The Foundry Layer Confirms: TSMC's Sixth Straight Record Quarter, a Q4 Guide Implying +55–61% YoY, and 2026–28 Capex "Significantly Higher"

Updated

The Foundry Layer Confirms: TSMC's Sixth Straight Record Quarter, a Q4 Guide Implying +55–61% YoY, and 2026–28 Capex "Significantly Higher"

The upstream physical evidence for the AI capex story (The Upstream Order Book: ASML Says Demand Will Outrun Its Supply Through 2028, With Order Interest Out to 2030 carries the equipment layer) got its foundry-layer confirmation this cycle: TSMC's Q3 2026 print (Oct 15) was the strongest yet — and it landed mid-selloff.

The numbers. Q3 revenue hit NT$1.494T ($46.72B), +50.9% YoY and +16.2% QoQ in USD terms — above the company's own $44.6–45.8B guidance. Net profit was NT$763.06B (~$23.59B), +68.7% YoY, beating the NT$746.3B LSEG SmartEstimate — the sixth consecutive record quarter and 11th straight quarter of YoY profit growth. Operating margin was 58.7%, above the guidance high end; HPC rose to 68% of revenue (+22% QoQ); 2nm already contributed 8% of wafer revenue, up from 3% in Q2. Q4 guidance is $52.3–54.3B — implying roughly 55–61% YoY growth — and full-year 2026 revenue growth was raised to ~44% (from >40%).

The demand language. CEO C.C. Wei, on the call: "the demand is very, very high. The gap between demand and supply are still quite big1." Pressed on whether the mid-to-high-50s% AI accelerator CAGR (2024–2029) still holds: "Whether it is a high 50 CAGR, it does not matter. It probably will be higher." CFO Wendell Huang on capacity: capex from 2026 to 2028 "is expected to be significantly higher than the past three years" — TSMC is building nearly 20 fabs worldwide (13 in Taiwan, 5–6 overseas), with N2 capacity to more than double in 2027 and the 2026 budget maintained at $60–64B.

The ground-truth check. Most directly relevant to the deployment-gap debate (Moody's Puts a Rating-Agency Number on the Committed-vs-Started Gap: $969B Committed to AI Infrastructure, $662B in Leases Not Yet Started), Wei said TSMC is now verifying its customers' buildouts itself: "we are now going one step further and sending our own teams of people to check on the data center build-out progress and power supply situation in the U.S. The feedback is so far so good, and the CSPs has been well-prepared."

The caution. Reuters: TSMC "struck a note of caution for 2027... due to macroeconomic uncertainties, tempering an otherwise upbeat outlook" — Wei: "we are being prudent in our business planning going into 2027," citing rising component prices and challenged consumer/price-sensitive end markets. And the beat "has done little to arrest a deepening selloff" across AI-linked semis — Taipei shares fell ~2% into the print.

What it means for the topic question. At the layer where capex converts to silicon, the story is not just intact — it is accelerating: record profits, a raised growth outlook, and a multi-year capacity commitment backed by TSMC's own balance sheet. The market's October selloff is not pricing a demand break at the foundry; it is pricing the financing path and the rate path (see Financing Layer: Anthropic Offers Blackstone 10%+ to Anchor the Junior Tranche — the Unguaranteed AI-Credit Price Arrives, 3.7%: September CPI Hot, an October Hike Back in Play at ~37% Odds — and AI Itself Named as an Inflation Driver).


  1. An instance of Supply, not demand, now rations every layer of the AI silicon stack — The foundry layer's sixth record quarter and 'significantly higher' 2026–28 capex commit its own balance sheet years out because physical capacity, not demand, rations the buildout. ↩︎

Part of

This finding is an example of a pattern recurring across your work:

Backlinks

Revision history

  • New subject: the foundry layer as the upstream physical-evidence gauge for AI capex — TSMC's Q3 2026 record print and raised outlook, the foundry-layer counterpart to the ASML equipment-layer note.
    · by the agent