Export controls on advanced AI chips are dividing the global compute ecosystem, fragmenting supply chains and forcing regional architecture choices.
Earlier this month, Taiwan announced new export controls on advanced artificial intelligence chips destined for China. These controls would extend existing restrictions beyond blacklisted firms like Huawei and classify personal imports of AI hardware into China as a criminal offense. As part of ongoing trade negotiations with the United States, the proposal pulls Taiwan—which manufactures the world's most sophisticated processors—into the American export-control arena. Though a marginal bureaucratic step, it carries immense implications: a world where computing power is segmented along geopolitical lines.
The race to lead in AI is often framed as competition to produce the most cutting-edge models and attract elite talent. Yet the more durable advantage lies in producing the physical infrastructure that enables complex AI computation. The jurisdiction controlling the supply of advanced semiconductors and the energy systems powering them gains authority to set standards for how the next generation of technology develops, including how it is regulated and where it is deployed. Fundamentally, this control determines which countries and firms can access high-performance processors.
Governments worldwide are weaponizing export controls to project power over computation development. By restricting which countries and firms can purchase high-performance processors and fabrication equipment, they shape the speed and direction of AI advancement across borders. AI ultimately depends on powerful computing, which depends on a set of advanced chips produced through a supply chain centralized in a select group of countries. When a jurisdiction controls any chokepoint along that chain, it gains leverage over the entire manufacturing ecosystem.
As export controls tighten globally, the AI ecosystem risks dividing into rival compute blocs—a fragmentation that raises production costs, slows innovation, fragments standards and research, and reduces the economic interdependence that has maintained stability in the Taiwan Strait region.
The United States recently strengthened enforcement of chip export controls, minimizing loopholes that previously allowed China to acquire advanced chips through overseas subsidiaries. It is now pushing to further restrict the supply chain by increasing oversight of contract manufacturers like Taiwan Semiconductor Manufacturing Company (TSMC), which fabricate most of the world's leading chips. Tighter oversight of these facilities grants the United States strategic advantage over global compute supply, governing the critical chokepoint.
Taiwan is aligning with the United States for both strategic and commercial reasons, establishing strict export controls on AI chips to China and penalizing circumvention. The country's security ties with the United States function through the same industrial base the export controls protect.
In response to this coordinated pressure, China is developing a nationwide AI infrastructure push reportedly worth $295 billion, aimed at reducing reliance on US-controlled chips by building independent compute capacity. The initiative would finance domestic chip production, including design and manufacturing, alongside the data center and energy infrastructure to run AI internally.
The AI ecosystem is thus splitting in two: one bloc built on US chip design and manufacturing in Taiwan and Japan, the other based on China's semiconductor self-sufficiency strategy. At the center sits US company Nvidia on one side and Chinese company Huawei on the other, with Taiwan and TSMC hedged between them as the chokepoint both blocs depend on. These parallel stacks for AI development feature separate chips, processes, and regulatory frameworks.
The fundamental risk is that these compute blocs will increase costs and slow progress through fragmented standards and research. A parallel system could also remove the economic interdependence that has long tempered geopolitical conflict over Taiwan. Export controls intended to slow rivalry are ultimately pushing toward a global AI landscape that is both more expensive and less innovative.