Asian Foundries Race To Match Taiwan's Capacity As Geopolitics Pressures Supply Chains

Semiconductor foundries across Korea, Japan, and Singapore are accelerating capacity additions in a coordinated push to reduce the industry's dependence on Taiwan, where TSMC still produces the overwhelming majority of advanced-node chips.

Samsung Foundry confirmed last week that its Pyeongtaek campus will house a third 2-nanometer line by 2028, eighteen months ahead of the previously communicated schedule. In Japan, Rapidus continues to receive METI funding to bring up its Hokkaido fab, with a clear target of competitive 2nm yields by 2027. Singapore's GlobalFoundries-Vanguard joint venture, often overlooked, is now tracking ahead of original capex on the 12-nanometer specialty line scheduled to ramp early next year.

The motivation is partly defensive. US export controls and the increasingly explicit defense framing of the chip industry have made customers — particularly hyperscalers — wary of single-source exposure. Apple, NVIDIA, and AMD have all hinted at a desire for second sources by the end of the decade, even if the immediate volume commitments remain modest.

The economics are still uncomfortable. Building leading-edge capacity outside Taiwan costs roughly 30% more on a depreciation-adjusted basis, and the talent pool — particularly in advanced packaging and lithography integration — remains concentrated in Hsinchu and the surrounding science parks. South Korea's MOTIE has earmarked targeted retention bonuses for senior process engineers; Japan is offering similar inducements to repatriated specialists.

Whether the diversification holds up to a real demand cycle remains an open question. AI infrastructure customers, who have been the loudest voice for second sources, may discover during the next allocation crunch that paying TSMC's premium feels less painful than waiting on a less-mature alternative. For now, the industry's geographic broadening is real but financially fragile.