A $4.5B Fund Lost Two-Thirds of Its Value Over a Wedding Weekend

Source: Patrick Boyle | Published: 2026-08-10T19:15:06Z

Leopold Aschenbrenner's fund ran into a margin call during his wedding weekend after a 4x leveraged long-short position moved against both legs simultaneously, wiping out two-thirds of the $4.5 billion fund within three weeks.

The day before, his wedding had officially begun at a Carmel-by-the-Sea estate: a multi-day affair, with a ceremony at a Tuscan-style villa, a send-off at a forest spa hotel, and a pre-wedding academic symposium complete with panel discussions. Guests had just survived the symposium and were finally unwinding when they found themselves watching a roughly $30 billion financial implosion unfold in real time.

This is the story of how a 24-year-old man vaporized two-thirds of a $45 billion fund in a matter of weeks — with no trading experience, no risk management team, armed only with a 165-page essay, a 4.5-hour podcast, and a brown turtleneck under a beige zip-up hoodie.

An Empire Built on One Essay

Leopold Aschenbrenner graduated from Columbia at 19, then worked at Sam Bankman-Fried's philanthropic organization until FTX collapsed in late 2022. He went on to OpenAI, and was fired less than a year later — he says for raising safety concerns; OpenAI says for "improperly disclosing sensitive information." He left without signing an NDA, forfeiting roughly $1 million in equity.

What made him famous was a 165-page essay he self-published in June 2024: Situational Awareness. It argued that AGI would arrive by 2027, that AI's recursive self-improvement would change the world faster than anyone could imagine, and called on the U.S. government to nationalize AI labs — despite the fact that just a few pages earlier, he had made clear he almost never supports government intervention in business.

The essay went viral in Silicon Valley's tech executive circles. Tim Ferriss called him "the Nostradamus of AI" — a more apt comparison than Ferriss probably intended, since Nostradamus's prophecies only seem to predict anything after you already know what happened.

On the strength of that essay and a 4.5-hour podcast with Dwarkesh Patel, Aschenbrenner raised $225 million in seed capital from Silicon Valley insiders including the Stripe founders, Nat Friedman, and Daniel Gross. He then leveraged heavily through prime brokers to scale assets under management to $45 billion. The team had eight people.

This Was Not a Hedge Fund

To understand the collapse, you first have to understand his strategy — or why it didn't really qualify as one.

The basic logic of a long-short hedge fund is to strip out market risk: buy the stocks you think will win, short the ones you think will lose, and the two legs cancel each other out, leaving you exposed only to what you're actually good at — picking winners and losers.

Aschenbrenner's trades looked like that on paper: long Korean chipmaker SK Hynix, SanDisk, and Bloom Energy; short software companies like Adobe. But both legs were betting on the same thing: whether the AI revolution would arrive on schedule. When markets began doubting AI's near-term profitability, his longs (chip stocks) fell and his shorts (software stocks) rose — both legs losing simultaneously. Bloomberg's Matt Levine put it well: it's like betting on one team to win and also betting against that same team losing — you think you have two bets, but you've made the same bet twice.

He hadn't hedged any risk. He'd amplified a single directional thesis from two sides.

More dangerous still, he layered in four-times leverage on top of that, borrowing three to four times investors' capital from prime brokers at Goldman Sachs and JPMorgan, and putting the whole pile into the same thematic wager.

The Math That Drives You to Zero

The mathematics behind this collapse deserve their own explanation, because they show why leverage is a systematic destructive force when applied to volatile assets.

Take a concentrated AI stock basket: expected return of 15%, volatility of 40%. With no leverage, the volatility drag is 8%, leaving a real compound growth rate of roughly 7% — a business you can run for decades.

Apply four-times leverage, and the expected return becomes 60%, which sounds appealing. But volatility drag doesn't scale linearly — it scales with the square of the leverage ratio. Four squared is sixteen, so drag goes from 8% to 128%. The actual compound return becomes -68% per year.

The root cause is the asymmetry of gains and losses: a 50% drop requires a 100% gain just to get back to even. Losses dig holes that equally-sized gains can't fill. Expected value gets pulled up by the extreme wins in the right tail, but almost nobody actually lives on that path. The path most people actually walk — the median outcome — points to losses.

The more lethal part: leverage kicks you out of the game before you ever reach that good path. One large enough drawdown triggers a margin call from your prime broker, positions get force-liquidated, and you exit bearing all the drag without ever collecting any of the right-tail reward.

Cliff Asness, founder of AQR and manager of hundreds of billions in assets, was asked about this by a journalist and said the lesson — "never be fully Kelly" — belongs in the folder labeled "things you should have already known." Situational Awareness was, of course, the name of the fund.

Silicon Valley Money Has a Ceiling

Early on, Aschenbrenner continued fundraising on strong early returns and Silicon Valley cachet, but quickly ran into a cultural wall.

