Dispersing the Fabs: TSMC's Expansion Beyond Taiwan and Its Geopolitical Limits
The political consensus that Taiwan’s concentration of advanced semiconductor production represents a strategic vulnerability has produced a global effort to disperse that production — or at least to replicate enough of it elsewhere that a Taiwan Strait crisis does not produce a complete collapse of advanced chip supply. TSMC is building fabs in Arizona, in Kumamoto and Hokkaido in Japan, and in Dresden, Germany. Intel is building in Ohio and Germany. Samsung is expanding in Texas. The CHIPS Act in the United States, the European Chips Act, and Japan’s semiconductor subsidy programs have collectively directed tens of billions of dollars at this dispersal objective. The effort is serious, expensive, and insufficient on the timescale that matters most.
The Arizona fab complex — TSMC’s most visible international expansion — has experienced delays, cost overruns, and cultural friction that have made it a case study in the difficulty of transplanting semiconductor manufacturing from an environment optimized for it over decades to an environment that is not. The first fab in Phoenix, producing at the N4 process node, has come online behind schedule and at higher cost than initial projections. The more advanced N2 fab, which would produce chips competitive with the most advanced TSMC Taiwan processes, has been pushed back. The workforce, supply chain, and operational culture that make TSMC’s Taiwan operations extraordinarily efficient cannot be recreated in Arizona on a capital budget, however large.
Japan’s TSMC expansion, centered on the Kumamoto fab in Kyushu, has moved more quickly than the Arizona project, partly because Japan’s semiconductor manufacturing ecosystem is deeper than Arizona’s and partly because the Japanese government has provided subsidies that effectively cover a substantial fraction of the construction cost. The Kumamoto fab produces at less advanced process nodes than TSMC’s leading-edge Taiwan operations — it targets the automotive and industrial semiconductor markets rather than the advanced logic chips that sit at the top of the capability hierarchy. A second, more advanced Hokkaido fab has been announced. The trajectory is real. It does not address the leading-edge gap on the timescale of near-term Taiwan contingency risk.
The leading-edge process nodes — TSMC’s 3nm and 2nm processes, and the N1.4 node that follows them — are the capability that matters most strategically. These processes produce the chips that go into the most advanced AI training systems, the most capable military electronics, and the highest-performance consumer devices. They are also the processes that require the most advanced equipment, the most experienced engineering workforce, and the most refined process control. They are the processes that TSMC’s Taiwan operations are uniquely positioned to execute at volume. None of the international expansion projects currently planned will produce leading-edge capacity comparable to TSMC’s Taiwan operations within the decade.
The equipment dependency compounds the geographic concentration problem. The extreme ultraviolet lithography machines that are essential for leading-edge semiconductor manufacturing are produced by ASML in the Netherlands. ASML ships approximately 50 EUV machines per year. TSMC’s Taiwan fabs use most of the world’s EUV installed base. Building parallel leading-edge capacity requires not just facilities and workers but EUV machines that must be ordered years in advance and that cannot be produced faster than ASML’s manufacturing capacity allows. EUV machines cannot be stockpiled in the way that chip inventories can. The bottleneck in the dispersal strategy is as much the equipment supply chain as the facility construction.
The dispersal strategy’s honest purpose is risk reduction rather than risk elimination. It aims to ensure that a Taiwan Strait crisis produces a severe shock to advanced chip supply rather than a complete and indefinite collapse. It provides enough leading-edge capacity outside Taiwan to sustain the most critical applications — military systems, essential infrastructure, the computing requirements of governments and financial systems — during a disruption that forces the rest of the global economy into rationing. It does not maintain the full capacity of the global electronics and AI industry through a conflict that destroys or captures TSMC’s Taiwan operations.
The fabs outside Taiwan matter. They are not the answer to the question that TSMC’s Taiwan concentration poses.