Hedge funds erased nearly all of their cumulative net buying in global artificial-intelligence-linked stocks since the start of 2025 during five trading days of selling from Oct. 8 to Oct. 14, according to Goldman Sachs’ prime brokerage data. The retreat ranked as the second-largest one-day net selloff in Goldman’s global AI basket over the past year and was driven chiefly by funds cutting long positions rather than adding bearish bets.
The speed of the unwind places renewed pressure on an equity trade that had concentrated heavily in software, chipmakers and AI infrastructure companies. Goldman’s data showed that funds did not fully abandon the theme built during the earlier 2023-24 rally, but they removed the net exposure added this year as valuations and market volatility came under strain.
Vincent Lin, who leads Goldman Sachs’ prime brokerage reporting cited in the material, said the selling extended beyond AI stocks. The bank’s prime book recorded broad net selling across global equities, including a one-day move of negative 2.1 standard deviations against its trailing one-year range. North American equities and emerging-market Asia saw the largest withdrawals, with South Korea and Taiwan the most concentrated regional areas of net selling.
Long-position cuts drove the AI selloff
The flow pattern points to funds reducing existing bullish positions rather than constructing a broad new short trade. Goldman calculated that long selling outpaced short selling by roughly 3.5-to-1 in its AI stock list, known as the GSXGGLAI basket.
That distinction matters for how the market may trade next. Heavy short selling can create future buying pressure when bearish positions are closed. A long liquidation, by contrast, reflects managers reducing exposure they previously held for upside participation, leaving less immediate forced demand once the initial selling ends.
Goldman’s figures showed net flows in the AI basket returning to levels seen at the beginning of 2025, while short flows remained about 13% below their earlier level. The bank said hedge funds had therefore wiped out the year’s accumulated net buying without fully reversing the longer-term AI positioning built during the previous two years.
Net allocation to global AI stocks stood at 26.1% of global single-stock net market value after the exit, down from 31.7% at an early-June peak, Goldman found. The allocation was near the 49th percentile against the past year, suggesting it had moved close to a normal recent range. Against a three-year comparison, it remained at the 83rd percentile, evidence that AI still occupies a large place in hedge-fund portfolios relative to the period before the sector’s surge.
The gross long-short ratio for AI holdings fell to 3.27 from 4.34 in June. That measure sat in the seventh percentile of its one-year range but the 64th percentile over three years, another sign that recent enthusiasm has cooled sharply without disappearing entirely.
Asian markets absorbed the sharpest pressure
The wider equity retreat hit Asia particularly hard. South Korea’s Kospi entered bear-market territory and triggered a circuit breaker in Seoul, according to the supplied material. Taiwan recorded a one-day foreign outflow of roughly $9.3 billion, while several Asian central banks intervened in currency markets to support their domestic currencies.
South Korea and Taiwan are central nodes in the global semiconductor supply chain, making them especially exposed when fund managers trim chip and AI hardware positions. Semiconductor companies have been among the major public-market beneficiaries of demand for data centers and advanced computing, but that connection also makes them vulnerable when positioning shifts quickly.
In the United States, Goldman recorded net selling in eight of 11 sectors, led by information technology and industrials. Software, semiconductors and semiconductor equipment, and IT services all registered net outflows. Health care and consumer staples attracted buying, reflecting a move toward sectors generally viewed as less sensitive to swings in economic expectations and technology valuations.
During Thursday morning trading cited in the material, health care and consumer staples each rose more than 2%, while the Nasdaq approached a near one-month low. CoreWeave declined another 8.5%, and the Philadelphia Semiconductor Index fell for a seventh consecutive session. The material estimated that related semiconductor market capitalization had fallen by about $2 trillion over that stretch.
High leverage remains a market constraint
The selling occurred while overall hedge-fund gross leverage remained elevated. Goldman reported total gross leverage of 310.3%, down 0.4 percentage points on the day. That figure ranked in the 72nd percentile over one year, the 91st percentile over three years and the 94th percentile over five years.
High gross leverage does not automatically signal imminent selling, since it measures both long and short exposures. It can, though, amplify portfolio adjustments when volatility rises or when crowded positions reverse. Net leverage fell 1.5 percentage points to 73.3%, including a 4.3-point decline during October, placing it in the sixth percentile of its one-year range.
Goldman had previously estimated that hedge-fund net exposure to the “Magnificent Seven” technology companies had reached nearly 22% of total U.S. exposure, a record. Such concentration means a retreat in the largest technology names can quickly spread through portfolios that use them as core long holdings.
Rich Privorotsky of Goldman Sachs’ Delta-One business described simultaneous reductions on both sides of a trade framed as long AI and short bonds. That combination suggests funds were reassessing a macroeconomic positioning structure, not merely reacting to weakness in a few technology stocks.
Crypto markets face a separate leverage test
The supplied material linked the equity move to $932 million in overnight forced liquidations across digital-asset markets and cited Bitcoin near $80,850 after a rise from a summer low of $58,000. It also said spot Bitcoin funds had recorded $760 million in outflows since early October.
Risk appetite can connect technology equities and crypto during broad deleveraging episodes, particularly where traders use borrowed capital. Yet the Goldman data tracks hedge-fund equity positioning, not direct cryptocurrency exposures, and does not establish that AI-stock selling caused the reported token-market liquidations.
The immediate lesson from the prime-brokerage flows is narrower: hedge funds have rapidly reduced the 2025 portion of a once-crowded AI trade while retaining meaningful longer-term exposure. Whether that becomes a temporary reset or develops into a deeper reassessment will depend on whether selling remains concentrated in expensive technology shares or expands further across global equity markets.
Worried about AI-driven volatility? Use AI copy trading to mirror expert strategies and navigate rapid hedge-fund rotations.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
