Goldman Sachs Trading Desk: After $20 Billion Flees TMT, Recommends Buying the AI Dip in the Order of Power, Cloud, Inference, Cybersecurity

Goldman Sachs FICC and Equities trading desk's latest assessment holds that the sharp correction in tech stocks over the past two weeks was essentially a hedge-fund-led deleveraging event, not a signal of weakening AI demand—a judgment that directly determines whether it is worth entering the market now, and how to enter.

According to news from the Zhuifeng Trading Desk, Goldman Sachs FICC and Equities trading desk said in an October 26 report that over the past two weeks, net selling across the market in TMT has totaled about $20 billion, approaching the historical peak of $23 billion within two weeks during the July AI selloff. Meanwhile, Goldman Sachs' risk appetite index posted one of the largest single-day declines in history, falling to its lowest level since April last year; the implied volatility spread between the Nasdaq 100 and the S&P 500 widened to the 98th percentile, a near-five-year high. Goldman Sachs' trading desk judges that the selling pressure is concentrated at the positioning level rather than the fundamental level, crowded positions have been largely flushed out, and now is the time to selectively re-establish AI exposure.

Goldman Sachs recommends re-adding AI exposure in the following order: first Power, then Hyperscalers, then Inference and Cyber. In addition, the report explicitly lists the verification conditions for whether the judgment of "deleveraging rather than weakening demand" holds, with a verification window from October 28 to November 6, 2026, covering the FOMC meeting, the earnings week of Microsoft, Google, Amazon and Intel, and Oracle's analyst day.

A positioning event, not a demand event

Goldman Sachs prime brokerage data show that U.S. equities have seen net selling for a third consecutive week (net selling in four of the past five weeks), with single-stock deleveraging at its largest since late July. The industrials sector was sold by hedge funds at the fastest pace in more than six months, while consumer staples and real estate were the most net-bought sectors last week, continuing the defensive positioning characteristic.

Structurally, during the previous correction the market saw clear rotation flows from AI into defensive sectors; last week, although consumer, utilities and some financial stocks all rose, this rotation almost disappeared. Goldman Sachs' trading desk believes this structural difference is a key signal—capital is not leaving the stock market, but is concentrated in reducing crowded AI-related positions.

The performance of two AI baskets confirms this judgment: Goldman Sachs' TMT Broad AI Basket fell 4.1% last week, the TMT Mo' Basket fell 7.8%, while the Nasdaq 100 as a whole fell only 80 basis points for the week. Selling pressure is highly concentrated in crowded AI positions rather than spreading systematically across the entire market.

Crowding has been released, index-level drawdown limited

From quantitative indicators, Goldman Sachs' AI versus AI risk pair (GSPUARTI) has fallen 13% year to date, while the S&P index excluding AI components (SPXXAI) has outperformed the S&P 500 year to date—meaning the damage from this correction is almost entirely concentrated in crowded AI positions, with index-level damage far smaller than that of the AI baskets.

The volatility structure also signals a clearing of crowding. The implied volatility spread between the Nasdaq 100 and the S&P 500 has widened to the 98th percentile, a near-five-year high, and the volatility of volatility is at the 83rd percentile over three years. Goldman Sachs' trading desk noted that a genuine panic-driven deleveraging occurred last Thursday, with the risk appetite index posting one of the largest single-day declines in history; volatility then fell rapidly before VIX expiry but soon rose again, showing a more steady upward trend.

Why order matters: risk stratification on the financing side

Musk posted on social platform X over the weekend that the bond market has closed financing channels to almost all companies without revenue independent of AI data centers. Although SpaceX successfully completed $40 billion in GPU debt financing, it has an investment-grade rating and independent revenue from Starlink and launch operations. Musk said bluntly that more than 200 neocloud companies will face massive death and consolidation, with few able to survive.

In contrast, SK Hynix has bucked the trend on the strength of real cash returns. According to Bloomberg, SK Hynix plans to pay a special dividend of about 16 trillion won (about $12 billion) to distribute the roughly 20 trillion won gain on its Kioxia stake; the company is also considering a shareholder return plan totaling 50 to 60 trillion won, including a potential buyback. SK Hynix shares thus rose against the trend, forming a sharp divergence from the broad decline in AI-related stocks.

Four verification conditions, window through November 6

Goldman Sachs' trading desk explicitly listed the verification conditions for whether the judgment of "deleveraging rather than weakening demand" holds, with a window from October 28 to November 6, 2026:

① Positioning side: whether TMT net selling in the next two weeks converges to less than half of the July round, and whether the risk appetite index can stop falling and rebound;

② Volatility side: whether the implied volatility spread between the Nasdaq 100 and the S&P 500 falls back from the 98th percentile—if it continues to widen and hit new highs, it is more likely a fundamental repricing rather than a clearing of positions;

③ Outcome side: whether hyperscaler capex guidance is maintained—if guidance is cut and neocloud financing continues to be closed, the premise of "deleveraging rather than weakening demand" will be overturned;

④ Verification window: this week, companies representing 44% of S&P 500 market capitalization report earnings, led by Microsoft, Google, Amazon and Intel; the 10-year U.S. Treasury yield is flat at 4.99%, and crude oil fell 2%. This week's hyperscaler and memory semiconductor earnings (including SK Hynix, SNDK, STX) and Oracle's analyst day will provide further verification.

Goldman Sachs Global Investment Research also notes that global growth risks are currently skewed to the downside, and renewed escalation in the Middle East has pushed energy prices back to March-April highs, putting negative pressure on consumers and energy-importing economies; this macro backdrop constitutes an external risk factor for the above AI trade judgment.

~~~~~~~~~~~~~~~~~~~~~~~~

The above excellent content comes from the Zhuifeng Trading Desk.

For more detailed interpretation, including real-time commentary and frontline research, please join [Zhuifeng Trading Desk ▪ Annual Membership] Risk Disclosure and Disclaimer

Markets carry risk, 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.