The most advanced computer chips in the world, the ones used in everything from F-35 fighter jets to iPhones to the AI systems that the entire US tech industry now runs on, are mostly made by a single company on a single island. The company is Taiwan Semiconductor Manufacturing Company, or TSMC. The island is Taiwan, which sits about 100 miles off the coast of China. Beijing has said openly and repeatedly that it considers Taiwan to be part of China and is willing to use force to bring it under control. American war planners spend a significant chunk of their time thinking about what would happen if it ever tried.
Reporting from the New York Times, the Financial Times, Bloomberg, and the Center for Strategic and International Studies all describe the same basic picture. Roughly 90 percent of the world’s leading-edge chips, meaning chips at 7 nanometers and below, come from TSMC. About 65 percent of all semiconductors used in US weapons systems originate in Taiwan. A 2024 RAND Corporation analysis concluded that a Chinese blockade or invasion of Taiwan would cause an immediate global GDP contraction larger than the 2008 financial crisis, primarily through the disruption of the chip supply chain.
The CHIPS Act, signed in 2022, set aside $52 billion in subsidies and incentives to bring chip manufacturing back to American soil. The early results are real but partial. TSMC is building plants in Arizona, the first of which began limited production in late 2024 and is ramping up through 2026. Intel is building new fabs in Ohio and Arizona. Samsung is building in Texas. Micron is expanding in New York and Idaho. The combined capacity, when all the announced plants are operating in 2028, will produce roughly 20 percent of the world’s leading-edge chips on US soil, up from essentially zero in 2022. That is real progress. It is also still a long way from making the country self-sufficient.
The concentration of leading-edge fabrication in Taiwan was not accidental. It is the result of three decades of capital allocation choices, in which US fabless design firms outsourced manufacturing to lower-cost Asian foundries while domestic capacity atrophied. Industry analysts at CSIS and the Semiconductor Industry Association have documented the timeline. Intel’s stumble at the 10nm and 7nm process nodes between 2015 and 2022 widened the gap further. The CHIPS Act addresses the symptom, but the labor pipeline for fab operators and process engineers remains thin. TSMC’s Arizona buildout has been delayed twice on staffing grounds.
The construction timeline for semiconductor fabs runs five to seven years, with another two to three to reach full yield. The CHIPS Act subsidies are obligated in tranches through 2030. The political question on the horizon is whether the next Congress sustains the appropriations cycle for Phase 2 of the program, currently estimated at an additional $40 to $60 billion to push US share of leading-edge production beyond the announced 20 percent floor. Reauthorization is expected to come up in 2027. The position of every member of the relevant committees is on the record.