Goldman Sachs Warns: Mag7 Positioning Crowdedness Hits Record, AI Trade Asymmetry Reverses
nashnova research
Goldman volatility trader Shawn Tuteja exits his AI long trade, citing Mag7 net exposure at 22% of total US exposure — an all-time high — while long-end rates keep climbing; the trade once defined by 'nobody owns it, options are cheap' has become the market's most crowded bet.
Why did a Goldman trader just flip bearish?
Shawn Tuteja, Goldman's volatility trader, closed out the AI long trade he had been championing.
His core call: the asymmetry — the risk-reward skew — in tech/AI has reversed. The macro backdrop is harder, and positioning is no longer a tailwind.
This means → buying AI used to be "under-owned, cheap, big upside." Now it is "over-owned, expensive, downside underpriced."
One index — but are there really two markets inside it?
Since August 27, the S&P 500 has gained 1.28%, while Goldman's "S&P 500 ex-AI" index (SPXXAI) has fallen 5.19% — a ~6.5 percentage-point gap, moving in opposite directions.
In September alone, the cap-weighted S&P 500 beat the equal-weight index by ~4.5 points, one of the widest monthly gaps in over 25 years. Fewer than 25% of constituents outperformed the index; the median stock sat 17% below its all-time high.
In plain terms = the index is rising, but most stocks are falling — the gains come from a handful of mega-caps. Goldman macro strategist Tony Pasquariello calls Big Tech "both the sword and the shield."
Why is all the money piling into just seven stocks?
Goldman's Financial Conditions Index has tightened ~40 basis points since late June, driven almost entirely by long-end rates. The 10-year Treasury yields 5.28%; the 30-year, 5.66%.
The SOFR — secured overnight financing rate, the benchmark for short-term interbank lending — Dec 2026/Dec 2027 spread has steepened ~50 bps since late August, as the market rapidly prices in more rate hikes.
This means → the higher rates go, the more pain for companies tied to the real economy. The only shelter for fast money is the few tech giants seen as "immune to high rates" — bond-market stress is forcing capital into Mag7.
Just how crowded is the positioning?
Goldman prime-brokerage data: Mag7 net exposure is near 22% of total US exposure — the highest since records began in early 2022.
Semiconductors and semi-equipment account for roughly 12% of total US exposure, down from a June peak of 14% but double the ~6% level at the start of the year.
US tech saw its largest monthly net buying since February 2025 in September — and was the only sector net bought on Goldman's prime book.
This reflects not genuine conviction but fear-driven concentration — passive crowding is more fragile than active chasing.
Is any AI trade still worth holding?
The only AI direction Tuteja still favors: plays tied to power infrastructure — what he calls the "needs-a-power-plant" AI trade.
This means → in his framework, AI certainty has shifted from "who has the best chip" to "who supplies AI with electricity" — the hardware side is too crowded; infrastructure positioning has room.
In plain terms = whether Big Tech can keep playing "sword and shield" is the market's next key test. If bond-market pressure persists, the crowded positioning itself becomes a downside amplifier.
市场有风险，内容仅供研究参考，不构成投资建议。