Here Is the 1 Dirt-Cheap Semiconductor Titan I Keep Loading Up on Repeat

Taiwan Semiconductor reported Q2 2026 revenue of NT$1.134 trillion, up 21.45% YoY, with net income rising 43.82% YoY to NT$572.8 billion and Q1 2026 EPS beating estimates by 8.39% at $3.49. Management said AI-related demand remains "extremely robust," guided to above 30% full-year 2026 revenue growth in U.S. dollar terms, and highlighted a 35% U.S. investment tax credit for Arizona. The bullish case is reinforced by a 36.2% ROE, 58.1% operating margin, a 27x forward P/E, and continued capital returns via NT$155.6 billion in cash dividends, though Taiwan/geopolitical risk remains a key overhang.

Analysis

The market is still treating the foundry like a cyclical supplier, but the better framing is as a capacity-constrained infrastructure asset embedded inside AI capex. That matters because the next leg of upside is not just unit growth; it is mix-driven pricing power from advanced nodes, advanced packaging, and customer willingness to prepay for supply assurance. The second-order effect is that every constraint TSM removes tends to widen the moat rather than normalize margins, because smaller peers cannot fund the same pace of node migration and geographic replication.

Consensus appears underappreciating the duration of demand visibility. If AI accelerator growth stays elevated through 2029, the relevant valuation multiple should compress less than a typical semiconductor name because earnings are becoming more annuity-like through multi-year tooling cycles and ecosystem lock-in. The real beneficiaries are upstream equipment vendors, specialty materials, and advanced substrate/package providers that get pulled along by TSM’s buildout; the losers are fabless rivals and foundry alternatives whose customers will prioritize guaranteed capacity over marginal cost savings.

The main tail risk is not a quarter or two of softer revenue; it is a regime shift in gross margin if geopolitical risk forces duplicated capacity before utilization is mature. The U.S./Japan/Germany footprint lowers headline concentration risk but also raises execution risk, because early-stage overseas fabs can dilute returns before subsidy capture fully offsets costs. A sharper-than-expected easing in AI capex, export controls, or customer inventory digestion would hit the multiple first, fundamentals later.

Contrarian view: the stock may be less cheap than it looks if investors are assuming both secular AI demand and structurally superior margins can coexist indefinitely. The market could be right to discount some of the growth because TSM is increasingly forced to spend ahead of revenue, making free cash flow more volatile than operating earnings suggest. That said, the setup still looks asymmetric if one believes AI infrastructure spend remains a 24-36 month story rather than a one-quarter trade.

Key Decisions for Investors

• Maintain core long TSM, but add on 5%-8% pullbacks only; near-term drawdowns are more likely to be sentiment-driven than fundamentals-driven over the next 1-3 months.
• Pair long TSM / short SOXX for 3-6 months: if the market continues to reward the whole complex indiscriminately, TSM should outperform on scarcity value and customer lock-in while broad ETF beta remains crowded.
• Buy TSM call spreads 6-12 months out instead of outright calls; the risk/reward is better if the stock grinds higher on earnings revisions rather than re-rates immediately.
• Add a basket long in ASML, AMAT, and KLAC versus short a weaker foundry or analog supplier for 6-12 months to express the capex-pull-through thesis.
• Use a trailing stop only on a geopolitical headline shock, not on routine volatility; the fundamental thesis should break only if AI capex guidance or TSM’s capacity-add cadence rolls over for multiple quarters.