Hedge funds claiming diversification suffered collective losses in October from concentrated AI bets
Hedge funds across nearly all mainstream strategies lost money simultaneously in October, showing that most gains over the past year were actually from implicit long AI exposure, not genuine diversified strategy returns.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Flagship funds claiming stability lost money collectively in October
Millennium, Point72, and Exodus Point—the funds with the most concentrated talent and assets, viewed by investors as most steadfast during crises—all lost money in October. Millennium, with ~14% annualized returns, fell 2.3% in a single month. By hedge fund logic, this is dereliction: investors buy hedge funds not to beat bull markets but to expect protection during downturns. In October, that protection failed entirely. The real shock is not the loss itself but the failure to hedge.
— Nishant KumarFunds claiming diversification were concentrated in the same few AI stocks
Vara, a fund with positions highly overlapping with certain AI narratives, crashed 33% in October yet remains positive year-to-date—showing how much that prior AI exposure returned. Whale Rock similarly fell 12%. These single-strategy funds' exposure is explicit: nearly all positions bet on the same small handful of AI stocks. When momentum reverses, no hedge can cushion the blow.
— Nishant KumarData shows multi-strategy funds were herding into the same stocks
Pivotal Path's research shows the top 10 multi-strategy funds control ~$2.2 trillion in positions including leverage, yet the actual capital supporting those positions is only $300-350 billion—indicating extreme system leverage. This does not mean they were momentum trading; likely these firms independently judged but converged on the same liquid stocks that fit their risk frameworks. The danger lies in what happens when they all change their minds at once: forced selling cascades in sync, exactly as occurred in October.
— Nishant KumarMulti-strategy funds show no mercy to losing traders
Nishant explained the brutal internal risk controls of multi-strategy funds: a PM managing $100 loses 5% (=$5) and their allocation gets cut in half; if they continue to 7.5% total loss, they are force-liquidated and fired on the spot. This discipline is the core pitch of the multi-strategy model—prevent any one team's losses from dragging down the whole platform. But when all strategies turned bad at once in October, this very mechanism meant massive positions were cut simultaneously, actually accelerating the decline.
— Nishant KumarMoney rushed to buy shares but top platforms closed to new capital
Millennium aimed to raise $20 billion this year, received over $30 billion in demand, ultimately took $22 billion. Meanwhile, D.E. Shaw, Point72, and Citadel combined returned $35-40 billion to investors. A paradox emerges: investors desperately want multi-strategy fund slots, yet the top platforms are capping size and returning capital. The reason: once you add enough teams, new positions no longer add real diversification—they just dilute the same underlying risk.
— Nishant KumarThursday's crash was not emotion but mechanical chain reaction
Michael Ball attributes Thursday's crash to options market mechanics: when investors buy puts for downside protection, put sellers (market makers) must continuously sell to hedge their delta risk as prices fall, which itself depresses prices in a self-reinforcing spiral. Layer on leveraged ETFs, systematic vol-control strategies, and CTA trend-following funds all forced to cut positions on the same day—four mechanisms in sync—and you breach the S&P level that held for months.
— Michael BallNew cloud operators halved in value, rate sensitivity is the weakness
Take Iron: a cloud stock once pitched as "Bitcoin miners transitioning to AI compute." Its price was $77 in November 2025, down to $21 at recording—roughly a 70% fall. These companies borrowed heavily to build data centers and are highly sensitive to interest rates and customer prepayment quality. Ball thinks the selloff is excessive because backlog orders and compute demand remain robust. But whether 2028 new projects get regulatory blocked remains genuinely unresolved.
— Michael BallHuang's governance turmoil failed to ripple through the market
Markets might have feared that Nvidia governance trouble would transmit through the AI chain. But Ball observed the actual impact was limited: Nvidia itself declined modestly and AI/semiconductor stocks rebounded days later. His reading: AI is now too large for Nvidia or Anthropic alone to move the needle. One company's governance headline cannot shake the whole narrative anymore. Stock moves now track mechanical forces like short covering more than fundamental narrative shifts.
— Michael BallIn their own words · checked verbatim
pots sound like an insult to many of the people in the industry, but that's what they are
Nishant Kumar4:10
So VARA, for example, lost 33% in October. That's like losing a third of your portfolio. But here's a killer statistic. They are still up for the year.
Nishant Kumar8:20
If a PM loses money, risk gets cut. So for example, if you have $ 100 to manage and if you lose 5% of that, so $ 5, at many of the firms, your capital will be cut into half. If you lose another 2.5%, so total loss becoming 7.5%, you are closed out. You can be liquidated and you can get fired immediately.
Nishant Kumar15:40
dozens and dozens of portfolio managers, big risk takers in equity space, in macro space, in credit space, they have been fired or they have been poached by others or they are spending their long non-competes on the beach
Nishant Kumar20:44
To me, the moment that sticks out in my mind is the CoreWeave guy's interview in the FT where he basically talked about raising all the capital in the world.
Michael Ball34:00
You're paying for the hedge. It's in the name. You're inherently long risk assets.
Tracy Alloway49:18
Figures
| Millennium October return | -2.3% | 4:10 |
| Millennium annualized return | ~14% | 4:10 |
| Vara October decline | 33% | 8:20 |
| Kirkoswell October decline (worst month on record) | 6.4% | 9:23 |
| Chris Rokos October decline | ~5% | 9:23 |
| Top 10 multi-strategy funds combined positions (including leverage) | ~$2.2 trillion | 12:33 |
| Millennium 2025 fundraising | target $20 billion, demand $30+ billion, raised $22 billion | 24:51 |
| Anthropic IPO initial valuation | $1.4 trillion | 42:12 |
Glossary
- multi-strat / pod
- Hedge fund model splitting capital among dozens of independent trading teams, managing overall risk through strict controls.
- negative gamma
- Options market makers holding short positions forced to continuously sell as prices fall, amplifying the decline.
- CTA
- Systematic funds trading commodity and futures trends, often accelerating selling during violent volatility.
- de-grossing
- Hedge fund simultaneous reduction of both long and short positions, lowering overall risk exposure.
How to listen
Institutional investors tracking hedge fund allocation, analysts researching multi-strategy funds, and founders and investors wanting to understand the AI bubble transmission chain.
The final section about oil and crude crack spreads has little connection to the main AI thread and can be skipped.