Hedge Funds’ October 2026 AI Rout: How Two Years of Buying Unwound in a Week

Key Takeaways

• In the week to 15 October 2026, global fundamental long-short funds — the stock pickers — lost money on every trading day, ending the week down 2.9 percent, though still up 14.3 percent for the year at that point, according to Goldman Sachs prime brokerage data reported by Reuters.
• Between 8 and 14 October, hedge funds net-sold global AI equities for five consecutive sessions — six straight including 15 October — erasing cumulative net purchases of AI stocks built up since the start of 2025. About two years of buying was unwound in about a week.
• On Thursday 22 October, all three major strategies — fundamental equity long-short, systematic, and multi-strategy — each lost more than 1 percent on the same day, the first simultaneous drawdown of that size since the COVID selloff of 2020.
• For October as a whole, JPMorgan estimates global hedge funds lost 3 percent; the Nasdaq 100 plunged 6.6 percent in the month, its biggest monthly fall in 18 months; Korea’s KOSPI fell 16 percent and China’s Star 50 fell 24 percent.
• Aspex Management, a Hong Kong pan-Asian fund managing about $14 billion, lost about 16 percent in October, one of its biggest monthly drawdowns, though it remained up 9 percent for the year through October.
• Walleye Capital, a New York multi-strategy fund managing about $17 billion, had its worst month since March 2020 and said in an investor letter on 5 November that it would trim its equity long-short business, cut a substantial portion of its equity investment team, and return a portion of capital.
• Search interest in “hedge funds october artificial intelligence rout” was a breakout topic, and searches for “hedge fund” and “hedge” rose sharply over seven days, according to the trend data behind this story — a rise in search interest, not a measure of trading volume.

A bad week that kept getting worse

October 2026 was the month hedge funds discovered, all at once, that the AI trade could go down as fast as it had gone up. The numbers, drawn from Goldman Sachs prime brokerage data and reported by Reuters, describe a selloff that was broad, fast, and concentrated in exactly the positions funds had crowded into over the previous two years.

Start with the week to 15 October. Global fundamental long-short funds — the stock pickers, the strategy most readers picture when they think of a hedge fund — lost money on every trading day that week and ended it down 2.9 percent. That was painful, but it was not yet catastrophic: the same funds were still up 14.3 percent for the year at that point. The damage was concentrated in the middle of the week. On Wednesday 15 October alone, global stock pickers fell 0.9 percent, Asia-focused managers fell 2.6 percent, tech-focused funds fell 2.0 percent, and systematic long-short and multi-strategy funds each fell about 1 percent.

The breadth of that single day is what made Goldman pay attention. 15 October was the third day in 2026 — after 4 February and 28 July — on which all three broad strategies fell more than 80 basis points in one session. Goldman noted that this had never happened from COVID through the end of 2025. In other words, a pattern that had not occurred once in roughly five years had now occurred three times in less than nine months, and the third occurrence was the worst of the three for Asia-focused managers.

The first half of October was worse still for Asia. Asia-focused stock pickers fell 6.8 percent in the first half of the month, erasing more than half of their gains from a late-June peak, when they had been up more than 40 percent for the year. A strategy that had been the year’s standout performer in June was, by mid-October, giving back most of what it had made. That is the shape of a crowded trade unwinding: the funds that had gained the most from a theme were the ones that lost the most when it turned.

Two years of AI buying, unwound in about a week

The most striking single fact in the Goldman data is about flows, not returns. Between 8 and 14 October, hedge funds net-sold global AI equities for five consecutive sessions — six straight sessions including 15 October, according to a later Goldman note. Those sales erased the cumulative net purchases of AI stocks that funds had built up since the start of 2025. About two years of buying was unwound in about a week.

The proceeds did not leave equities altogether. Funds rotated into healthcare and consumer staples, the defensive corners of the market where earnings are steadier and valuations less dependent on distant growth. That rotation is a classic risk-off move: when a growth theme turns, money does not necessarily go to cash; it goes to the parts of the market that are least correlated with the theme that just broke.

