TSMC's Arizona Second Fab Moves Into Pilot Production Ahead Of Schedule

Taiwan Semiconductor Manufacturing Company has begun pilot production at its second Arizona fab roughly six months ahead of the schedule that had been internally targeted, with risk wafers running through the line and volume production now formally guided for the first quarter of 2027.

The Phoenix-area facility, which the company refers to as Fab 21 Phase 2, is configured for the 3-nanometre node — a meaningful upgrade from the 4-nanometre process that runs in the first phase and a partial closure of the technology gap with TSMC's Taiwan-located leading-edge fabs. The acceleration has been driven partly by lessons learned from the first phase's longer-than-expected ramp and partly by aggressive scoping of the construction sequence by the same teams that delivered the company's Hsinchu and Kaohsiung facilities.

The customer-mix story is what most distinguishes the second fab from the first. Apple, AMD, and NVIDIA are all confirmed customers, and several other major US-based fabless customers have committed wafer-starts allocations through the early ramp window. The combination of a slightly older node than the absolute leading edge and a US-located capacity with confirmed eligibility for the CHIPS Act incentive structure has produced a customer-demand profile that is sold out for the first eighteen months of volume production.

TSMC's wider Arizona programme is also gathering momentum. The third fab, originally guided for late 2030 commissioning, is now targeting an earlier groundbreaking on a dedicated leading-edge process — most likely 2-nanometre — and the supply-chain and packaging-services ecosystem around the Phoenix campus is consolidating in a way that several US-based analysts have flagged as the more important second-order effect of the original investment decision.

For investors, the most important framing question is on the unit-economics gap. TSMC's Arizona output is structurally more expensive per wafer than equivalent capacity in Taiwan, and that gap has not closed materially through the first-phase ramp. Whether the second fab can narrow that gap, particularly with the higher-value 3-nanometre product mix and a more disciplined ramp curve, is the test that decides whether the wider US-located leading-edge programme is commercially viable on its own merits or remains substantially policy-dependent.