The best of the AI trade could be behind us. Here’s how to prepare, according to Barclays

Barclays said the AI trade is no longer a “one-way trade,” urging investors to be more selective and diversify. The bank cited hotter-than-expected CPI and PPI readings over the past week as increasing the probability of another Federal Reserve rate hike.

Analysis

The key risk is a split between AI earnings and AI duration. Higher real-rate expectations can compress valuations immediately, especially for businesses whose AI revenue is distant or difficult to verify, without yet changing infrastructure suppliers’ reported orders. The second-order test is whether sustained financing costs cause large technology buyers to raise project-return hurdles: that would eventually pressure data-center buildout, power and networking demand, and supplier order visibility—not just software multiples.

Over the next 1–3 months, watch real yields, Fed pricing, and company commentary on AI monetization and capex budgets. A pause in rate pressure could revive high-beta AI names, while persistent inflation paired with capex scrutiny would broaden downside beyond valuation. Over 6–18 months, returns should increasingly separate firms converting AI spend into measurable revenue or productivity from those relying on investment plans and narrative.

The contrarian case is that the macro shock may be priced faster than AI earnings expectations adjust; if buyer budgets and supplier guidance hold, indiscriminate de-risking could create relative-value opportunities. There is not enough evidence here to justify a wholesale short of AI exposure. The thesis weakens if real yields retreat and earnings revisions remain positive; it strengthens if buyers trim capex or AI-linked order growth and guidance roll over.

Key Decisions for Investors

• Reduce concentration rather than exit the theme: use strength to trim the highest-duration, least-proven AI monetization exposures and keep core positions in businesses with current cash generation and verifiable demand.
• Consider a relative-value pair: long profitable AI infrastructure exposure against a smaller short in speculative, long-duration AI application exposure. Size the short conservatively; reassess if real yields fall or application revenue evidence improves.
• Treat persistent inflation as a catalyst, not a standalone short signal. Track real yields and Fed-rate expectations over the next 1–3 months; avoid adding broad AI downside hedges unless rates rise alongside weakening capex or guidance.
• Set a thesis check on large technology buyers’ capex plans and supplier order commentary. A material budget reduction or downward guidance revision would support further de-risking; stable budgets and firm orders would argue against extending shorts.