The scale of the selling was extreme by Goldman’s own measures. Net selling of global equities measured minus 2.1 standard deviations against the one-year average — a statistical way of saying that the pace of selling was far outside the normal range. Net leverage, the amount of borrowed exposure funds were running, fell to historically low levels. Funds were not just selling; they were cutting the leverage that had amplified their gains on the way up, and that was now amplifying their losses on the way down.

That deleveraging dynamic is central to understanding October. In a crowded trade, losses force selling, and selling causes more losses, which force more selling. A fund that loses money in a crowded position may have to cut exposure to meet risk limits, even if it still believes in the position. When many funds hold the same positions — and PivotalPath warned of high and increasing overlap among the ten largest multi-strategy funds, meaning an unwind could hit many of the same securities — those forced sales land in the same stocks at the same time. The result is a feedback loop, and October was a month of feedback loops.

The later data confirmed the scale. In the week to about 26 October, trend-following systematic funds had their worst week since 2024, down 2 percent for the week, while the Nasdaq 100 fell 5.5 percent that week, its biggest weekly fall since April 2025, and Meta fell 10 percent in the week after flagging more capital-raising for AI. Goldman reported $20 billion of tech-stock selling over two weeks, against $23 billion in July’s unwind. July had been the warning; October was the larger event in most of the measures that mattered, even if the two-week dollar total was slightly smaller.

Thursday 22 October: the day all three strategies fell together

If 15 October was the warning, Thursday 22 October was the confirmation. On that day, all three major strategies — fundamental equity long-short, systematic, and multi-strategy — each lost more than 1 percent. It was the first simultaneous drawdown of that size since the COVID selloff of 2020. A Goldman prime brokerage note put it bluntly: “If last Thursday felt bad, today has been worse.”

The catalysts that day were specific and visible. Meta fell 7 percent, its largest drop since July, after announcing a new share class to raise funds for AI. Nvidia fell 4 percent. The Nasdaq Composite fell 2.6 percent, its largest decline since 5 June. Meta’s announcement mattered because it connected the two worries driving the selloff: that AI spending is enormous, and that it is being funded with new capital — in this case, new shares — that dilutes existing holders. Investors were not just questioning whether AI spending would pay off; they were watching the bill for it arrive in real time.

By 22 October, the monthly tallies told the story. Global fundamental long-short funds were down 3.8 percent for October, though still up 8.8 percent for the year. Tech-focused funds were down about 10 percent for October. Asia-focused funds were down about 9 percent. The losses were driven by China and technology holdings, plus a squeeze in crowded short positions — funds that had bet against certain stocks found those bets moving against them at the same time their long positions were falling, a combination that hurts both sides of a long-short book at once.

The full-month figures, reported on 6 November, completed the picture. JPMorgan estimates global hedge funds lost 3 percent in October. The Nasdaq 100 plunged 6.6 percent in the month, its biggest monthly fall in 18 months. The Nasdaq Composite fell 5 percent in October. The damage was worst in Asia: Korea’s KOSPI fell 16 percent, Taiwan fell 6 percent, and China’s Star 50 fell 24 percent. The October rout was a global event, but its centre of gravity was Asian technology, where the year’s gains had been largest and the crowding heaviest.

The funds that were hit hardest

Behind the strategy-level numbers are individual funds, and two of them illustrate the month in concrete terms.

Aspex Management, a Hong Kong-based pan-Asian equities fund managing about $14 billion, founded by Hermes Li, the former head of Asia equities at Och-Ziff, lost about 16 percent in October, according to a source cited by Reuters — one of its biggest monthly drawdowns. It remained up 9 percent for the year through October, which shows how strong its earlier gains had been, and how much of them a single month could erase. Aspex had backed AI startup MiniMax and optical-transceiver firm Zhongji Innolight, two names squarely in the AI supply chain that the selloff hit hardest. The fund’s experience is a case study in concentration: backing the right theme, in the right region, at the right time, and then watching the same theme reverse faster than the position could be reduced.

