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

AI-linked semiconductor winners are cooling off: the iShares Semiconductor ETF (SOXX) is up ~80% YTD but has fallen double-digits over the last seven weeks. In contrast, software is catching bids, with the iShares Expanded Tech-Software ETF (IGV) up double-digits over the same period as investors rotate from AI hardware to AI software beneficiaries (e.g., SaaS). The article argues investors shouldn’t sell chips outright, citing continued strong AI-chip earnings growth and reasonable valuations for some names (e.g., Nvidia forward P/E ~22.9, PEG ~0.55), though it flags that the rotation could reverse.

Analysis

This looks more like a positioning unwind than a true fundamental regime change. The first-order winner is software with perceived AI monetization leverage, but the second-order winner is actually the most scarce growth names inside software — MSFT, PANW, and PLTR — because they can absorb capital without needing the same incremental capex intensity as hardware. The losers are the most crowded, valuation-sensitive semi names, especially those where the market has already priced in flawless execution; INTC is the clearest example of multiple fragility if earnings momentum stalls.

The key issue for chips is not demand disappearing, but the marginal buyer getting more selective after a huge run. That means SOXX can underperform even if NVDA keeps compounding, because basket investors will rotate out of the entire group when they want lower-duration exposure. By contrast, software can rerate quickly if investors decide AI spend is finally turning into budgeted enterprise software revenue rather than just vendor capex. The setup favors relative-value rather than outright macro-beta longs.

The contrarian read is that the move may be overstated: semis and software are not substitutes over a 6-18 month horizon; they are different legs of the same AI buildout. If hyperscaler capex guides stay elevated and order books do not roll over, any SOXX drawdown should attract dip buyers. The near-term falsifier is a weak set of AI infrastructure guides over the next 1-2 earnings cycles; the medium-term falsifier is software monetization missing enough to force multiple compression back to fundamentals.

Key Decisions for Investors

• Tactically pair long MSFT/PANW against short SOXX for 4-8 weeks; this expresses AI monetization > AI buildout and should work if the rotation is flow-driven rather than fundamental. Risk/reward is favorable if semi multiples compress another 5-10% while software recovers 5-8%.
• Stay long NVDA, but use it as the higher-quality hedge inside semis rather than adding to SOXX. NVDA is the name most likely to hold up if capital rotates within hardware, but trim only if forward estimates stop rising.
• Avoid chasing INTC at current levels; it is the most vulnerable to multiple compression if the market keeps rewarding growth-duration elsewhere. A 1-3 month underperform versus MSFT/PANW is the cleaner expression than a broad tech short.
• If you need a cleaner sector bet, buy IGV/XLK-style software exposure on weakness, not strength, and monitor next earnings season for proof that AI is actually expanding software billings. If billings or RPO decelerate, exit quickly.
• Watch SOXX around the next hyperscaler capex print: a re-acceleration in capex guidance would falsify the bearish rotation thesis and likely trigger a sharp reflexive rebound in semis over the following 2-6 weeks.