The TSMC Arizona fab is finally producing chips

TSMC’s Arizona fab is producing 4nm chips.

That sentence took years to become true. Construction delays. Cultural friction between Taiwanese engineers and American workers. Arguments about work hours, management styles, and what “on schedule” means. Water allocation negotiations in a desert state. A pandemic. Supply chain disruptions. The kind of grinding, unglamorous difficulty that doesn’t make headlines.

But chips are coming off the line now. American-made, leading-edge semiconductors. For the first time in decades.

Why one fab matters

The CHIPS Act committed $52 billion to reshoring semiconductor manufacturing. The thesis was strategic: Taiwan makes 90% of the world’s most advanced chips. A single earthquake, a single geopolitical crisis, a single disruption in the Taiwan Strait, and the global economy loses access to the components that power everything from phones to AI data centers to military systems.

One fab doesn’t fix that concentration. Not even close. But it proves the concept. It demonstrates that advanced semiconductor manufacturing can happen outside of Taiwan and Korea. It creates institutional knowledge, supply chains, and a workforce that didn’t exist in Arizona three years ago.

The second fab will be easier than the first. The third easier than the second. That’s how manufacturing learning curves work.

The culture clash, revisited

I wrote about this in 2023, when reports emerged of friction between TSMC’s Taiwanese management and American workers. The details were predictable: TSMC expected the work intensity of Hsinchu (12-hour days, weekend shifts, minimal pushback on targets). American engineers expected the work culture of Phoenix (reasonable hours, work-life balance, union-adjacent expectations).

Neither side was wrong. Both sides were working from deeply held assumptions about what “work” means. The resolution, from what I’ve heard, was messy and ongoing. Some compromises. Some departures. Some adaptation on both sides.

This is the part of reshoring that policy documents skip over. You can move the equipment. You can build the clean rooms. You can install the ASML lithography machines. But the culture that produces a 92% yield rate at a 3nm process node isn’t something you can ship in a container. It’s built over decades. It lives in the habits and expectations of the people on the floor.

The water question

The fab uses approximately 4.5 million gallons of water per day. In Arizona. A state that’s been in a drought for over two decades. TSMC is recycling a significant portion and investing in water treatment, but the fundamental tension remains: advanced chip manufacturing needs a lot of water, and Arizona doesn’t have a lot of water.

This tension isn’t going away. It’s going to define where fabs can be built in the US. Not just “where are the engineers” or “where are the subsidies” but “where is the water.” The geography of chip manufacturing is also a geography of water rights.

What I’m thinking

One fab. 4nm chips. After years of delays and billions of dollars. It’s a beginning, not a solution.

The concentration risk remains. If something happens in Taiwan tomorrow, this single Arizona fab doesn’t replace what’s lost. The scale gap between “one fab producing” and “strategic independence” is enormous.

But beginnings matter. The first Falcon 9 landing was just one rocket landing on one barge. Seven years later, rockets land routinely. The first TSMC Arizona wafer is just one wafer. What matters is the second, and the tenth, and the hundredth.

I’ll take the beginning.

Related thinking: