9% of my book, added around $380 recently. Posting the whole thing because I’d rather be told I’m wrong now than find out later.
The case: they’re the bottleneck for advanced nodes. Every hyperscaler’s capex has to physically pass through them, and there’s no second supplier at the leading edge. Capex guidance went up rather than down , but, management committing that kind of money is a stronger demand signal than anything they say on the call.
What the market is seems to be assuming is that Arizona and the other overseas fabs don’t meaningfully dilute the margin story, and that 2nm pricing holds.
My kill condition: overseas fab margins compressing two quarters running with no pricing offset. That’s the assumption above breaking, and it’s checkable each quarter rather than a vibe.
The risk I’m least able to size: the Shanghai domestic DUV lithography development. I don’t think it threatens the leading edge for years. But it doesn’t have to be competitive to hurt it just has to be good enough at the low end to compress pricing from below.
What I can’t figure out: how durable 2nm pricing is if Samsung fixes yields. TSM’s pricing power is the investment. I don’t have a good way to handicap it and I’d rather someone told me I’m underrating the risk.
Customer concentration is the other thing I’m carrying without a good answer, NVDA and Apple are a large share of revenue between them, but I don't think they'll leave anytime soon or something.