TSMC Arizona faces water, labor constraints

Semiconductor fabs are basically giant utility machines that also happen to make chips. They need absurd amounts of ultra-pure water, huge blocks of reliable power, and a workforce that knows how to build and run some of the most finicky factories on earth. That is why the latest update on TSMC’s Arizona buildout matters. On May 12, Taiwan’s National Development Council minister Yeh Chun-hsien said the Phoenix project is going better than expected financially and operationally, but water shortages and labor constraints are still real problems. ### What changed today? The news is not that TSMC hit a delay. The news is that a senior Taiwan official came back from visiting the Arizona site and said the project is working better than TSMC expected — while also spelling out the bottlenecks more clearly than before. Yeh said the Arizona operation earned NT$16.14 billion, about $514 million, last year in its first full year of mass production. (taipeitimes.com) ### How far along is Arizona now? TSMC’s own Arizona page says the first fab started high-volume N4 production in Q4 2024. The second fab’s structure was completed in 2025, with N3 volume production targeted for the second half of 2027. The third fab broke ground in April 2025 and is aimed at N2 and A16, with production by the end of the decade. ### Why is water such a big deal? (taipeitimes.com) Chipmaking uses water at every step — cleaning wafers, rinsing chemicals, controlling contamination. In Arizona, the catch is obvious: the desert is not where you want loose assumptions about water supply. Yeh said TSMC still sees securing enough water as a major concern and wants support from Arizona’s state government. TSMC has also said it plans extensive recycling and treatment systems, which tells you the company knows this is not a side issue but a core operating constraint. (tsmc.com) ### Why is labor still hard? A fab is not just a building project. It needs construction workers now, then technicians, engineers, tool specialists, and managers for years after that. Yeh said labor shortages remain a concern and that TSMC has also had trouble getting visas for overseas hires. That matters because overseas staff are often the people who transfer process know-how from an existing fab to a new one. If those people arrive late, the whole ramp gets harder. (taipeitimes.com) ### Is this just an Arizona problem? Not really. It is the cost of trying to rebuild advanced chip capacity in the US at speed. TSMC says the Arizona project has grown from an original $12 billion announcement to a planned $165 billion footprint — six fabs, two advanced packaging facilities, and an R&D center. That is massive, but it also means every local bottleneck scales with the project. (taipeitimes.com) ### Why does this matter for buyers? Because “more domestic capacity” does not automatically mean “cheap and easy supply.” TSMC is still spending hard into AI demand — its 2026 capital budget is $52 billion to $56 billion. But Arizona shows the real constraint: upstream chip capacity depends on water, power, permits, and people, not just money. OEMs get more geographic diversification, which is good. They do not get a frictionless supply base. (tsmc.com) ### So what is the real takeaway? TSMC Arizona is no longer a symbolic reshoring project. It is a working fab cluster with profits, output, and a much bigger expansion map behind it. But the hard part is now visible. The US can attract leading-edge manufacturing capital. The harder trick is feeding it enough water, power, and skilled labor to keep the whole machine running. (taipeitimes.com) (investor.tsmc.com)