Walleye Capital, a New York multi-strategy fund managing about $17 billion, had its worst month since March 2020 — the COVID crash. In an investor letter on 5 November, it said it would trim its internal equity long-short business, cut “a substantial portion” of its equity investment team, and return a portion of capital to investors. That is not a routine monthly update. A multi-strategy fund shrinking its equity business, cutting staff, and returning capital is a fund deciding that the strategy that lost the money should be smaller, permanently. Several large multi-manager platforms cut leverage and laid off loss-making teams in recent weeks, according to Reuters, with more layoffs expected if declines continue. October’s losses are now reshaping the industry’s structure, not just its monthly returns.

A third voice adds a different kind of warning. Masaki Gotoh, a partner at Tokyo-based TriVista Capital, said his firm had trimmed hardware investments since late June, and added: “the concentration level is staggering… the circular financing is also a source of some concern”. That is an analyst’s opinion, not a reported loss figure, and it should be read as one. But it captures the two worries that ran through the month: that too much money was concentrated in too few AI names, and that some of that money was circulating between companies that were simultaneously each other’s customers, suppliers, and investors. Circular financing does not prove a bubble, but it does mean that a problem at one company can travel quickly to others.

The 5 November rout deserves a brief mention here as a coda, not as part of the October story. On that Thursday, after the US midterm elections, tech-focused equity long-short funds fell more than 4 percent in a day, leaving their year-to-date returns “roughly flat” against a peak of plus 24 percent in June. Reuters links the November momentum to Democrat midterm wins fuelling fears of regulatory scrutiny — a November development, distinct from the October rout. It is included here because it shows the October damage was not repaired in the first days of November, but the causes were different, and this article keeps its focus on October.

Why it happened: rates, debt, and the returns question

The reported causes of the October rout fall into four linked groups, and they reinforce each other.

First, inflation. Hotter-than-expected US September CPI data repriced the rate path higher. When expected rates rise, the present value of distant future earnings falls, and AI companies are valued precisely on distant future earnings. Higher rates also raise the cost of the debt that funds the AI buildout. That connection was made explicit by the Federal Reserve Chair himself in the same month: Kevin Warsh said at the 28 October press conference that “interest rates get in the cracks, interest rates affect directly or indirectly, almost everything”, and expressed scepticism that AI investment is insensitive to borrowing costs. Our report on Kevin Warsh’s October remarks covers his comments in full, including what he did and did not say about an AI bubble.

Second, debt. The AI infrastructure buildout is heavily debt-financed, so higher borrowing costs bite directly. The stress showed up in credit markets: CoreWeave bonds sold off, with yields above 15 percent. A yield above 15 percent on a company’s bonds is a market verdict that its debt is risky, and it is also a practical problem, because it raises the cost of the next borrowing. Companies that had raised large sums to fund AI infrastructure — as our report on Oracle and Alphabet Borrowing Billions for AI on the Same Day documents — were doing so in a market that was becoming more selective about who it would lend to and at what price.

Third, the returns question. Investors questioned whether massive AI spending will generate returns soon enough to justify its scale. This is not a claim that AI has no value, or that the spending will never pay off. It is a claim about timing and magnitude: the spending is certain and immediate, the returns are uncertain and future, and in October the market decided to charge more for that uncertainty. OpenAI’s experience in the same period is a useful parallel — it cut its funding target from about $30 billion to about $20 billion as AI stocks slid. Our report on OpenAI Cuts Its Funding Target by $10 Billion as AI Stocks Slide covers that episode, and it shows the same selectivity operating in private markets that Goldman measured in public ones.

