Leveraged ETFs Face $47B Single-Day Forced Selling as AI Bonds Hit by Massive Hedge Fund Dumping
nashnova research
Nomura estimates leveraged ETFs had to force-sell roughly $47 billion at Thursday's close — the second-largest single-day forced sale in their data history — while hedge funds simultaneously capitulated out of AI-related bonds, creating a three-way deleveraging shock.
How extreme is a $47 billion forced sell?
Nomura quant strategist Charlie McElligott estimates leveraged ETFs had to force-sell approximately $47 billion at Thursday's close — second only to June 5's $51.3 billion and sitting at the 0th percentile of history.
Morgan Stanley's derivatives desk, using 1 p.m. price levels, pegged the figure higher at roughly $52 billion.
This means → this is not a discretionary sell-off. Leveraged ETFs must rebalance daily to match their target multiple; the steeper the drop, the larger the forced sale.
In plain terms = the funds are not choosing to sell — the rules force them to, and bigger losses trigger bigger forced sales, creating a self-reinforcing downward spiral.
Why did the selling land squarely on AI and chips?
By sector: semiconductors $20.5 billion, tech $17.3 billion, the "Magnificent Seven" $3.9 billion, S&P 500 $4.2 billion — every one at the 0th percentile of history.
McElligott notes that 86% of leveraged ETF AUM is concentrated in "tech leadership" / AI themes.
This means → leveraged ETFs act as an amplifier for AI exposure — they double-buy on the way up and double-sell on the way down, and an 86% concentration funnels nearly all the selling pressure into AI-linked names.
How did the "trade of the year" collapse in 48 hours?
McElligott called the "semis-plus-energy barbell" the "trade of the year" — at its peak weeks ago, it had returned 66% with a Sharpe ratio of 3.3.
In the past 48 hours: energy fell 10.2% on the week, semis fell 13.1%, and the barbell strategy dropped roughly 13% overall.
The "AI enablers long vs. hyperscaler short" pair trade has retraced 15% from its June 22 peak of +80.4%, with weekly Sharpe ratios of -17.2 and -10.1.
In plain terms = the strategy bet that "both sides of AI win" — buy the chips and buy the power. Instead both sides collapsed at once, turning outsized gains into outsized losses overnight.
How did "front-running the front-run" amplify the drop?
McElligott titled his report "the front-running of the front-running of mechanical flows."
This means → the market knew a $47 billion forced sell was coming at the close, so traders sold ahead of it to front-run the impact — and that front-running itself pushed prices lower, triggering even more forced selling. The decline was artificially magnified.
Nomura's crowding scorecard: U.S. equity momentum factor down 11.9% over five days and 23.1% quarter-to-date; energy ETF (XLE) down 10.4% month-to-date.
Multiple "capitulation catalysts" stacked at once: hedge fund Walleye returned capital and fired its equity head after its worst month since Covid; Goldman's John Flood said his desk was seeing "phones blowing up"; UBS trader Torsten Sippel warned hedge funds were capitulating, with a buy/sell/short ratio of 23/75/2.
What happened simultaneously in AI bond markets?
Andrzej Skiba, head of BlueBay U.S. fixed income at RBC Global Asset Management, said on Bloomberg TV: "We are seeing massive capitulation from hedge funds in AI-related bonds."
A roughly $500 million data-center loan portfolio has been put up for sale — a rare auction occurring days after the midterm elections. The loans include a $25 million tranche backing the Stargate project's large data center in New Mexico, a joint venture of OpenAI, SoftBank, and Oracle.
The credit spread for investment-grade tech borrowers over the broader U.S. market has widened to its highest since 2003.
This means → deleveraging is not confined to equities — AI-linked bonds are being unwound by hedge funds too, and credit markets are repricing AI assets in real time.
After this three-way shock, is deleveraging nearly done?
Three forces are acting simultaneously: leveraged-ETF mechanical selling, CTA systematic deleveraging, and AI bond-market liquidation.
CTA trend-following strategies — quantitative funds that trade automatically based on price trends — still hold legacy bond shorts. McElligott estimates a 2% rise in U.S. Treasury prices would trigger roughly $225 billion in global bond buying, including $69.5 billion in 2-year euro, $48.5 billion in 2-year dollar, and $23.9 billion in 10-year JGBs.
Skiba had previously called AI "a losing trade" for fixed-income investors, but said he is now "more comfortable" holding AI-related debt as spreads have widened sharply.
This reflects whether deleveraging is near its end depends on whether the CTA bond short-squeeze has run its course — if Treasuries keep rising, the scale of forced buying could unleash another wave of dislocation.
市场有风险，内容仅供研究参考，不构成投资建议。