TSMC

TSMC enters mid-2026 as the single chokepoint of the entire AI infrastructure cycle. The most recently reported quarter, Q1 2026, delivered revenue of $35.90 billion (+40.6% YoY in USD), beating consensus of $35.5 billion. Net income surged 58.3% YoY to NT$572.48 billion, diluted EPS of NT$22.08 (about $3.49 per ADR) topped estimates of NT$20.88, and gross margin expanded to 66.2% — above the guided range. The advanced-node mix is the real story: 7nm and below now make up 74% of total wafer revenue, with 3nm contributing 25%, and the High Performance Computing platform reaching 61% of revenue (up from 51% a year earlier) after a 20% sequential jump.

The most important shift since the start of 2026 is that 2nm (N2) is no longer a future catalyst — volume production began in Q4 2025 at Fab 22 in Kaohsiung, and reported yields have already surpassed 65%, unusually strong for a first-generation gate-all-around node. Apple booked over half of initial 2nm capacity for the A20 processor and M6 chip. An enhanced N2P with backside power delivery follows in late 2026. Q2 2026 guidance calls for revenue of $39.0-40.2 billion at gross margins of 65.5-67.5%, and CEO C.C. Wei lifted the full-year 2026 USD revenue-growth target to above 30% with capex toward the high end of the $52-56 billion range. One yellow flag: April 2026 monthly revenue of NT$410.73 billion grew just 17.5% YoY — the slowest pace in months — as comparisons toughen. The May print is due around June 10.

TSM has unusual smart-money composition: it is one of Warren Buffett’s best-known sales (initiated and exited within roughly six months in 2023, with Buffett citing geopolitical risk). That sale was at roughly $80 per share; the ADR now trades near $447, and Buffett has not re-entered. Among other BMI-tracked managers, Mohnish Pabrai built a position in late 2023 and has held through 2025-2026; it appears in Pabrai Investment Funds 13F filings at roughly 6% of portfolio. Stanley Druckenmiller cycled through TSM in 2024 but does not currently hold.

The institutional flow narrative for 2026 has shifted as the geopolitical-risk discount Buffett cited has compressed materially — the ADR is up well over 5x from his exit, and the 52-week range now spans roughly $200 to $449. Insider activity at TSM is structurally less observable than at U.S. peers because of Taiwan disclosure rules, but tracked Q1 13F flows from large institutional accounts showed net buying through the quarter as the AI-capex thesis broadened beyond NVIDIA. The most-watched structural item remains the Arizona expansion: the second fab (3nm) is tracking equipment move-in around October 2026 with volume targeted for Q4 2027, while 2nm in Arizona is now guided toward the end of the decade — roughly 2029.

TSM trades near $447 per ADR, a trailing P/E of roughly 36 and a market cap near $2.19 trillion, after climbing from a 52-week low around $200. On a forward basis the multiple is more moderate — roughly 22-24x consensus 2026 EPS and into the high-teens on 2027 — still a discount to fabless customers like NVIDIA and AMD despite superior gross margin (66%+) and lower customer-concentration risk than the smaller foundries. On EV/EBITDA, TSM sits well above Samsung Foundry and Intel Foundry, both correctly discounted for weaker process competitiveness. Wall Street is at Strong Buy across roughly 17-19 covering firms, with an average target near $468 (high $600, low $354) — implying only mid-single-digit upside after the run. The geopolitical-risk overhang Arizona is dismantling is now largely priced; the bull case rests on AI-capex durability rather than on cheapness.

1. Around June 10, 2026: May 2026 monthly revenue print — the first read on whether growth re-accelerates after April’s 17.5% YoY (slowest in months), a high-frequency tell on AI-foundry demand
2. Mid-July 2026 (estimated): Q2 2026 earnings — confirms whether revenue lands within the $39.0-40.2 billion guide; gross margin near the 67.5% top would signal 2nm yield is exceeding plan
3. September 2026: Apple iPhone 18 Pro launch with the A20 (2nm) — first mass-market product on the N2 node, validating yield, capacity ramp, and per-wafer pricing power
4. Late 2026 / Q4 2027: N2P with backside power delivery (late 2026) and Arizona second-fab (3nm) volume targeted Q4 2027 — milestones that further reduce the geopolitical discount

TSM remains the cleanest AI-infrastructure compounder I track, but the easy money is gone. When I first flagged it near $80-190, the geopolitical discount was the whole story. Today, near $447 with a 36x trailing multiple and the ADR pressed against all-time highs, you are no longer paid to take Taiwan risk — you are paying up for confirmed 2nm execution and a durable AI-capex cycle. The thesis is intact: 66%+ gross margins, 30%+ revenue growth, a structural foundry monopoly, and 2nm already shipping at healthy yields. But with consensus targets only mid-single-digits above the current price, I am no longer an aggressive buyer here. I trim into strength above $440 and would add seriously only on a meaningful pullback toward the mid-$300s — the low end of the analyst range — or on any AI-capex scare that overshoots fundamentals. The April monthly slowdown to 17.5% is worth watching: the comparisons are getting harder, and at this multiple the market is pricing perfection. The thesis breaks if Taiwan Strait tensions escalate to direct kinetic risk — under 5% probability over 24 months in my book, but not zero.