A global AI safety regime should learn from nuclear arms control by focusing on the physical infrastructure that enables strategic capabilities. Instead of just seeking promises, it should aim to control access to chokepoints like advanced chip manufacturing and the massive data centers required for frontier models.

The brazen smuggling of NVIDIA chips to China signals that the competition for AI dominance is an "all-out sprint" and a matter of national security. Control over compute infrastructure is now as geopolitically critical as energy, making it the central battleground of a new technological Cold War.

Former White House advisor Ben Buchanan argues that contrary to the popular phrase "data is the new oil," computing power is the true bottleneck and driver of AI progress. This physical reality—advanced chips primarily made by democracies—creates a powerful geopolitical lever to influence nations like China.

With 97% of high-end chips and 72% of the global foundry market controlled by Taiwan, specifically TSMC, any disruption—from military blockade to cyberattack—would trigger an 'economic apocalypse.' This massive over-concentration creates a singular, fragile chokepoint with no short-term alternative, threatening the entire global economy.

Beyond financial metrics, the most significant 'tail risk' to the AI boom is the high concentration of advanced semiconductor manufacturing overseas, particularly in Taiwan. A geopolitical conflict could sever the supply of essential hardware, posing a much more fundamental threat to the industry's growth than market volatility or corporate overspending.

Taiwan's TSMC dominates advanced chip manufacturing not only through technical excellence but also its business model. By acting as a pure-play foundry that doesn't compete with its clients (unlike Intel or Samsung), it fostered unique trust and partnerships, making it the central hub of the semiconductor ecosystem and a critical geopolitical asset.

The current AI landscape mirrors the historic Windows-Intel duopoly. OpenAI is the new Microsoft, controlling the user-facing software layer, while NVIDIA acts as the new Intel, dominating essential chip infrastructure. This parallel suggests a long-term power concentration is forming.

The long-term ability to scale AI compute is not constrained by power or data centers, but by the production of advanced semiconductors. The ultimate chokepoint is ASML, the world's only manufacturer of EUV lithography tools, which can only produce just over 100 units annually by 2030.

While energy is a concern, the highly consolidated semiconductor supply chain, with TSMC controlling 90% of advanced nodes and relying on a single EUV machine supplier (ASML), creates a more immediate and inelastic bottleneck for AI hardware expansion than energy production.