Higher rates will separate the winners and losers in the AI trade: T. Rowe Price

T. Rowe Price’s Rahul Ghosh says a more hawkish Fed is unlikely to derail the AI trade if earnings and free-cash-flow growth stay strong. He argues markets have underestimated returns from AI investment, citing strengthening recent earnings that support an extended hyperscaler data-center and semiconductors capex cycle.

Analysis

The market is still treating AI like a pure duration trade, but the next phase is more about self-funding capacity than multiple expansion. If hyperscalers keep converting incremental AI spend into visible revenue and FCF, higher discount rates should matter less for the platform leaders and more for the long-tail of speculative software where monetization is still theoretical.

The clearest winners are the semiconductor and datacenter infrastructure stacks with pricing power and backlog visibility: NVDA, AVGO, ANET, VRT, ETN, and the memory/packaging chain (MU, TSM, AMAT) if capex remains disciplined. A hawkish Fed also widens the gap between cash-rich hyperscalers and weaker AI-adjacent names, because the former can keep funding the cycle internally while the latter face funding-cost pressure and slower customer conversion.

The contrarian mistake is assuming the main risk is a higher 10Y alone; the real break would be a capex pause or evidence that AI spend is still producing weak incremental returns. Over 1-3 months, any earnings beat from cloud or semiconductor names should keep forcing investors to chase the trade; over 6-18 months, only firms with pricing power and real FCF accretion will deserve premium multiples. Falsifier: a broad reduction in hyperscaler capex guides or a sustained deterioration in FCF conversion despite heavy AI spend.

Key Decisions for Investors

• Buy SMH on rate-led pullbacks over the next 2-6 weeks; pair against ARKK or unprofitable software baskets to express 'cash flow beats discount-rate' over 1-3 months.
• Overweight AVGO/NVDA/ANET versus high-multiple AI software peers for a 3-6 month horizon; risk/reward is strongest where backlog and pricing power can offset yield pressure.
• Add VRT or ETN on any weakness tied to higher yields; the trade is that datacenter power/cooling capex has a longer runway than headline AI enthusiasm.
• Set an alert for the next hyperscaler earnings cycle: if capex is maintained and FCF remains stable, add to the AI basket; if capex rolls over, cut exposure quickly because the trade would shift from earnings compounders to multiple compression.