Key Notes
- A Steno Research reconstruction shows an approximate 7.8% cumulative position unwind through October 27, the deepest in its series beginning in 2016.
- Mid-October Goldman data showed AI net buying since 2025 being reversed, with existing long positions driving much of the selling.
- These figures measure trading activity across different periods, rather than investor redemptions or the elimination of every AI holding.
Hedge funds have sharply reduced stock-market risk as October’s artificial intelligence selloff forced a retreat from crowded trades. A chart shared by Steno Research founder Andreas Steno Larsen shows a position unwind equivalent to approximately 7.8% of gross exposure, the deepest in its reconstructed series beginning in 2016.
The broader retreat follows an earlier burst of AI selling. Goldman Sachs Group Inc. (NYSE: GS) figures reproduced in an October 16 analysis showed five consecutive sessions of net sales, from October 8 to 14, reversing the cumulative net buying of global AI stocks recorded since the start of 2025.
Those measures describe changes in trading positions, rather than investors withdrawing the same percentage of their money from hedge funds. They also cover different periods: the AI flow figures concern mid-October, while the Steno chart uses data through October 27.
The biggest deleveraging event in more than 10 years
— Andreas Steno Larsen (@AndreasSteno) October 30, 2026
A Large Reduction in Trading Positions
The Steno chart tracks cumulative gross trading flows from a running peak, expressed as a percentage of total global gross exposure. It is labeled as an approximate reconstruction of Morgan Stanley (NYSE: MS) prime-brokerage data, rather than a complete public census of the hedge-fund industry.
Its roughly 7.8% October drawdown exceeds the 2020 episode shown in that particular series. The comparison is specific to the position-flow measure; it does not establish that today’s financial conditions are worse than the pandemic crisis.
Gross exposure adds the size of long positions and short positions. Selling shares a fund already owns reduces the long side, while buying back borrowed shares to close a short reduces the short side. Both can shrink a portfolio’s overall trading footprint.
Net exposure measures something different: longs minus shorts. A manager can reduce gross positions substantially without making an equally large change in its net bet on the direction of the market.
AI Selling Reversed Net Purchases Since 2025
The Goldman figures, reproduced in Stephen Innes’s October 16 analysis, show cumulative net AI flows returning to their starting point after five selling sessions.
That does not mean funds sold every AI share they owned or eliminated positions built before 2025. The measure combines purchases, sales and changes in short positions over the period, rather than identifying an empty portfolio.
Innes described the latest session in that mid-October update as the second-largest AI net-selling day over the preceding year. Sales of existing long positions outpaced new short sales by roughly 3.5 to one in that session.
The distinction helps explain the pressure: much of the selling came from investors reducing an existing position, rather than exclusively from traders opening fresh bearish bets. The ratio should not be read as a final total for the entire month.
Later Reports Showed Continued Technology Selling
The unwind extended beyond the initial five-day window. Reuters reported on October 16 that Goldman had counted six consecutive sessions of AI selling. Global fundamental long-short funds lost 2.9% in the week through October 15, while Asia-focused stock pickers were down 6.8% in the first half of the month.
In a separate October 26 Reuters report, Goldman’s prime desk put technology-stock selling at about $20 billion over the preceding two weeks, approaching the $23 billion unwind recorded in July. American equities had been net sold for three consecutive weeks in the sample.
Prime brokers provide financing and trading services to hedge funds, making their books useful windows into institutional positioning. Their samples, classifications and reporting dates nevertheless differ, so the figures cannot simply be combined into one industry-wide cash-outflow total.
The reports show a sustained reduction in risk through several October episodes. They do not establish that every fund followed the same strategy or that managers collectively abandoned the long-term AI investment thesis.
Memory Stocks Bore a Heavy Cost
The market damage was particularly visible in memory suppliers. Yahoo Finance’s October 16 market analysis put the combined market-value loss at more than $700 billion since October 6 across four companies:
- Samsung Electronics (KRX: 005930), the South Korean memory and electronics group.
- SK hynix (KRX: 000660), a major supplier of memory used in AI systems.
- Micron Technology Inc. (NASDAQ: MU), the American memory-chip manufacturer.
- Sandisk Corp. (NASDAQ: SNDK), the flash-memory and storage company.
That was a mid-October estimate, not a newly measured seven-session loss ending November 1. A reduction in market capitalization also differs from the dollars hedge funds received by selling shares.
As we previously reported, South Korea’s chip-heavy market faced pressure from competition concerns and questions over AI investment returns. Those operating issues matter alongside the positioning changes amplifying price moves.
A Rotation With an Uncertain Endpoint
The retreat has also exposed differences across sectors. As we reported Friday, 25 stocks reached record intraday highs even as the broader market fell, with healthcare, energy and financial companies prominent on the list.
That divergence supports a picture of investors reassessing where to hold risk. It does not prove that all proceeds from AI sales went directly into those stocks, or that the selling has reached exhaustion.
Lower positions can reduce the amount left to liquidate, but weaker earnings expectations or tighter financing can still generate further pressure. The next positioning updates will show whether the retreat is slowing; company cash flow and funding disclosures will help determine whether the underlying investment case is changing.
