The AI Trade Rotation: Money Is Moving Out of Chips and Into This

AI chip stocks have surged in 2026 (SOXX up ~80% YTD) but the ETF has fallen double-digits over the past seven weeks, while a software ETF (IGV) has risen double-digits over the same window—suggesting a rotation from chips to software. The article argues investors shouldn’t sell chips because AI data-center demand is still a tailwind and some chip valuations remain reasonable (e.g., Nvidia forward P/E ~22.9 and PEG ~0.55; Intel trades at ~79x forward earnings after being up ~170% YTD). Overall, it frames the move as valuation/positioning-driven and possibly reversible in coming weeks.

Analysis

This looks less like a true regime change than a crowded-factor unwind inside the AI complex. The first names to get hit are the most valuation-sensitive chip exposures, while software with clearer monetization and less immediate capex dependency can catch a bid; that favors MSFT, PANW, and to a lesser extent PLTR, while SOXX is the cleaner expression of de-grossing. The second-order effect is that AI infrastructure spend is still likely intact, but the market is temporarily paying more for proof of monetization than for raw compute capacity.

The key catalyst window is the next 1-3 months of earnings and hyperscaler capex commentary. If NVDA and adjacent suppliers keep showing accelerating order books, the rotation should fade quickly because the fundamental earnings delta in semis remains superior; if not, the group’s elevated starting valuations make it vulnerable to another leg of multiple compression. INTC is the fragile link: any disappointment there can spill over into the whole semiconductor basket because it reinforces the idea that the market has already priced perfection.

Contrarian view: investors may be overreacting to a short-term style rotation and underestimating how quickly semis can re-rate when guidance stays strong. The market is also still assuming software can monetize AI without showing margin leakage or longer sales cycles; that proof point is weaker than the current relative performance suggests. Net: this is a relative-value setup, not an all-out sector call, and the thesis breaks if AI capex or NVDA guide accelerates again.

Key Decisions for Investors

• Long MSFT/PANW basket vs short SOXX for 1-3 months: play the rotation while keeping beta-balanced; cover the short if SOXX reclaims relative strength after the next NVDA/hyperscaler print.
• Maintain core NVDA exposure, but hedge event risk into earnings with a put spread or collar; the thesis is invalidated if forward guide or data-center growth decelerates meaningfully.
• Avoid chasing INTC on momentum rallies: the stock is the most exposed to multiple compression if investors decide AI winners should be bought on earnings quality, not price performance.
• Watch PLTR as a high-beta software beneficiary rather than a pure fundamental re-rate; add only if commercial growth and operating leverage continue to beat in the next quarter.
• Set a 1-3 month alert on SOXX relative to IGV-style software proxies: if the spread reverses, the rotation trade is probably exhausted and should be unwound.