Semiconductor Supply Chains

A single facility in Taiwan makes the chips that run the world. Everyone has noticed.

The brief

A single firm — Taiwan Semiconductor Manufacturing Company — fabricates, as of the mid-2020s, the large majority of the world's leading-edge logic chips: the processors driving flagship phones, frontier AI systems, and some advanced weapons. Most of that leading-edge capacity remains in Taiwan, spread across fabs in Hsinchu, Tainan, and Kaohsiung, roughly a hundred miles across the strait from mainland China. The machines that print those circuits — extreme-ultraviolet lithography tools made by a single Dutch company, ASML — cost upward of $150 million each, weigh as much as a jumbo jet, and draw components from dozens of countries. The supply chain that produces a smartphone, a missile guidance system, or a frontier AI model passes through an extraordinary chain of chokepoints.

This concentration is the result of forty years of radical specialization under conditions of stable globalization. Different stages of chip production migrated to whichever location was best at them — design in California, fabrication in Taiwan and South Korea, packaging in Malaysia, lithography in the Netherlands, ultra-pure materials and photoresists in Japan. The model that made it possible was the fabless–foundry split, pioneered when TSMC founder Morris Chang opened a pure-play foundry in 1987: firms like Apple, Nvidia, and AMD design chips they never manufacture, while TSMC manufactures chips it never designs, each side specializing so deeply that a single leading-edge fab now costs around $20 billion. The economics compound — every new process node, from 7nm to 5nm to 3nm, demands costlier tools and bigger volumes, so the number of firms able to stay at the frontier has collapsed to three — TSMC, Samsung, and an Intel fighting to stay in the race — with TSMC dominant in leading-edge foundry production. By economic measures the system is extraordinarily efficient. By geopolitical measures it is as concentrated as a single point of failure can be. A Chinese invasion or blockade of Taiwan would not merely be a regional war; it would crater the global electronics economy within months, which is why the chips themselves are sometimes called a silicon shield — the very dependency that makes Taiwan a target also makes its destruction unaffordable to everyone. Recognizing the exposure, the United States launched one of the largest peacetime industrial policies in its history — the 2022 CHIPS Act, some $52 billion in subsidies — to bring fabrication onshore, while tightening export controls to deny China the most advanced tools and the engineers who run them. China is racing the other way, pouring state money into a domestic industry it cannot yet make self-sufficient.

Why nowThe semiconductor question sits behind almost every major geopolitical contest of the late 2020s — Taiwan's security, the AI race, US–China relations, the trajectory of European industrial policy. Export controls and counter-controls now reach down to individual machines and engineers, and China has retaliated by restricting exports of gallium, germanium, and the rare-earth elements that chipmaking itself depends on. The bottleneck has become the central lever of statecraft. Whichever country secures redundant access to leading-edge chips — through onshore fabs, allied supply, or breakthroughs that loosen the lithography bottleneck — will compound advantages across the rest of the technological stack, from artificial intelligence to precision weapons, for a generation.