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

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.