After hedge funds' record AI stock selloff, Goldman Sachs recommends selective allocation to AI assets

Hedge funds' concentrated deleveraging and the decline in AI trade crowding have created buying opportunities in some U.S. tech stocks, but Goldman Sachs remains cautious about a rebound. Its strategists recommend selectively allocating to AI assets, with the core rationale being that earnings expectations for leading companies have not yet been revised down; however, continuously swelling capital expenditure and still-unclear investment returns leave it uncertain whether this correction can become a good entry point.

This week, AI-related assets suffered a sharp selloff. Goldman Sachs' Hedge Fund VIP basket posted its worst single-day performance since 2011 on Thursday, and Goldman's prime brokerage division also saw its largest tech stock selloff since September 2024. Meanwhile, the S&P 500 ex-AI constituents index outperformed the overall S&P 500 year-to-date for the first time, as the AI outperformance advantage that had persisted for nearly three years clearly weakened.

Behind the selloff is the market beginning to re-examine the costs and returns of AI investment. Meta raised the lower bound of its 2026 capital expenditure guidance to $135 billion, free cash flow turned negative, and the stock fell 7% on Thursday. This shows that revenue beating expectations is no longer enough to offset investors' concerns about continuously rising capital expenditure.

Hedge funds accelerate deleveraging, tech stocks see concentrated reductions

Goldman Sachs Prime division data show that fundamental long/short fund managers' beta-adjusted net leverage fell 3.5 percentage points this week and has fallen 10.3 percentage points cumulatively since October, dropping to a one-year low and sitting in the 14th percentile of the distribution over the past three years. Over the past 8 trading days, hedge funds deleveraged on 7 days, with the pace of adjustment exceeding any period in all of 2025.

Thursday's deleveraging ranked fifth for the year and was entirely driven by long selling. Fundamental long/short funds' net leverage ratio fell to 47%, in the 1st percentile of the three-year lookback range, indicating that funds are concentrating on cutting risk exposure.

Goldman Sachs' Tony Pasquariello said in a Friday report that, judging from client communications, this selloff has hit the market's core holdings, and Thursday's reduction in tech stock exposure was the largest single-day scale in the past two years.

Fund flows also showed clear divergence. On Thursday, the Philadelphia Semiconductor Index fell 6.7%, yet 337 S&P 500 constituents rose, with the index overall down only 1.2%. Meanwhile, the S&P 500 ex-AI constituents index rose 6.5% in October and is on track for its best monthly performance in nearly three years. AI stocks' advantage relative to the broad market is narrowing, and the market is no longer willing to pay a premium indiscriminately for AI-related assets.

The ebbing of the AI trade is particularly evident in the performance of Goldman Sachs' related portfolios. Goldman Sachs' AI and AI Risk Hedging basket (GSPUARTI) peaked at 583.45 points on June 22 and last quoted at 261.71 points, a drawdown of about 55% from its high. Goldman Sachs' High Beta Momentum basket (GSPRHIMO) has also given back all of its year-to-date gains, down about 43% from its June high and retesting its July low.

Goldman Sachs' AI Beneficiaries basket (GSTMTAIP) relative to the S&P 500 ex-AI constituents index (SPXXAI) fell to its worst single-day performance since the DeepSeek shock, and broad AI stocks have also fallen back to summer lows relative to the broad market.

However, Goldman Sachs strategist Louis Miller believes that position adjustments and stock price declines do not mean the AI investment thesis has become invalid. He offers three reasons for buying on dips: first, investors' momentum exposure has fallen to its lowest level since June 2025; second, the scale of this correction is relatively large, and the risk-reward ratio of some assets has improved; third, earnings expectations for AI leaders remain stable.

Among these, earnings expectations are the most critical support. Goldman Sachs points out that earnings expectations for hyperscalers have not been revised down since July, and the data center equipment sector has shown no signs of slowing earnings growth. If earnings remain stable while stock prices have already fallen sharply, some AI assets may have room to rebound.

But the clearing of positions does not mean risks have been fully released. Miller also acknowledges that although momentum exposure has declined, it remains in the 73rd percentile of the historical distribution, indicating that related trades have not completely shed their crowded status.

Goldman Sachs' recommendation is not a broad bullish call on AI, but rather selective positioning while retaining hedges. On one hand, the bank is optimistic about rebound opportunities in some AI beneficiary stocks; on the other, it is wary of non-AI stocks rising too fast in the short term, and is shifting investment focus to areas such as behind-the-meter power generation and self-built campus power generation, while avoiding cyclical and low-quality stocks.

This strategy reflects Goldman Sachs' judgment that opportunities in the AI supply chain are diverging: position clearing and stock price pullbacks have created entry opportunities, but that does not mean the entire sector has bottomed. Pasquariello candidly admits he has "limited conviction" on market direction, and points out that there is a solution path for computing power demand (POWER), but the capital cost (CAPITAL) problem still has no answer.

As AI capital expenditure continues to expand, the market's focus is no longer just on whether cloud computing providers will continue to increase spending, but on whether these expenditures can be converted into sufficient revenue, profit and free cash flow. Meta's stock price reaction shows that even with strong business growth, if investment returns cannot keep pace with spending expansion, valuations may still come under pressure.

Historical performance also reminds investors to remain cautious. Goldman Sachs data show that after a 10% drawdown in momentum portfolios, the median return over the next three months is 16.4%; but the rebound after the March 2026 low was ultimately fully given back, and after bottoming in July, the portfolio fell a further 15.4% in the following month.

Next week, Microsoft and Alphabet will release earnings successively within hours after the Federal Reserve announces its interest rate decision, with Amazon following on Thursday. The results will further test the return capacity of AI investment. With the contradiction between capital expenditure and cash flow not yet eased, Goldman Sachs' strategy is to capture pullback opportunities in some AI assets rather than betting on a broad rebound across the entire sector. Risk Disclosure and Disclaimer

Markets carry risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation or needs of individual users. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular circumstances. Investment based on this is at the user's own risk.