Earnings From Taiwan Semiconductor and ASML Show Soaring Demand, So Why Are AI Stocks Falling? (And Here's What Investors Should Do Next.)

The market is not disputing AI demand; it is discounting the durability and payback of the spend. That creates a widening gap between firms with backlog/order visibility and those whose valuation depends on monetizing the narrative, so the near-term loser set is more likely the high-multiple application layer than the wafer-fab/equipment complex. In other words, if capex remains elevated, TSM/ASML should convert it into earnings with less ambiguity, while NVDA and PLTR are more exposed to any pause in incremental budget authorizations because the market is already pricing perfection.

The second-order risk is a flow-driven multiple reset, not an earnings air pocket. When investors rotate toward “safer” growth, the first names to de-rate are crowded AI leaders with the longest duration cash flows; that can persist for 1-3 months even if fundamental demand stays intact. The catalyst to reverse that trade is any hard evidence that hyperscaler AI budgets are still expanding into the next budget cycle, or that ASML/TSM order momentum accelerates again after this digestion phase.

Contrarianly, the consensus may be overestimating near-term capex fatigue and underestimating how much of the AI buildout is now structurally embedded in cloud infrastructure and advanced-node supply chains. A pullback in the leaders does not automatically mean the AI trade is broken; it may simply be a valuation reset before the next leg. The cleanest falsifier is a visible slowdown in order growth or guidance from TSM/ASML over the next 1-2 quarters; absent that, selling infrastructure on soft tape looks more tactical than fundamental.