Panic signal! Goldman Sachs traders: AI credit risk has started to spread to broader markets

```

The credit risk of AI hyperscale data center operators (hyperscalers) is breaking through the boundaries of the technology sector and spreading to broader markets.

Goldman Sachs’ top derivatives trader Brian Garrett issued a warning in the latest weekend prep report: Pressure in the AI-related bond market—including widening credit spreads for hyperscalers, significant expansion in credit default swaps (CDS), and increased concessions for new bond issuance—is being transmitted throughout the entire market. Meanwhile, the S&P 500 Index is increasingly unable to accurately reflect the actual performance of individual stocks; the divergence between the index and single stocks is becoming apparent across asset classes.

Internal market pressure signals have clearly intensified. Goldman Sachs’ "panic index" soared by 5.5 points in a single week—a magnitude usually seen during periods of extreme market stress. Garrett bluntly stated that the actual experience on the trading desk last Friday was far worse than what the VIX reading of 18 would suggest and noted that there were signs of "partial capitulation selling" in the technology sector. However, he added that further observations are necessary.

AI capital expenditure concerns spread to credit markets

Garrett’s core assessment is that the wave of capital expenditure from AI hyperscale data center operators has become the main source of the global credit pulse. If this logic develops cracks, its impact will not be limited to the equity market.

In recent weeks, market doubts over the investment returns of hyperscalers have resurfaced and directly affected the credit market, causing related credit spreads to widen and putting pressure on semiconductor and memory chip stocks. Garrett pointed out that Google’s parent Alphabet and Tesla will report earnings this Wednesday, both viewed as "capital expenditure return" benchmarks, and the market will use these to once again test the validity of the AI investment thesis.

Garrett’s core judgment this week hits the mark: "If you performed well in the first half of 2026, your days in July could be extremely difficult."

The tech sector faces historic-level selloff

Goldman Sachs’ prime brokerage data shows the scale of tech sector selling has reached a historic record. Hedge funds have had net sales of US information technology stocks in 6 of the past 8 weeks, with cumulative selling being the largest seen in Goldman’s over ten-year record, comparable to the selloff wave of summer 2024.

Total and net exposure for the information technology sector as a proportion of US prime brokerage books rose to five-year highs of 23.4% and 26.3% respectively in early June, but just about six weeks later have fallen back to 19.4% and 14.7%—bringing their percentile rankings over the past year down to the 32nd and 2nd percentiles respectively.

Garrett specifically pointed out: "The continued and large-scale selloff since early June shows that tech investors are significantly reducing their long positions, with signs of partial capitulation starting to emerge." This week, the TMT sector became the worst-performing and most sold US sector.

Panic index surges, market structure becomes increasingly distorted

From derivatives market signals, internal market pressure has exceeded what surface data reveals. Goldman Sachs’ panic index jumped from low levels near 1 to highs above 6 this week, with a weekly increase of 5.5 points—a magnitude usually associated with extreme market stress events.

Meanwhile, stock volatility has increased markedly, but trading volumes have not expanded in tandem. Garrett believes this indicates investors still lack confidence to make significant adjustments to their positions. Last Friday, DRAM daily swing reached 13% but closed flat; the Philadelphia Semiconductor Index swung 7% but closed down 1%. Garrett reminded that the average market cap of SOX constituent stocks approaches $500 billion, and the occurrence of such dramatic volatility for companies of this size is already an abnormal signal.

Additionally, Garrett noticed a phenomenon that puzzled him: in the closing period, there was large-scale market-on-close (MOC) imbalance in the opposite direction of the S&P 500’s intraday movement. He attributed this to the prevalence of leveraged products, inverse ETFs, and zero-day options (0DTE), and judged that the current intraday technical structure of the S&P 500 is "completely different from the past."

Momentum strategies near their end, but risks remain unresolved

Goldman's high-beta momentum long-short portfolio has fallen 32% from its peak. Garrett’s team believes momentum unwinding is in its "late stage" for three reasons: first, the decline is similar to comparable historical events (46% in 2020, 45% in 2021); second, positions are much cleaner, with a sharp reduction in momentum long exposure; third, there is currently a lack of clear fundamental catalysts, and AI capital expenditure pressure is the main drag.

As for implied correlation, despite a slightly worsened macro backdrop, implied correlation remains at a 20-year low. Garrett believes this means the cost of hedging the real pain in individual stocks is much higher than hedging the S&P 500 index, which itself is increasingly unable to represent the performance of actual investment portfolios.

Risk warning and disclaimerThe market has risks; investment requires caution. This article does not constitute personal investment advice nor does it take into account any user’s specific investment targets, financial situation, or needs. Users should consider whether any opinions, views, or conclusions herein fit their particular circumstances. Investments based on this are at your own risk. ```