Goldman Sachs notes shift in fund positioning towards semis and away from software

Goldman Sachs says hedge funds and mutual funds continued rotating out of software and into semiconductors at the start of Q2 2026, with mutual funds holding their lowest software exposure since at least 2012 and hedge fund semiconductor weight at a record high. The SOX is up 72.3% year to date while the iShares Expanded Tech-Software Sector ETF is down 11.1%, underscoring a sharp AI-driven divergence between winners and losers. The backdrop remains risk-on for semis but pressured by higher global rates, surging oil prices, and Middle East tensions.

Analysis

The positioning message is more important than the headline move: semis are becoming the crowded, consensus expression of AI monetization while software is being repriced as a low-duration cash flow trap. That creates a second-order setup where the leaders can keep working even on mediocre fundamentals, but the marginal buyer is now increasingly flow-driven rather than fundamental, which raises the odds of sharp factor reversals if rates back up or AI capex growth slows.

Within semis, the best relative trade is not the highest-beta AI narrative names but the capital-expenditure enablers: equipment and process-control beneficiaries should see demand hold up longer than headline GPU-exposed names because hyperscalers can defer some compute purchases but cannot easily pause fab and node transitions. That favors names tied to wafer-fab intensity and EUV ecosystem spend, while also leaving room for a catch-up bid in selective mature-tech beneficiaries as investors rotate from "platform winners" to "picks-and-shovels".

The underappreciated risk is that the current leadership becomes self-defeating if real yields continue higher. Semis are still long-duration assets; if markets reprice terminal rates up by even 25-50 bps, the highest multiple cohort will likely de-rate fastest, and the crowded long book in semis becomes a source of air pockets. Software may stop falling well before it becomes a true outperformer, but a stabilization in rates could produce a violent mean reversion because positioning is now so one-sided.

On stock-level flow, Microsoft looks vulnerable as the most obvious source of funding for both hedge funds and mutual funds: it is widely owned, less scarce, and more exposed to the "AI is expensive and late monetizing" debate than the semiconductor enablers. Apple and Meta are better relative hedges inside big tech because they combine balance sheet support with less direct AI valuation scrutiny, while Intel and SiTime can work tactically as lower-consensus semi exposure if investors keep chasing broad semiconductor beta rather than just the most obvious AI winners.

Key Decisions for Investors

• Long LRCX / AMAT / ASML basket versus short MSFT for 4-8 weeks: best risk/reward is in equipment names that monetize the capex cycle, while MSFT is the most likely funding source if positioning mean-reverts; stop if real yields break lower and software starts outperforming.
• Buy INTC on a tactical 1-2 month basis versus short an equal-dollar software basket: if the market keeps rewarding semiconductor breadth over software, the upside comes from lagging semi catch-up rather than crowded AI leaders; trim if semis lose relative strength versus Nasdaq.
• Sell downside protection on AAPL or META against long semis for a market-neutral AI rotation trade: both have less direct exposure to the software-disruption narrative and can act as ballast while semis remain the preferred factor; risk is a broad de-risking in growth.
• Avoid chasing SMCI and the most crowded AI beta for new longs; use rallies to fade into strength via short-dated call spreads or pairs against equipment names, because multiple expansion is most fragile where expectations are already maximal.
• Set a 2-4 week trigger on U.S. 10Y yields: if yields rise another 25 bps, reduce gross in high-multiple semis; if yields fall, cover software shorts first, as that is the fastest channel for a squeeze given current underownership.