGI race. Each presents real obstacles to Taiwan’s long-term viability as a self-governing democracy. A diplomatic chip strategy may be able to defend against the first threat. Further research and analysis of the AI infrastructure of Taiwan’s key partners and their ability to integrate AI chips for crisis-ready military applications is necessary.
Lastly, a growing sense of wariness about supporting Taiwan by some traditional allies in the United States should add urgency to Taiwan’s navigation of the trilemma. Taipei must act to convince its partners why it should be defended in the event of a full-blown crisis with Beijing. For the foreseeable future and barring major government intervention from Taipei or Washington, Taiwan will remain home to the leading manufacturer of AI chips. Connecting access to AI chips to AI progress to strategic gains in national power is crucial for Taiwan to promote and solidify allied commitments. After all, those military concessions could be the difference between a thriving democratic Taiwan in the next twenty-five years or something worse.
About the Author: Seth Poling
Seth Poling is a joint BA/MA candidate at the Committee on International Relations, University of Chicago. He previously conducted research on cyberspace entanglement at the College of Information and Cyberspace; surveyed the Chinese semiconductor industry for the Silverado Policy Accelerator; and analyzed PLA information forces and long-term US-Taiwan-China trends for the Center for the Study of Chinese Military Affairs. Seth is a prior-enlisted Airman of the United States Air Force.
Image: William Potter/Shutterstock