Fourth, crowding and deleveraging. The AI trade was crowded — too many funds holding the same names, in the same direction, with leverage. When the trade turned, losses in crowded positions forced funds to cut exposure, and the forced selling pushed prices down further. That is a mechanical process, not a verdict on AI as a technology, but its effects on returns were real. It also explains why the damage was worst in Asia and in technology: those were the most crowded corners of the market, and the most leveraged to the AI theme.

What the rout does and does not prove

October’s numbers are dramatic, but they need the same discipline as any market story: separate what happened from what it means.

What happened is documented: funds lost money, leverage fell, two years of AI buying was unwound in a week, named funds suffered historic monthly losses, and the industry began cutting teams and returning capital. Those are facts from Goldman Sachs prime brokerage data and JPMorgan estimates, reported by Reuters, and they are not in serious dispute.

What it means is less settled. It does not prove that AI is a bubble, or that AI spending will collapse, or that the technology has failed. Funds that lost money in October were, in many cases, still up for the year — global stock pickers were up 8.8 percent year-to-date at 22 October, Aspex was up 9 percent through October. A drawdown that leaves a fund up for the year is a correction in a profitable strategy, not a wipeout. Nor does it prove that the AI buildout will stop: the spending plans that worried investors in October were still, in most cases, spending plans, not cancellations.

What it does prove is narrower and more useful. It proves that the AI trade had become crowded enough that an unwind could be fast, broad, and self-reinforcing. It proves that leverage amplifies losses as efficiently as it amplifies gains, and that risk limits force selling at exactly the moment when selling hurts most. It proves that higher rates and debt-funded spending are a dangerous combination when sentiment turns. And it proves that the funds most exposed to a single theme, in a single region, are the ones that suffer most when that theme reverses — a lesson Aspex’s 16 percent month and the Asia-focused funds’ 9 percent month both illustrate.

For a broader view of how investors were repricing the AI boom in the same weeks, see our coverage of AI Bubble Fears Surge as Markets Reprice the AI Boom. The two stories are companions: one is about what public-market investors feared, and this one is about what professional funds actually lost.

What to watch next

If you want to follow this story honestly, watch a short list of verifiable milestones rather than daily commentary. First, November prime brokerage data: whether the net selling of AI equities continued after October, or whether funds began rebuilding positions at lower prices. Second, leverage: whether net leverage stays at historically low levels, which would suggest funds remain cautious, or rebuilds, which would suggest confidence is returning. Third, the funds themselves: whether more multi-manager platforms follow Walleye in trimming equity teams and returning capital, and whether layoffs continue. Fourth, credit markets: whether CoreWeave’s bond yields come back down, and whether AI-related borrowers can still raise debt at workable prices. Fifth, earnings and spending plans: whether the companies at the centre of the AI buildout maintain, cut, or increase their capital expenditure, and whether they can show revenue growth that justifies it.

October 2026 will be remembered as the month the AI trade stopped being a one-way bet. The funds that made the most from it in the first half of the year gave back a large part of those gains in a few weeks, and some of them are now smaller, less leveraged, and less exposed to the theme than they were. That is not the end of AI investing. It is the end of the easy phase of it — the phase when buying the theme was enough, and when crowding felt like consensus. The next phase will reward discrimination: between companies that can fund their spending and those that cannot, between spending that generates returns and spending that does not, and between funds that manage their crowding and those that do not. October was the month the market started charging for those distinctions.

Sources: Reuters, Hedge Fund Returns Worsen This Week as Tech Selloff Deepens; Reuters, Trend Hedge Funds Saw Worst Week Since 2024 as AI Stocks Fell, Goldman Says; Reuters, Tech Selloff Wipes Out Hedge Fund Gains, Prompting Risk Cuts, Layoffs; Reuters, Asian Hedge Fund Aspex Suffers 16% Loss in October Tech Rout, Source Says; DZRH, AI Selloff Hammers Hedge Funds This Week as Asia Managers Lead Losses.