A 4x-Levered Bet on the AI Narrative: How a Genius Fund Died of Volatility Drag
Aschenbrenner's failure was not a failure of intelligence. It was what happens when you put 3-4x leverage behind a single high-volatility AI theme: volatility drag and asymmetric payoffs made losing control a matter of arithmetic. Markets do not reward concentrated, narrative-driven bets.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
You can get the expected value right and still be ruined by repeated bets
Aschenbrenner is a believer in effective altruism, a school of thought that encourages taking risks with high expected returns even when repeating the wager leads to certain ruin. That is the seed of everything he did later: correct in theory, fatal in practice. The show connects that philosophy directly to how FTX ended — you think you are managing risk, when in fact you are just talking yourself into ignoring the tail.
— Patrick BoyleThe 165-page AI prophecy was a Silicon Valley group chat, written down
Boyle read every page of ‘Situational Awareness’ and concluded it is not prophecy but a roundup of views the San Francisco crowd had already talked to death. The essay contradicts itself — professing faith in the private sector on one page while calling for the government to nationalize the AI labs on another. Its popularity came not from insight but from restating the industry group chat back to readers in a form that made them feel they had a grip on the future.
— Patrick BoyleOne PDF underwrote a $45 billion trading book
Aschenbrenner had no prior trading experience. He claims he shorted the market in 2020, but he was 17 at the time and could not have opened a brokerage account. On the strength of a 165-page PDF and some podcast appearances, he raised $225 million from Silicon Valley, and after leverage the trading book reached $45 billion — all of it run by 8 employees. This is not a question of competence. It is structural absurdity.
— Patrick BoyleNew York passing on the deal, unanimously, was the smartest decision that night
Silicon Valley asks: can it be built? Washington asks: whose permission do we need? Wall Street asks: what if we're wrong? Aschenbrenner's essay converted into enormous checks in Silicon Valley, but New York institutional investors asked the standard question: what happens if the AI revolution doesn't arrive on schedule? His answer was ‘I genuinely believe it will happen’. No detailed plan, no contingency for the position. The New York money passed, unanimously — in hindsight, the smartest risk decision of the night.
— Patrick BoyleBoth the longs and the shorts were bets on the same AI story
The logic of a traditional long/short fund is to buy the good companies and short the bad ones, stripping out broad market risk and keeping only stock-selection skill. Aschenbrenner bought chip stocks and shorted software stocks — both sides of the book betting on the same AI story. The moment the market doubted that AI would make money soon, the longs fell and the shorts rose, and he lost on both sides at once. In Matt Levine's words: this did not reduce risk, it concentrated risk in a different direction.
— Patrick BoyleAfter losing $30 billion, he called it a rare opportunity
In the letter he sent investors after the losses, he described the market turmoil as ‘ructions’ — Boyle's dig is that the word suits a neighbor failing to pick up after their dog, not losing $30 billion during your own wedding. But the letter is the real highlight: he called this a rare opportunity and invited investors to add to their positions on August 1, while he took one-on-one calls with furious investors from his honeymoon.
— Patrick BoyleCitadel taking the book was not a rescue, it was an auction
Citadel bought the entire collapsing portfolio, but this was not a rescue, it was an auction. Griffin beat out Jane Street and Millennium to win the bid. Jane Street was itself an investor in Aschenbrenner's fund, coming full circle to bid for his wreckage. The same thing happened with Enron, Amaranth and SoCo — Griffin is always the first to smell smoke, and he is the last person you want to see when you are the one on fire.
— Patrick BoyleAt 4x leverage, even a great expected return compounds to zero
This is the mechanism that killed the fund: leverage scales returns linearly, but volatility drags on compounding as a square. Suppose the AI stocks you hold have an expected return of 15% a year with 40% volatility; unlevered, you compound at 7%. Lever 4x and the expected return becomes 60%, but volatility drag goes from 8% to 128%, pulling you back to negative 68% growth. The expected value looks fine; the path goes to zero with certainty.
— Patrick BoyleIn their own words · checked verbatim
His actual trading experience before starting the fund appears to have been non-existent.
Patrick Boyle5:29
Wall Street asks a colder question. What happens when you're wrong?
Patrick Boyle14:22
It's a bit like betting a hundred dollars that one football team will win a match, and then as a hedge, betting another hundred dollars that their opponents will lose.
Patrick Boyle17:35
It's not traditionally the word used for losing 30 billion dollars during your own wedding.
Patrick Boyle21:49
So, Leopold was, in the end, saved by the fact that some of his assets were too difficult to sell in a panic.
Patrick Boyle27:03
The expected return is propped up by a thin sliver of wonderful outcomes way out in the right tail.
Patrick Boyle33:28
Figures
| Fund drawdown | roughly 67% lost in one month | 1:05 |
| Assets under management | $45 billion | 2:15 |
| Aschenbrenner's age | 24 | 0:00 |
| Fund employees | 8 | 11:13 |
| Initial capital raised | $225 million | 11:13 |
| Leverage | 3 to 4x | 18:39 |
| Citadel assets under management | $71 billion | 23:53 |
| Single investment in a Sequoia portfolio company | $400 million | 38:38 |
Glossary
- volatility drag
- The compounding loss caused by high volatility: the realized growth rate is roughly the mean minus half the variance.
- margin call
- The broker's forced demand that a fund immediately post more cash or liquidate positions when collateral value falls short.
- expected return
- The probability-weighted average of all possible returns; easily inflated by extreme good outcomes on the right tail, which masks the real risk of a losing median.
- long/short equity
- Holding longs and shorts simultaneously to hedge out broad market moves while retaining stock-selection skill — but if both sides bet on the same theme, the hedge is meaningless.
- lockup agreement
- The contractual term by which a fund bars redemptions for a set period, restricting investors from pulling money out during a crisis.
- full Kelly
- Betting a single position up to full leverage in pursuit of maximum logarithmic growth, ignoring volatility and the risk of permanent loss. Cliff Asness used it to needle the fund's approach.
How to listen
VCs and hedge fund practitioners who follow narrative investing and leverage blowups, plus any founder who has been sold on the idea that AI solves everything.
Skip the wedding gripes and celebrity anecdotes at the top. Start at 4:28 with effective altruism and 16:29 with the deconstruction of the hedge — that is enough.