Running Away in Panic! Hedge Funds Offload Two Years' Net Purchases in One Week, Goldman Sachs Warns: Deleveraging May Have Only Just Begun

The AI positions that hedge funds had accumulated over nearly two years were completely liquidated within five trading days. The intensity and speed of this sell-off have raised market concerns over a larger-scale deleveraging.

According to the latest report from Vincent Lin, Head of Prime Brokerage at Goldman Sachs, hedge funds net sold global AI stocks for five consecutive trading days from October 8 to 14, fully reversing all cumulative net purchases since the start of 2025—in other words, every single net purchase made in nearly two years was wiped out within a week.

Meanwhile, the Goldman Sachs Prime book shows that global equities witnessed a "significant net sell-off," with the single-day net sell amount hitting -2.1 standard deviations on a one-year lookback basis. North America and emerging markets in Asia led the decline, with Korea and Taiwan being the most heavily net sold markets.

The direct impact of this sell-off has already been felt across multiple markets. Korea’s Composite Index entered a bear market, Seoul triggered a circuit breaker, foreign capital net outflow from Taiwan’s market hit a record high of about $9.3 billion in a single day, and several Asian central banks were forced to intervene in the foreign exchange market. TMT directional long-short funds lost 2.7% in a single day—an extreme move of -2.6 standard deviations—bringing cumulative October losses to 4.5%.

Two Years of Net Purchases Wiped Out in Five Days

The scale of this AI stock sell-off is second only to this July’s crash in Goldman Sachs’ historical data. According to Goldman Sachs Prime, the net sell amount for global AI stocks (GSXGGLAI constituents) on that day was the second largest in the past year, with a ratio of long liquidation to short selling at about 3.5 to 1—indicating the sell-off was mainly driven by long positions actively reducing exposure, rather than increased shorting.

From the flow chart, hedge funds’ net flows in AI names (black line) have returned to the starting point of early 2025, while long flows (blue line) have continuously fallen from the peak between February and March this year, and short flows (red line) have remained near a low point of around -13% throughout the period. Goldman Sachs’ chart clearly shows this round trip: The entire net purchase cycle from 2025 to 2026 was completely reversed within a week, while shorts continued to add during the decline.

Just last week, Goldman Sachs also pointed out that the net exposure of hedge funds to the Mag7 in its Prime book accounted for nearly 22% of total US exposure—a record high. Five days later, the same group of funds collectively rushed to exit.

Positions Remain Heavy, Deleveraging Far from Over

Although the sell-off this week was intense, from a longer-term perspective, AI positions remain elevated, and the room for further deleveraging cannot be ignored.

Goldman Sachs data shows that the proportion of global AI stock net allocation to global single-stock net market value is currently at 26.1%, down from the early June peak of 31.7%, ranking at the 49th percentile in the past year—but still at the 83rd percentile over the past three years. The AI portfolio’s long-short ratio has fallen from a June peak of 4.34 to 3.27, ranking at the 7th percentile in the past year, but still at the 64th percentile over three years.

In short, one-year data shows positions have been relatively cleared, but three-year data indicates hedge funds remain significantly overweight overall. Rachel Hu from Goldman Sachs’ Asia team points out that trading feedback from the Taiwan market shows the level of "capitulation" in this sell-off is less severe than in July, suggesting more selling pressure may follow.

J.P. Morgan’s TPM positioning indicator currently sits around the 46th percentile—a "neutral" zone—well above the 19th percentile at the late July low, indicating there is still room for "further deleveraging" before key tech catalysts emerge.

Defensive Shift: Selling Tech, Buying Toothpaste

This sell-off is not limited to the AI sector; instead, it presents a broad defensive shift. Goldman Sachs Prime book shows that eight out of 11 US sectors saw net selling, with information technology and industrials leading the declines. All sub-sectors in information technology—software, semiconductors and semiconductor equipment, IT services—suffered net selling.

Meanwhile, hedge funds’ buying activity focused on healthcare and consumer staples, which Goldman Sachs characterizes as a "marginal increase in defensive posture." In early trading on Thursday, healthcare and consumer staples both rose over 2%, supporting the Dow, while the Nasdaq slid to its lowest in nearly a month. CoreWeave dropped another 8.5%, the Philadelphia Semiconductor Index fell for a seventh straight day, and about $2 trillion in related market cap was wiped out.

In this collective bloodletting among human fund managers, quant funds were the only winners. Global systematic long-short funds rose 0.5% on the day, with a cumulative gain of 1.9% in October; year-to-date returns lead all strategies at 21.7%.

Fragile Leverage Structure, Violent Deleveraging Risk Remains

Leverage data helps explain the intensity of this sell-off. Goldman Sachs Prime book’s total leverage ratio only dropped slightly by 0.4 percentage points to 310.3%, ranking at the 72nd, 91st, and 94th percentiles over one, three, and five-year periods, respectively—near the five-year high. Net leverage fell by 1.5 percentage points to 73.3%, down 4.3 percentage points cumulatively in October, ranking at the 6th percentile over the past year.

Goldman Sachs’ logic: a fund with total leverage of 310% and net leverage of 73% has an actual sellable scale far exceeding what net exposure shows. When crowded longs (AI) and crowded shorts move in opposite directions, deleveraging becomes the only way out, but the exit is extremely narrow and everyone rushes to the same door at once—this is a repeat of the late July crash.

Rich Privorotsky from Goldman Sachs Delta-One business summarized the current predicament in an overnight report: "Looking back, the painful trade was long AI, short bonds. Now we’re seeing both sides being covered at once... The market is discovering that the demand for computing power can grow exponentially, but the capital available to finance it cannot."

Goldman Sachs: This Looks More Like the Beginning Than the End

The market’s main debate now is: is this a buying opportunity, or the beginning of a larger-scale deleveraging?

J.P. Morgan leans toward the former, seeing dips as buying opportunities. Goldman Sachs’ flow desk was still watching positive technical signals for year-end the previous two days. However, Goldman Sachs’ overall judgment is more cautious: AI net exposure remains at the 83rd percentile over three years, total leverage is at the 94th percentile over five years, and Asian foreign capital is pulling out at record scale—all these signals point to this round of deleveraging being closer to the beginning than the end.

Hedge funds have already sold all AI positions bought since 2025. But those accumulated in 2023 and 2024 still remain fully on the books.