TSMC reportedly weighs Texas expansion, but key US tax incentive could decide the move as Singapore enters the frame

TSMC's potential Texas expansion is not a capacity play—it is a structural reorganization forced by the geopolitization of AI compute. Technical cascade: A second US fab demands a full EUV tool-cluster redeployment in Texas. ASML's delivery schedule is already locked by Arizona; a parallel Texas ramp would compress the 2nm yield-curve window to under 12 months—unprecedented in industry history. More critically, advanced-node knowledge density is dispersing from Hsinchu toward Texas and Singapore, eroding the geographic moat that has protected the Taiwan, China cluster for three decades. Compliance leverage: Singapore entering the frame is not a fallback; it is a negotiation weapon. TSMC is converting multi-site optionality into binding demands on tax credits, water allocation, and power guarantees from Washington. US fab OPEX runs 30-40% above Taiwan, China; without adequate subsidy, sub-3nm gross margins face structural impairment. Competitive response: Intel's 18A remains below yield target. A Texas second site ramping by 2027 closes the last window for Intel's domestic foundry ambitions. Samsung's Taylor fab faces a timing disadvantage—TSMC's client lock-in with NVIDIA and Apple creates an insurmountable switching-cost barrier. 12-24 month outlook: Global advanced capacity shifts from single-pole concentration to a three-pole architecture. But if all planned fabs ramp simultaneously by 2027, a structural 3nm oversupply emerges, and pricing power migrates from foundry back to design houses.