After nine consecutive weeks of gains in the U.S. stock market, undercurrents are emerging! Goldman Sachs warns: the market is becoming increasingly crowded, concentrated, and leveraged around AI, with downside protection nearly gone.

After nine consecutive weeks of gains in the U.S. stock market, undercurrents are emerging! Goldman Sachs warns: the market is becoming increasingly crowded, concentrated, and leveraged around AI, with downside protection nearly gone.

```

The S&P 500 Index has risen for nine consecutive weeks, but beneath the calm surface, structural risks are quietly building up.

On June 2, Goldman's Sales and Trading team top trader Lee Coppersmith issued a warning in his latest report: The rally at the index level appears smooth, but things beneath the surface are becoming "increasingly unsettling." He pointed out that the market's AI bets have evolved from being fundamentally driven to a self-reinforcing loop strengthened by market structure itself—positions are more crowded, leverage is higher, concentration is stronger, yet the cost for downside protection has fallen to historic lows.

The most telling indicator of this contradiction is single-stock skew: The average one-month put/call option skew of S&P 500 constituents has fallen to the lowest level in Goldman's data history, meaning investors are abandoning downside protection en masse in favor of chasing upside exposure—yet at the same time, realized volatility at the individual stock level continues to climb.

Goldman’s US Vol Panic Index has also dropped to a near two-year low.

Meanwhile, Goldman Prime data shows hedge funds’ total leverage rose again this week by 2.1 percentage points to about 323%, a five-year high.

Index calm masks intense factor volatility, extreme divergence between surface and underlying

Since ChatGPT's launch on November 30, 2022, the S&P 500 Index has gained about 85%, having gone through the regional bank crisis in March 2023, Japan’s rate hike shock in August 2024, the so-called "reciprocal tariff" turmoil in April 2025, and this year’s US-Iran war, recovering and hitting new highs each time.

However, Coppersmith stresses in the report that what matters most now is no longer the gains themselves, but the growing divergence between the calm at the index level and the intense rotation among factors, positions, and individual stocks.

The Nasdaq 100 Index (NDX) rose about 10% in May, marking the first consecutive two months with double-digit gains since 2009—a 32% rebound from the March low. But behind this dazzling number, Goldman’s TMT momentum pair index (GSTMTMOM Index) has seen more than 25 days this year with daily swings of ±5%; last year there were only 6 such days.

The software sector rose again about 8% last week, driven by favorable positioning, better-than-expected earnings, and renewed optimism about AI monetization "velocity." Yet even within the sector, the divergence is obvious—Goldman's trading desk continues to observe a clear split between data infrastructure and cybersecurity as perceived infrastructure winners, and the more challenging SaaS business models.

AI positions self-reinforcing, leveraged ETF sizes double, increasing market fragility

Coppersmith highlights a forming structural risk: The AI trading drive no longer comes solely from fundamentals; the market structure itself is creating reflexive reinforcement.

Global leveraged/inverse single-stock ETF assets under management have exceeded $60 billion, doubling just since early April this year. Products linked to SK Hynix, Samsung, and broader memory exposure have seen explosive inflows, as investors continue to chase concentrated exposure to the AI construction wave.

These products mechanically short Gamma, meaning continued inflows amplify momentum, squeeze effects, and intensify rotation at the factor level.

Meanwhile, hedge fund position data confirms this crowding. Goldman Prime data shows hedge funds are net buyers of consumer discretionary stocks for the fourth consecutive week, across all major regions, despite the sector’s allocation still near its five-year low; the tech sector had its largest weekly deleveraging in over a month, though overall exposure remains near five-year highs.

Downside hedging costs drop to record lows; "reverse dispersion" hedging regains attention

The correlation regime implied by current market pricing is very mild, which is precisely what Coppersmith worries about most.

S&P 500 implied correlation is near all-time lows, with index volatility much lower than component stock volatility. This pattern historically provided the theoretical basis for "reverse dispersion" hedging—shorting single-stock volatility while going long index volatility.

This strategy works when the market shifts from stock differentiation to macro panic and correlations surge—examples include the global financial crisis, the COVID-19 shock, 2018’s "Volmageddon," the August 5, 2024 Japan market stress event, and last April’s "reciprocal tariff day."

However, Coppersmith admits that in recent years this hedging strategy has been tough to hold, since what dominates is the opposite pattern: leaders narrow, correlations keep falling, stocks swing wildly but the index remains well-behaved.

His summary is direct: The market is increasingly convinced about AI, increasingly concentrated, more leveraged, and more inclined to pay for upside convexity than to buy downside protection. Macro panic has subsided, but single-stock idiosyncratic risk has not disappeared—indeed, single-stock skew pricing is headed in the opposite direction.

Earnings growth provides fundamental support, but uncertainty remains

Still, Coppersmith acknowledges this rally is not simply valuation expansion or blind speculation—earnings growth is truly the main driver.

Prior to ChatGPT's launch, Goldman's US strategy team predicted 2023 S&P 500 EPS at about $224; now, the team forecasts 2026 EPS will reach $340, about 52% above prior pre-AI expectations. Since November 2022, roughly two-thirds of the S&P 500’s gains can be explained by earnings growth.

At the AI industry chain level, Goldman's tech research team recently estimated that Agentic AI could push token consumption up to about 24 times current levels, creating massive incremental demand for GPU, CPU, power systems, cooling infrastructure, network architecture, and full data center redesign.

Coppersmith points out that the market still doesn’t fully know where AI's value will ultimately settle in the industry chain, and this uncertainty itself may be helping sustain the breadth of global AI capex enthusiasm.

He also notes in his report that historically, major technology shocks tend to intensify concentration, not diversification. Firms best equipped to deploy intangible capital at scale—data, computing power, distribution channels, software ecosystem, and network effects—usually capture most of the economic value in the end.

Risk Warning and DisclaimerThe market involves risk, and investment needs caution. This article does not constitute personal investment advice, nor does it take into account any individual’s specific investment goals, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in this article fit their own circumstances. Acting accordingly is at your own risk. ```