Five weeks ago, this Goldman Sachs trader strongly recommended buying AI options; now, he warns of "record-breaking AI congestion."
Five weeks ago, Goldman Sachs derivatives trader Shawn Tuteja strongly recommended buying call options on AI stocks, citing "no open positions and cheap options," a trade that subsequently yielded substantial profits. Now, he has reversed his judgment—the AI sector has gone from being ignored to the most crowded trade in history, and his latest report is intriguingly titled: "Not So Clear Now."
The S&P 500 closed at a record high of 7818.93 on Tuesday, while the Nasdaq 100 also hit a record high of 31224. However, a recent research report by Tuteja shows that Mag7 net exposure has risen to the highest level ever recorded in Goldman Sachs' database (dating back to early 2022), accounting for approximately 22% of total US equity exposure. He warns that the initial logic for buying AI—"low valuation, low holdings"—has largely disappeared, leaving only a bullish bet that requires continued fundamental performance.

At the market level, the impact of this judgment is direct and clear: Goldman Sachs advises clients to shift their holdings from the already sought-after AI leaders to lagging sectors represented by SPXXAI (the S&P 500 index excluding AI stocks), while regarding energy infrastructure AI assets as one of the few exceptions that still have old logic to support them.
Bond market pressures drove funds into seven stocks
Tuteja analyzed the macroeconomic background behind this renewed crowding of AI trading. Since the end of June, the Goldman Sachs Financial Conditions Index has tightened by about 40 basis points, almost entirely due to rising long-term interest rates—credit spreads, the US dollar, and equities contributed negligibly. Meanwhile, the SOFR December 2026/December 2027 spread has surged by nearly 50 basis points since the end of August, indicating that the market has significantly priced in further rate hikes before the end of 2027. Currently, the 10-year US Treasury yield is at 5.28%, and the 30-year yield is at 5.66%.
Faced with a comprehensive tightening of interest rates, quick money funds have adopted a highly consistent strategy: buying a few tech stocks that seem immune to high interest rates and selling all assets linked to interest rate sensitivity. "Clients have indeed repurchased tech stocks due to this macroeconomic uncertainty," Tuteja wrote.
Data confirms this assessment. In September, US tech stocks saw their largest monthly net inflow since February 2025, making it the only sector to achieve net buying in Goldman Sachs's block trades book. Simultaneously, retail investor funds also made a strong comeback—total assets in the US long leveraged semiconductor ETF rebounded from a low of approximately $72 billion on July 29th to approximately $115 billion, a recovery of about 60% in ten weeks.
Record concentration means AI is no longer an undervalued deal.
The data cited by Tuteja clearly reveals the core of the current risks. Mag 7 net exposure accounts for approximately 22% of total US equity exposure, the highest ever recorded by Goldman Sachs; semiconductor and semiconductor equipment exposure accounts for approximately 12%, more than double the approximately 6% at the beginning of the year; Goldman Sachs' commodity brokerage book net leverage ratio is 46.4%, down 8.5 percentage points from the beginning of the year, placing it at the 0th percentile of the past year's data, while the long-short ratio is at the 1st percentile of the past five years.

Market breadth data is equally alarming. In September, the S&P 500 market capitalization-weighted index outperformed the equal-weighted index by approximately 4.5 percentage points, one of the largest monthly gaps in over 25 years; less than 25% of its components outperformed the index; the median stock price is down 17% from its all-time high; and the Nasdaq 100 outperformed the Russell 2000 in 14 of the past 16 trading days. Goldman Sachs' Tony Pasquariello described large-cap tech stocks as "continuing to play both the spear and the shield."
Jonathan Krinsky of BTIG pointed out that the Nasdaq 100 has outperformed the Russell 2000 for 14 consecutive days, "a feat previously only seen in early 2000," and noted that the Philadelphia Semiconductor Index's performance in 2000 is "continuing in an amazing way," currently around the top of that rally.
Neither of the two paths is "continuing to chase Nvidia's stock price."
Tuteja's assessment of the market's future path is the core of this report. He outlines two scenarios, both of which are highly consistent in their conclusions: if the macroeconomy improves, the remaining 490 lagging stocks will outperform the AI leaders; if the macroeconomy deteriorates, AI stocks will ultimately fall along with the broader market. In either scenario, chasing after already heavily held AI stocks is not the right strategy.
His preferred approach is to take long positions in lagging sectors with limited risk: buying 102% out-of-the-money call options on SPXXAI expiring in November, costing approximately 1.25%, which he describes as a convex right-tail hedging tool should the macroeconomic environment improve. It's worth noting that, from Goldman Sachs' perspective, average U.S. stock call options are now considered "tail hedging tools."
Morgan Stanley's Mike Wilson said over the weekend that the current poor market breadth "must be addressed in some way"; Deutsche Bank's Henry Allen was even more direct. Meanwhile, JPMorgan's trading arm turned "tactically bullish" again last week, but the basis for this confidence remains questionable given the high interest rate environment.
Hedge funds have few assets left to sell.
The report is not entirely bearish, as the other side of the positioning chart provides downside support. Although hedge funds are heavily invested in technology, their overall leverage is at a historical low, with the net leverage ratio at the 0th percentile of data from the past year. This means that even if the market falls, there is limited room for large-scale deleveraging.
Tuteja estimates that the implicit "buy zone" for the S&P 500 is around 7400 points, about 5.4% lower than Tuesday's close. He cites three reasons supporting the "relatively manageable" decline in the index: the overall leverage level of funds is already low, seasonal factors, and the reopening of the stock buyback window.
However, critics point out that this bottoming logic has a clear cyclical nature: the so-called "support" depends on hedge funds that just made a record increase in their holdings of Mag 7 and are willing to continue buying more shares after a 5% drop.
The only AI transaction that still has logical support: buying electricity.
Within the AI sector, Tuteja remains bullish on one sub-sector—AI-powered infrastructure. Valuations in this sector have fallen from approximately 26 times forward P/E at the beginning of the year to around 21 times, with a relatively light position and substantial catalysts. Goldman Sachs' AI Power Basket (GSENEPOW) rose approximately 5% on the day the report was released, while Constellation Energy (CEG) saw a single-day increase of 12% due to Google's announcement of integrating its nuclear power plant.
Tuteja's year-end trading strategy is to sell 85% out-of-the-money put options expiring on December 31, and use the premiums to fund the purchase of a 110%/135% call option spread, with a net cost of approximately 1.75%.
It's important to note that the midterm elections on November 3rd are a significant variable. Goldman Sachs options data shows that the market is pricing in an additional 1.69% volatility for the AI power sector, exceeding the S&P 500's 0.31% and higher than the corresponding figures for any party's policy basket. As Tuteja's colleagues warned last Sunday, the midterm elections are "likely not going to be a positive event for the market."
The bond market is the true arbitrator.
Tuteja believes that the core risk in the current situation still lies in the CPI data.
A rate hike in October was priced in by the market at a 64% probability before last week's non-farm payroll data release. The "moderate" data did not fundamentally change the argument for inflationary pressures, especially given that Brent crude oil was near $100 per barrel and the 30-year Treasury yield was at 5.66%. The last time the S&P 500 and the "S&P 500 excluding AI" showed such a large gap was in June, and holders of leveraged semiconductor ETFs still remember what happened in July.
Risk Warning and DisclaimerInvesting involves risk; please exercise 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 specific circumstances. Any investment decisions made based on this information are at your own risk.