Journalist Ezra Klein, interviewing tech writer Jasmine Sun on a podcast, observed something interesting: people in Silicon Valley spend their entire careers watching seemingly impossible problems dissolve once enough engineers start writing code. In that culture, every obstacle is temporary, every wall is just a bug that hasn't been fixed yet.

Washington runs on the opposite logic — things that should be simple get stuck permanently because everyone wants different outcomes. And the things Silicon Valley most wants to build — gigawatt-scale power plants and massive data centers — are Washington-style problems, not coding problems. Permits, land rights, utility commissions, angry local residents: none of that gets solved with better code.

Wall Street asks a different, colder question: what happens if you're wrong?

Aschenbrenner took his fundraising roadshow to New York. Blackstone, the world's largest hedge fund investor, took the meeting and passed. According to the New York Times, a wealthy New York investor put him through standard due diligence in his office: if the AI revolution doesn't arrive on schedule, what's your plan? Aschenbrenner reportedly didn't have a detailed answer. He said he genuinely believed everything would work out.

That's not a risk management framework. New York money didn't invest — and in hindsight, that was an excellent risk decision.

Margin Calls on the Wedding Weekend

In July, Korean retail investors — locally called "ants" — had borrowed heavily on margin to load up on Samsung and SK Hynix as AI plays. When the tech rally stalled, the ants got margin calls and started selling, sending SK Hynix into a sharp decline. At the same time, the software stocks Aschenbrenner was short began rallying. Both legs of the portfolio were bleeding simultaneously.

It was his wedding weekend.

Goldman Sachs and JPMorgan began issuing margin calls. His team started calling rival funds, trying to offload positions to raise cash. According to the New York Times, his representatives described the positions for sale as "distressed hedge portfolios." The counterparties opened the files and found a pile of directional bets in the process of blowing up — once competitors know you have losing positions you're desperate to sell, they front-run your liquidation and profit from the selling pressure you're about to create.

On July 24, he sent investors a letter acknowledging that the fund had "not been immune to market turbulence," adding a postscript: this might be a good opportunity to add exposure, and new capital would be welcome as of August 1. That week was his honeymoon.

Citadel Takes the Trade, and the Logic of a Clearance Sale

The fund began shopping the entire portfolio to major hedge funds. Jane Street and Millennium both examined the positions and declined. In the end, only Citadel stepped in — Griffin took what was left.

This wasn't a rescue. It was a clearance sale: Wall Street's most patient, best-resourced institutions examined a collapsed fund's holdings one by one, and each made their own judgment.

One detail stands out: Jane Street also walked in to look at the positions — and Jane Street was also one of Aschenbrenner's own investors. Your own backer walking in to examine the wreckage and then walking away says something about how Wall Street processes emotion.

Griffin is a regular at these occasions. When Enron collapsed in 2001, he chartered a jet to Houston the same day to recruit talent. He was there when Amaranth Advisors blew up on a natural gas bet. He appeared when SAC Capital came undone. If your fund is on fire, Griffin is the first one to smell smoke — and the last person you want to see walking through the door.

The people who hadn't moved as fast as Aschenbrenner on the way up were sitting on cash, calm-faced, signing acquisition paperwork at dawn when the fire broke out.

Anthropic, and the Smaller World Inside This Story

What kept Aschenbrenner from total wipeout was his private equity positions — primarily a multi-billion-dollar stake in Anthropic. Private assets can't be marked to market daily; prime brokers can't force you to sell at the current price overnight. He tried to sell the Anthropic equity to Sequoia and Greenoaks to raise cash, but the Citadel deal closed first.

The asset that saved him happens to be the company his wife helps run. She serves as Chief of Staff to Anthropic's CEO. The two met at Sam Bankman-Fried's philanthropic organization. Nearly everyone in this story met there.

After all of this, the fund shrank from $45 billion to roughly $8 to $10 billion.

Two Coasts, Two Logics

A week after the collapse, Bloomberg reported that Silicon Valley money was lining up to get in, and the fund was politely turning away new investors for now. When Sequoia partner Pat Grady was asked about it, he said: "Our judgment is that he will be around Silicon Valley for a long time."

Days after the Citadel deal closed, Aschenbrenner wired $400 million into a Sequoia-backed private startup — a company he had already put $100 million into just a month earlier. Having nearly been destroyed by one enormous concentrated bet, he placed another one in exactly the same way.

Silicon Valley's tolerance for failure isn't entirely without basis — in an ecosystem that genuinely produced Google and SpaceX, statistically speaking, betting on supremely confident young people isn't a bad strategy. But betting on a fund manager and betting on a founder are entirely different things: the former is playing with other people's money, subject to the mathematics of volatility drag; the latter is running their own company, where the rules of the game are fundamentally different. The wholesale transplant of venture logic into hedge fund management was the foundational category error in this story — and the two coasts' divergent reactions to that error reveal more than the collapse itself ever could.