Hedge Funds Liquidate AI Positions in Five Days, Second-Largest Net Selling in Nearly a Year
nashnova research
Hedge funds erased nearly two years of accumulated net buying in global AI stocks within five trading days, the second-largest single-day net selling in a year; yet positions remain elevated on a three-year basis, suggesting the unwind may not be over.
Two years of buying, gone in five days — what happened?
According to Goldman Sachs Prime data, hedge funds net-sold global AI stocks for five consecutive trading days from October 8 to 14. Cumulative net buying since early 2025 has been fully reversed to zero.
Single-day net selling ranked as the second-largest in the past year, behind only the July selloff.
This means → positions built steadily over nearly two years were liquidated in a single week — far faster than they were assembled.
Who was selling — longs exiting or shorts piling in?
The selloff was driven overwhelmingly by long liquidation. The ratio of long selling to short selling was roughly 3.5 to 1.
In plain terms = this was not bears attacking AI — it was funds that *believed* in AI walking away. "Faith shaken" explains this round better than "short squeeze."
Short additions played a supporting role, not the lead.
Where do positions stand now?
Hedge-fund net allocation to global AI stocks fell from a June peak of 31.7% to 26.1% of global single-stock net market value, landing at the 49th percentile over one year.
In plain terms = on a one-year view, positioning is back to the middle — close to a "washout."
But on a three-year window the same metric sits at the 83rd percentile. This means → relative to longer history, AI allocation is still elevated and room to de-risk may remain.
What do other firms say — was this a full "capitulation"?
JPMorgan's TPM positioning indicator (a gauge of overall market exposure) sits near the 46th percentile, above the 19th-percentile low hit in late July. This means → the market still has room to deleverage further.
Goldman's Asia team noted that Taiwan market signals show this selloff's "capitulation" intensity fell short of July, hinting at potential further pressure ahead.
The long-short ratio dropped from a June peak of 4.34 to 3.27 — 7th percentile on one year, but still 64th percentile on three years.
Is the AI selloff an isolated event?
No. On October 14, global equities saw significant net selling overall — a single-day move of roughly negative 2.1 standard deviations versus the past year, again led by long sales.
Nearly every region saw net selling. North America and emerging-market Asia led the decline, with South Korea and Taiwan the most heavily sold markets.
This reflects a broader risk-appetite contraction, not a verdict on AI alone.
Is the unwind over? Where is the key tension?
The core contradiction: one-year positioning signals "washout," while three-year positioning still signals "overweight" — the two readings conflict.
This means → which signal matters more depends on whether the market believes the 2026 AI boom can continue.
Put simply = if you think AI momentum holds, the one-year "washout" is a buy signal; if you think the cycle has peaked, the three-year "overweight" is the true reading — and the unwind is far from done.
市场有风险，内容仅供研究参考，不构成投资建议。