Leading-edge semiconductor capacity disperses and the concentration argument weakens
Concentration falls materially for the first time in two decades. Taiwan remains first and retains the most advanced node, but its share of the most…
Claude · 2032–2042 · plausible
Prior state
The overwhelming majority of leading-edge logic capacity sits in Taiwan. Subsidised fabs in the United States, Japan and Germany were financed in the early 2020s and are ramping. Export controls have slowed but not stopped Chinese progress, which has reached competitive performance at trailing-edge nodes and through advanced packaging without access to the most advanced lithography. The concentration of production in Taiwan has been widely argued to deter conflict.
Material change
Concentration falls materially for the first time in two decades. Taiwan remains first and retains the most advanced node, but its share of the most advanced logic capacity declines substantially as overseas fabs reach volume. Separately, Chinese domestic production reaches adequacy — not frontier parity — for its own military, industrial and consumer needs at a node one or two generations behind, supported by domestic packaging. Export controls consequently shift away from logic and toward equipment servicing, design software and high-bandwidth memory, where chokepoints persist.
Why now
Fabs financed in 2022–2026 reach volume production five to eight years later, which centres on this window. Chinese domestic toolchain programmes have a plausible maturity window in the same period. Neither of these is a surprise; what changes in this decade is that both mature simultaneously, which alters the strategic argument rather than merely the industrial map.
Mechanism and resistance
Overseas fabs face persistent cost disadvantages in construction, utilities and skilled labour, and yield ramps have been slower than announced. Talent is the binding constraint and is not quickly transferable. Domestic lithography at the most advanced level remains genuinely difficult and may not be achieved. Packaging, substrate and memory chokepoints do not disperse on the same timetable, so dependence persists in a different layer.
Consequences
The security implications run in both directions and should not be collapsed into one. Reduced concentration lowers the global cost of a disruption in Taiwan, which weakens both the deterrent value of that cost and the external incentive to prevent disruption. Allied industrial-policy spending comes under sharper domestic scrutiny once the strategic rationale is partly satisfied. Chinese adequacy reduces the leverage of controls, which pushes competition toward standards, memory and equipment servicing. Semiconductor employment becomes a regional political interest in three new places, with the durability that implies.
End state
A multi-region leading-edge base in which Taiwan is first but no longer nearly exclusive, Chinese domestic supply is adequate for domestic requirements, and export controls have migrated to equipment, memory and design tools.
Observable test
Share of sub-3-nanometre-class wafer capacity by location; volume production of a logic node using a substantially domestic Chinese toolchain; the composition of export-control entity and equipment lists.
Disconfirming sign
Overseas fabs sustain persistently low yields, leading-edge capacity remains concentrated at 2020s levels, and controls continue to focus on logic.