The U.S. stock market bull run of "only one trade left"

The U.S. stock market is standing at an unsettling high: the index is at a record high, but only one theme is supporting it. Tony Pasquariello, head of hedge fund coverage at Goldman Sachs, admitted in his latest client report that the current U.S. stock rally is "basically just one big trade"artificial intelligence. He maintained his broadly bullish stance on U.S. equities, but his language has clearly turned more cautious, no longer willing to assert that the next 5% move will be upward. At the same time, Goldman Sachs data shows that about one-sixth of S&P 500 constituents have fallen more than 50% from their highs, while the index itself is just a step away from a record high. This divergent structure is fueling deeper market concerns about concentration risk. Interest rates keep climbing, market breadth is deteriorating, and positioning is extremely concentrated in the AI sector. Multiple stress signals are flashing at once, and the bulls' line of defense now rests only on whether Nvidia and Micron can deliver on expectations in the upcoming earnings season. Extreme divergence: a bull market that exists only in certain corners The S&P 500 briefly touched a record high of 7,818 this week before falling back to 7,765 on news of an OpenAI revenue warning. On the surface, the bull market flag is still flying, but the internal structure has become highly distorted. Data compiled by Goldman Sachs colleague Brian Garrett reveals the depth of this divide: over the past three months, about 45% of S&P 500 constituents have been negatively correlated with the index itself, a record high. Goldman Sachs' Adrien Simonet further quantified it: since August 27, the S&P 500 as a whole has risen 1.28%, while the "S&P 500 ex-AI" (SPXXAI), which strips out AI-related stocks, has fallen 5.19%. The rolling 30-day gap between the two is close to its largest since the AI trade began in January 2023. Meanwhile, the Russell 2000 has underperformed the Nasdaq on 16 of the past 20 trading days, lagging by about 9 percentage points. Pasquariello characterizes this market as an "extremely narrow rally"U.S. large-cap tech stocks are racing ahead like a locomotive, while rate-sensitive and cyclical sectors remain under persistent pressure. Positioning is "clean," but risk is concentrated in one place Despite the index being at a record high, overall market positioning is surprisingly light. Goldman Sachs Prime Book data shows that in September, hedge funds' net exposure to U.S. equities was at one-, three-, and five-year lows; net leverage in fundamental long/short strategies fell for a third straight week to its lowest level since Trump announced reciprocal tariffs last year; and Goldman Sachs' sentiment indicator is hovering near multi-year lows. Pasquariello's interpretation: being underweight U.S. equities has itself become a crowded trade, and for institutions worried about missing an S&P 500 run to 8,100, the cost of call options is relatively cheap. However, low net exposure does not equal low risk, because gross exposure remains elevated and the residual positioning is extremely concentrated in the same direction. In September, tech was the only sector that hedge funds bought on a net basis, with purchases at their largest scale since February 2025. Goldman Sachs Prime Book data shows that net exposure to the "Mag 7" now accounts for about 22% of total U.S. equity exposure, the highest on record since data began in early 2022; semiconductor positioning stands at 12%, double the level at the start of the year. Simonet added that leveraged semiconductor ETF assets have reached about $115 billion, which would create a mechanical amplification effect in a market downturn. On Thursday, after the OpenAI news hit, Goldman Sachs' TMT trading desk recorded more than $1 billion in net selling in semiconductors, AI, and large-cap techa small-scale preview of what could come. Earnings growth looks strong, but is highly dependent on a handful of companies With earnings season about to begin, Pasquariello raises a third point: who wants to short in the face of an expected 28% year-over-year earnings growth for the S&P 500? That figure is still being revised up from last week's 27% forecast. But this, too, is a story that "depends on where you look." Goldman Sachs' Ben Snider's Q3 earnings preview shows that earnings growth for the median S&P 500 stock has slowed from 14% in Q2 to 9% in Q3. Broken down by sector, tech growth is as high as 64%, while energy reaches 122%. Goldman Sachs' Wilson estimates that 68% of S&P 500 earnings growth this quarter will come from the top 10 contributors, up from 48% last quarter; Micron and Nvidia alone account for half of that 68%. More notably, Goldman Sachs' 27% full-year earnings growth forecast is already a cycle high in its own model, and growth will slow in each of the