AAPL
AMD
NVDA
TSM
The AI chip dinner rush
Taiwan Semiconductor is acting like the world’s most popular restaurant just got hit with a line out the door. The company plans to spend nearly $56 billion this year to expand capacity for AI customers including Apple, Nvidia, and AMD, because apparently “just make more chips” is not as easy as hitting copy-paste.
CEO C.C. Wei said the brutal part is the clock. A new fab can take 2 to 3 years to build and then another 1 to 2 years to ramp, which means today’s demand frenzy is still being solved with tomorrow’s supply. In other words: the AI boom is moving at internet speed, while chip factories are running on construction timelines.
The bottleneck isn’t going away
TSM is pushing new 3-nanometer fabs in Taiwan, the U.S., and Japan, with production timelines stretching into 2027 and 2028. It’s also expanding 3-nanometer output for the first time, because demand is spilling beyond AI into smartphones, automotive, and IoT. That’s a nice problem to have, but it’s still a problem.
Analysts say the company remains the tech and market leader, but competitors are circling:
- Samsung is getting more competitive at 2-nanometer tech
- EUV tool supply remains a choke point
- Demand for 2-nanometer-and-below chips could outstrip capacity well into 2030
Why investors should care
This is the kind of update that tells you the AI trade is still early — and still supply-constrained. If TSM can keep widening its lead in advanced nodes and packaging, it stays in the catbird seat for one of the biggest buildouts in tech.
The stock was down slightly on the day, but that’s hardly the point. The bigger story is that TSM’s moat is looking less like a moat and more like a castle with a decade-long waiting list.
Big picture: when the whole AI economy is thirsty, the company controlling the water pipes tends to do pretty well.