They are spending $165 billion to build structurally higher-cost capacity for customers who have absolutely no other choice.
Thesis: TSMC owns the bottleneck of the century. The bearish fixation on a 2-3% margin dilution from overseas expansion misses the forest for the trees: AI accelerator revenue is doubling, N2 tape-outs are breaking records, and they are the sole enabler of the $100B+ hyperscaler capex war. The 'dilution' is simply the cost of doing business in a fragmented world, and with 80% market share in advanced nodes, they will pass it on.
Verdict: LONG — Conviction: HIGH
Catalyst: Q2 execution proving the 57-59% gross margin floor despite Arizona ramp. Confirmation of price hikes for 'geographic flexibility' in late 2025.
Key Risk: China revenue evaporation that isn't fully backfilled by US hyperscalers, or a faster-than-expected mix shift to edge inference where TSMC has less leverage.
The Tell: C.C. Wei downgrading demand language from 'insane' to 'very strong' and 'a little bit balanced.' The panic-buying phase of the AI cycle is over; we are entering the execution phase. This isn't weakness, it's normalization, but the hyper-growth narrative junkies might twitch.
Friction Level: MODERATE_FRICTION — The Street models Arizona/Kumamoto as a permanent margin drag. The reality is it's a feature hyperscalers must pay for to secure supply.
Report not found
The report data is no longer available. Please return to the archive.