TSMC posted record $40 billion revenue. Its stock fell 4%. Investors are no longer buying the AI spending story on faith.

The market is treating this as a free-cash-flow story, not an operating story. For TSM, that usually means multiple compression first and only later a debate about whether the spend is actually building a wider moat; in the near term, higher reinvestment intensity tends to matter more than incremental revenue beats. The spillover is more interesting: if the capex is truly being pushed into leading-edge capacity, the clearest beneficiaries are the tool vendors and service names with backlog leverage, especially ASML, AMAT, LRCX and KLAC, because the spend is effectively pre-committed demand with better visibility than end-market semis.

The second-order read-through is that TSM is signalling confidence in multi-quarter AI and advanced-node demand, which should ease supply constraints for NVDA/AMD/AVGO over a 1-3 month horizon. But the same spend also deepens depreciation and overseas-fab drag over 6-18 months, so the equity market can punish TSM even if the business remains structurally strong. If the next call or investor day shows capex is being pulled forward for geopolitical duplication rather than node leadership, that is the key falsifier for a bullish interpretation.

Contrarian view: the move may be overdone because investors are extrapolating one quarter of heavy spending into a permanent margin reset. If utilization stays tight and operating cash flow keeps funding the buildout, the selloff could reverse quickly once the market reclassifies the spend as a moat investment. The bigger risk is not demand destruction, but that the mix shift toward non-Taiwan capacity lowers long-run economics and keeps the stock capped even if revenue growth remains exceptional.