Panic Signals Emerge? Goldman Warns: AI Credit Risk Begins Spreading to Broader Markets

Brian Garrett, a senior derivatives trader at Goldman Sachs, has warned that the market is beginning to question whether the massive artificial intelligence (AI) investments by large technology companies will generate sufficient returns. The credit risk associated with capital expenditures on hyperscale data centers is no longer confined to the technology sector but is gradually spreading to broader financial markets.

In his latest weekly market report, Garrett noted that credit spreads for hyperscale data center operators have continued to widen in recent weeks, while credit default swap (CDS) costs have risen. New bond issuances now require higher spreads to attract buyers, indicating that pressure in the AI-related bond market is being transmitted across the broader market.

He believes that AI capital expenditures have become a key driver of the global credit cycle, and if investors begin to doubt this investment logic, the impact will extend beyond tech stocks and could affect the overall market.

Market attention will now focus on the upcoming earnings reports from Alphabet (GOOGL-US), Google's parent company, and Tesla (TSLA-US) this week.

Garrett emphasized that both companies are critical benchmarks for testing whether AI capital spending can translate into actual profits. Their earnings results will once again test market confidence in the AI investment narrative.

He bluntly stated: "If you perform exceptionally well in the first half of 2026, July might be quite painful."

Hedge Funds Withdraw Heavily from Tech Stocks

Data from Goldman Sachs' prime brokerage shows that the scale of hedge funds reducing their positions in U.S. tech stocks has reached the largest level in over a decade. Over the past eight weeks, six weeks have seen net selling, with selling pressure comparable to the tech stock correction in summer 2024.

Total and net exposure to the information technology sector in U.S. prime brokerage accounts peaked at 23.4% and 26.3% respectively in early June, but dropped to 19.4% and 14.7% within just about six weeks.

Garrett said the sustained and large-scale capital outflows since early June indicate signs of "capitulation selling," and the technology, media, and telecommunications (TMT) sector has become the weakest-performing and most capital-outflowing group in the U.S. stock market this week.

Panic Index Rises, Market Structure Shows Anomalies

In addition to capital flows, derivatives markets are also signaling greater stress.

Goldman's proprietary "panic index" surged from near 1 to over 6 this week, rising 5.5 points in a single week—a movement typically seen only during periods of extreme market stress.

Garrett said the level of tension on trading desks last Friday (the 17th) was far more severe than what the VIX index's level of around 18 would suggest.

Notably, individual stock volatility has clearly increased recently, but trading volume has not risen correspondingly, suggesting investors are under pressure but have not yet made broad portfolio adjustments.

For example, the Roundhill Memory ETF (DRAM-US) saw an intraday swing of 13% last Friday but closed nearly flat. The Philadelphia Semiconductor Index (SOX) swung 7% that day but closed down only about 1%.

Garrett pointed out that the average market cap of SOX components is close to $500 billion, and such large-cap stocks moving in sync with extreme volatility is itself a warning signal.

Additionally, he observed a significant imbalance in "market-on-close" (MOC) orders at the end of trading sessions that moved in the opposite direction of the S&P 500's daily movement. He speculated this is likely due to the popularity of leveraged ETFs, inverse ETFs, and zero-day-to-expiration (0DTE) options, indicating that the current market's closing auction structure has fundamentally changed from the past.

Momentum Trading Nears End, But Market Risks Remain

Goldman's tracked high-beta momentum long-short strategy has declined 32% cumulatively from its peak.

Garrett's team believes the current deleveraging of momentum trades may be in its late stages, citing reasons such as the current drawdown approaching historical correction levels seen in 2020 and 2021, investors having clearly reduced holdings, and the lack of new fundamental negative catalysts—AI capital spending concerns remain the primary drag.

However, he cautioned that implied correlation among individual stocks remains at a near 20-year low, meaning stock price movements are increasingly diverging. Investors seeking to hedge individual stock risk will face much higher costs than shorting the S&P 500 index, making the S&P 500 increasingly inadequate at reflecting the true performance of a typical investment portfolio.