According to Woofun AI, the divergence in the US stock and bond market is approaching a tipping point. Goldman Sachs (GS.US) strategist Louis Miller issued a severe warning: the credit spreads of AI hyperscale cloud computing companies have expanded to the highest level in the year, and corresponding stock prices have not fully reflected this pressure. The market faces a dualistic option where either the bond market is “too pessimistic” or the stock market still has about 15% downside undigested.
Warning signals in the credit market are spreading rapidly from the core to the margins. The Goldman Sachs AI Hyperscale Cloud Computing Credit Spread Index (GSUCHS30) has reached the widest level in the year. Looking back at history, the last time the spread was close to the current value, the relevant stock price was about 15% lower than it is now.
Meanwhile, Goldman Sachs's broader basket of AI beneficiaries has accumulated a cumulative decline of about 25% from the June high compared to the S&P 500 index, which excludes AI components. The pressure isn't limited to large cloud computing companies; their credit default swaps (CDS) have just hit a record high. Lower in the capital structure, the risk is even more severe: Of the 20 junk rated AI data center bonds, 7 rely on a single tenant CoreWeave (CRWV.US) as a repayment source, and CoreWeave's own bonds have fallen to a difficult level.
According to data compiled by Woofun AI, almost none of the AI sector was spared this week. NeoClouds fell 24.8% on the 5th, optical networks fell 21.4%, and memory chips fell 16.6%. The only theme that recorded significant increases was “AI-AT-risk” (traditional industry facing the impact of AI), which rose 3.5%.
The core source of the current pressure is a shift in capital expenditure growth. Miller warned that the widening spread across the sector is forcing investors to confront a question: what will happen when AI spending growth slows. AI capital spending is likely to have peaked, and financing sustainability and return on capital have replaced recent revisions to earnings expectations and are at the heart of the market debate. The year-on-year capital expenditure growth rate of hyperscale cloud computing companies peaked at 116% in the third quarter of 2026, and the market generally expects it to slow quarterly thereafter, falling to about 20% by the end of 2027.
The impact of this trend on S&P 500 earnings has been clearly estimated: Goldman Sachs estimates that capital expenditure (net value after depreciation) of hyperscale cloud computing companies contributed about 11 percentage points to S&P 500 earnings growth in 2026, falling to 7 percentage points in 2027, and turning -1 percentage point by 2028 due to depreciation and amortization of large data centers. In other words, capital expenditure, which once boosted profits, will in turn become a drag once the growth rate slows down.
Internal market indicators simultaneously alert, and pressure on the AI sector is not an isolated phenomenon. On the emotional side, put option trading volume reached “the second highest this year, after June” on Friday. On the risk appetite side, Goldman Sachs Beta Factor Hedging Package (GSP1BETA) fell 15% from recent highs, and high-beta cyclical stocks have fallen below the upward trend channel. On the growth forecast side, the economic growth forecast implied by Goldman Sachs's cyclical stock to defensive stock ratio (GSP1CYDE, excluding commodities) has fallen to about 2%, indicating that the degree of pessimism about growth in the market has exceeded the level supported by fundamental data. On the leverage and position side, the beta exposure held by hedge funds is higher than the level implied by the level of leverage — the beta-adjusted net exposure is at the 67th percentile in five years, while the net leverage ratio is only at the 17th percentile. Miller also pointed out that financial conditions have “changed from a positive driver supporting growth to a negative drag”, higher interest rates and wider interest spreads are directly impacting debt-dependent stocks, and Goldman Sachs's debt-sensitive stock basket (GSX1DEBT) continues to weigh on the market.
Despite this warning, Miller did not predict a complete collapse of the market. He believes that the market is still absorbing pressure through sharp declines on a rotating rather than exponential level, and proposed five trading strategies. 1. Expand the “AI-enabled” broad-based market. The S&P 500 Index (SPXXAI), which excludes AI components, has outperformed the S&P 500 by 586 basis points since October 6, and the correlation between the AI sector and the rest of the market has turned negative. 2. Increase the quality factor but pay attention to technical risks. Goldman Sachs's Quality Factor Hedging Package (GSP1QUAL) has risen 20% in three months, and the RSI has already exceeded 80, and there is a risk of a sharp reversal; the European Quality Factor (GSPEQUAL) has risen 7.5% in the past month.