next three quarters; hyperscaler capex growth (estimated by Snider at 116%) may also peak this quarter. The peak in growth and the peak in capex arriving together, combined with stock prices and positioning at historical extremes, is the kind of combination that tends to be described only in hindsight as "having been foreshadowed all along." Rates escalate from headwind to systemic risk This is the most notable shift in Pasquariello's language. Ten days ago, he characterized rising rates as a "headwind" for stocks, but explicitly ruled out the possibility of it being "disruptive." In his latest report, his wording has escalated: rising rates are becoming a "greater risk" for sovereign and corporate debt. Market data confirms this shift. On Thursday, the 30-year Treasury auction cleared at a yield of 5.618%, the highest since August 2000; the 10-year Treasury yield closed at 5.23%. Goldman Sachs economists expect the "Warsh Fed" to raise rates again in December after hiking to 3.75%-4% in September. Goldman Sachs' Rikin Shah calculates that of the 108-basis-point selloff in 10-year yields, 98 basis points came from real rates, and the 30-year real yield of 3.33% is already near the upper end of its range since 2010. His conclusion: the Treasury market is looking for one of two tipping pointsthe level at which AI financing demand begins to become sensitive to rates, or the level at which some other part of the economy breaks first. On the corporate credit side, Goldman Sachs credit strategists note that USD investment-grade credit has returned -292 basis points year-to-date, and they are focused on software issuers facing refinancing pressure from maturing loans. Goldman Sachs economists estimate that current rate levels could drag on economic growth by 0.5 percentage points in 2027 through housing, consumption, and capex channels. Simonet's warning is especially direct: a term premium shock will not distinguish between good and bad AI stories; it will reprice all ten-year cash flows at once. He sees the true left-tail risk as the Fed being forced to hike more than expected to preserve long-end credibility, with CPI data as the near-term trigger. Bull and bear sides, the same bet In his report, Pasquariello quotes an investor he clearly holds in high regard, and it is the line he considers most important in the entire piece: "You know, this is basically just one big trade." He does not dispute it. Lining up six judgments side by sidetrend, positioning, earnings, rates, concentrationreveals that they are merely different facets of the same observation: the trend is up, because of AI; positioning is light, except for AI; earnings growth depends on AI and its energy demand; rates are rising, partly because of AI's financing needs. The bull case and the bear case both point to the same target. Pasquariello also offers a friendly "tech and energy in tandem" perspective: over the past 11 years, a 50/50 portfolio of the two sectors has posted positive returns in 10 of them, with an average annual return of 18.7% and a Sharpe ratio of 1.2. But this, too, is describing the same trade and its energy constraint. Seasonality and midterms: the final variable Pasquariello's sixth judgment is relatively brief: Q4 seasonality is favorable, but he expects the November 3 midterm elections to bring a wave of volatility. Goldman Sachs' Alec Phillips notes that prediction markets give Democrats a more than 90% chance of retaking the House, 65% for the Senate, and a generic ballot lead of 8.9 percentage points, with a "sweep" now the market's base case. Meanwhile, the VIX closed at 15.4 on Thursday, but Goldman Sachs' volatility trading desk reported that buyers purchased about $10 million in vega of year-end S&P 500 puts over roughly six hours that day. Someone is buying insurance against "friendly seasonality." A market with one exit Pasquariello ultimately offers three conclusions: the bull market holds, but it depends on where you look; near-term risk-reward is unclear, and he advises "watching the speed limit"; stick with the two fastest horsesthe U.S. and Japan; and hedge with a combination of long equities paired with short rates. This hedge itself is telling: if the recommendation for holding stocks is to simultaneously short bonds, then it is effectively saying the two will fall together, and "shorting rates" only profits if yields keep risingwhich is precisely what the 85% of constituents already in deep drawdown can least withstand, and which will ultimately become a financing-cost pressure on the other 15%. The bull scenario requires Micron and Nvidia to deliver, hyperscaler capex not to peak, the Fed to stop after two hikes, and someone willing to buy long bonds at a 5.6% yield. The bear scenario requires only one of those to fail. One big trade is wonderful when it is rising. But by definition, it also means the entire market has only one exit.