Taiwan Semiconductor Manufacturing Company announced a second $100 billion capital commitment to U.S. manufacturing, bringing its Arizona megaproject to $265 billion and marking the largest foreign direct investment in American semiconductor history. The expansion lands eighteen months after the first Phoenix fab began equipment installation and six months after Washington tightened export controls on advanced nodes to China.
The commitment extends TSMC's domestic footprint beyond the initial three-fab complex in Maricopa County. The company disclosed plans for at least two additional advanced packaging facilities and a fourth logic fab capable of 2-nanometer process technology by late 2028. Current Arizona operations house one 4-nanometer production line scheduled for volume output in Q2 2025 and a second 3-nanometer facility targeting 2026 ramp. The fourth fab represents the first time TSMC will manufacture its leading-edge node on U.S. soil within twelve months of Taiwan deployment.
The scale shift reflects two pressures. First, hyperscaler AI infrastructure spending now runs at $200 billion annually across the top four cloud providers, with semiconductor content representing 42% of that capital allocation according to Morgan Stanley's January buildout survey. NVIDIA alone consumed 18% of TSMC's total wafer capacity in Q4 2024, up from 11% a year prior. Domestic U.S. fab capacity allows the chipmaker to serve this customer base without tariff exposure or logistical friction as AI accelerator volumes triple through 2027. Second, Washington's October 2024 update to semiconductor export restrictions effectively cut TSMC's addressable China revenue by $8 billion annually, removing 14% of trailing sales. The Arizona buildout replaces that revenue base with strategic customer proximity and eliminates the political risk embedded in cross-strait manufacturing concentration.
The project timeline matters for three constituencies. Defense primes including Northrop Grumman and Lockheed Martin already contracted for $4.2 billion in TSMC Arizona capacity through 2029 under Pentagon direct-pay agreements finalized in November 2024. The 2-nanometer node coming online in 2028 enables the next generation of satellite processing and autonomous platform compute that current 5-nanometer defense chips cannot support. For hyperscalers, domestic advanced packaging co-located with logic fabs cuts AI accelerator delivery lead times from 16 weeks to 9 weeks, according to TSMC's client engineering roadmaps shared in December. That velocity advantage alone justifies the 8-12% wafer price premium for U.S.-manufactured silicon. For TSMC, the move diversifies geopolitical risk while locking in $6.6 billion in CHIPS Act subsidies and $5 billion in federal loan guarantees that reduce the effective cost of capital by 140 basis points compared to Taiwan-based expansion.
Operators should track three milestones. First, Arizona's first fab production yield data in Q3 2025 will reveal whether TSMC can replicate Taiwan's 92% good-die rates in a new labor market with limited process expertise. Second, the U.S. Commerce Department's subsidy disbursement schedule, with the next $1.8 billion tranche conditional on hitting construction benchmarks by August 2025. Third, any announcement of European or Japanese co-investment in the Arizona complex, which TSMC's chairman referenced as under discussion in the January earnings call without naming partners.
The $265 billion total now exceeds Intel's entire market capitalization by $103 billion and represents 5.2 times TSMC's trailing twelve-month free cash flow, financed over eight years.