Taiwan Semiconductor Manufacturing Company Limited TSM

Measured against its own history, TSM looks expensive, not cheap: it trades near 35 times trailing earnings and ~25 times expected earnings, versus a ten-year average closer to 20. Conservative valuation models put the ADR’s intrinsic value around $142–267, well below the ~$424 price, so the stock leans on the AI cycle continuing rather than on current cash flows. The offset analysts cite: at ~25x forward against >30% guided growth, the price-to-growth ratio sits below 1 — high growth priced in, but not outright euphoria.

Not really — and that’s the counterintuitive part. A genuine “Taiwan discount” would show up as a cheap multiple, but TSM trades at a ~35x premium to its ~20x ten-year average. The price is paying for the AI cycle, not marking the stock down for geopolitics. Taiwan risk sits as a tail — improbable but severe given the world depends on the island’s fabs — that a premium multiple doesn’t contemplate. The practical question for a reader is how much of that AI premium would survive if the strait returned to the headlines.

Less than the headlines suggest: sales to Chinese customers fell to about 9% in 2025 — and 7% in the first quarter of 2026 — down from 20–22% in 2019-2020, and the Nanjing plant is only ~2.4% of the business. The bigger issue is political plumbing: Washington revoked Nanjing’s Validated End-User status effective December 31, 2025, then granted TSMC an annual, renewable license to keep importing US tools. So China is now a supply-chain and licensing risk more than a demand risk.

TSMC’s Q1 2026 gross margin was 66.2%, and management guided the second quarter to 65.5–67.5% — extraordinary for a manufacturer. It holds despite two drags: the 2-nanometer ramp (a transitory 2–3 points) and overseas fabs in Arizona, Japan and Germany (a structural 2–3 points, widening to 3–4). What absorbs them is the moat: roughly 70% foundry share and a node lead that confers pricing power, letting TSMC pass much of the higher overseas cost on to customers because there’s nowhere else to fabricate leading-edge AI chips at scale.