The AI trade’s worst day in a year became a buying opportunity by Monday

This is less a clean AI-de-risking than a violent factor rotation caused by duration sensitivity and positioning. The tape is telling us the marginal buyer still believes in the AI capex supercycle, but the marginal seller is now using any macro excuse—rates, geopolitics, or inflation—to cut the highest-duration exposure first. That means the near-term leader/laggard split is likely to persist even if the index levels stabilize: semis with direct AI leverage can bounce hard, while broader tech breadth remains fragile until rate expectations stop ratcheting up.

The more important second-order effect is in supply chain dispersion. If the market starts questioning whether hyperscaler demand is timing-shifted rather than cancelled, the first casualties are not the obvious AI winners but the second-tier beneficiaries priced on terminal demand assumptions: networking, memory, and equipment vendors with little pricing power. Memory in particular is where a small change in perceived demand can turn into a large multiple compression because the market has been underwriting a prolonged shortage narrative; that makes rallies in the chip complex vulnerable to sharp reversals if CPI/PPI come in hot.

The geopolitical layer matters because it hits the same crowded trade from a different angle: higher oil and higher yields are both anti-multiple forces for long-duration growth. A temporary de-escalation in the Middle East can keep the risk premium contained, but if crude re-accelerates while inflation data surprises up, the market will reprice not just Fed cuts but the entire AI cash-flow discount rate. That is a more dangerous setup than last week’s selloff, because it combines a valuation headwind with a macro excuse to keep selling into strength.

Consensus is probably underestimating how much of the AI trade was being supported by momentum and passive flows rather than fundamental revisions. The rally can survive a one-week reset; it is much less resilient if the next two macro prints confirm that rates are staying higher for longer. In that case, the right move is not to short the winners outright, but to fade the most duration-sensitive names against higher-quality cash generators that benefit from real earnings now, not promises later.