$45 Billion to $10 Billion in a Week: The AI Prophet Who Saw Everything but the Trade Beneath His Feet
He read the AI build-out years before Wall Street did and turned a viral essay into one of the fastest fortunes in hedge-fund history. What he never saw was the position sitting directly under him, and the people who could see it perfectly.
The wedding was meant to be the event of the season. A Tuscan-style villa above Carmel-by-the-Sea, a multi-day celebration, even a pre-wedding colloquium of panels and breakout sessions for guests who like their romance served with ideas. The couple asked for no gifts. By the time the first cars were climbing the coast road, phone lines were lighting up in London, in Chicago and in lower Manhattan, and the groom's $45 billion hedge fund was being taken apart in the dark.
The groom was Leopold Aschenbrenner. Two years earlier, almost no one in finance had heard the name. By that weekend he was the most closely watched young investor in America, and he was watching, in real time, roughly three-quarters of his fund's value disappear.
The obvious moral was reached within the hour, and it is not wrong. A 24-year-old with no track record ran extreme leverage into a volatile sector bet and got what extreme leverage eventually delivers. True. Also the least interesting thing about this story. Because the man who lost the money was, by the fund's own returns, right about almost everything that was supposed to matter. The forecast held. What failed was the structure he wrapped around it.
The prophet trade
Aschenbrenner's rise reads like a parody of the era that produced it. Columbia valedictorian at 19. A stint at the philanthropic arm of Sam Bankman-Fried's FTX, which he left the day the exchange filed for bankruptcy. A year on OpenAI's superalignment team, ended by a firing he disputes. Then, weeks later, a 165-page essay called "Situational Awareness" arguing that artificial general intelligence was arriving faster than governments understood, and that only a few hundred people alive genuinely grasped what was coming. He counted himself among them.
Most people would have taken the speaking fees. Aschenbrenner built a hedge fund and named it after the essay. He raised $225 million in July 2024 from a roster that told you exactly which world he moved in: the Stripe founders Patrick and John Collison, the former GitHub chief Nat Friedman, the investor Daniel Gross. Even Jane Street, a quant firm that almost never hands money to outside managers, came in. His thesis was a single straight line drawn deep into the future: intelligence would demand compute, compute would demand chips and memory and data centers and power, and the companies supplying that build-out would win for a decade.
For nearly two years, the line held. The fund returned more than 1,000 percent since inception, ballooning from a few hundred million to a peak of $45 billion, a scale other managers spend decades reaching. Retail investors pored over its quarterly filings like scripture, hunting for the next name. His dark turtlenecks and his reclusive reputation did the rest. The financial internet had found its oracle.
That worship was the first crack, though no one read it that way at the time. When your positions are public and your every filing is studied, everyone knows exactly what you own. Keep that thought. It is the whole story.
How the machine turned on him
In the middle of July 2026, sentiment on the AI trade wobbled. Cheaper open-source models out of China spooked traders into asking, again, where in the supply chain the value actually sat. Over three trading days, hedge funds cut their positions at a pace not seen in three years, according to Goldman Sachs. For an unleveraged investor, a wobble is a question: did the thesis break, and do I have time to answer. Aschenbrenner did not have time, because he had borrowed $3 to $4 for every $1 of his own, and layered options on top of that.
His prime brokers, Goldman Sachs, JPMorgan and Bank of America among them, began watching the fund's performance by the day and then by the hour. Then came the part that makes this more than a leverage cautionary tale. Rival funds knew, roughly, what Situational Awareness held. Its biggest bets were no secret. So they shorted his largest positions, betting he would soon be forced to sell them himself, and drove the very prices that would trigger his margin calls. Falling prices produced calls for collateral. The calls forced him to sell. The selling pushed the same stocks lower, which produced more calls. Aschenbrenner later described the dynamic to his investors in the plainest possible terms: a bank run, each sign of weakness manufacturing the next.
Consider what actually killed him. Not obscurity. Visibility. The firm named for seeing what others could not was destroyed because everyone could see it. The oracle whose book was studied like scripture had handed his executioners the map.
The twist: he was not wrong
Here is where the tidy morality tale falls apart. When the smoke cleared, the stocks Aschenbrenner had been forced to dump did not keep falling. They soared. Many of the semiconductor and infrastructure names roared higher the day after he let them go, because the thing dragging them down had never been their fundamentals. It had been one enormous, known, forced seller, and once he was gone the pressure lifted.
Which means the rebound proves nothing flattering about him and nothing damning either. It does not show the companies were undervalued, or that his AGI thesis was sound, or that anyone else shared it. It shows only that a forced seller trades at a discount to a patient one. His conviction may have been correct to the decimal. It simply stopped being his call.
The cruelest document in the whole affair is a letter he wrote on July 24, six days before the end. In it he told investors the selloff was one of the best buying opportunities since early 2025, and invited them to add fresh capital on August 1. He was, in other words, trying to buy the dip. He may even have been right about the dip. But an investor on margin carries a second question the patient investor never has to answer: not only "did my thesis break," but "will my broker let me keep holding long enough to find out." Six days after he called the bottom, Goldman, JPMorgan and Bank of America answered it for him.
I keep returning to a moment from 2024, on a podcast hosted by his friend Dwarkesh Patel, when Aschenbrenner was asked how these things blow up. Not blowing up, he said, is task number one and task number two. He knew. He said it out loud, on tape, and then built a machine that could not survive a bad fortnight.
The oracle was reading his own reflection
There is a subtler failure underneath the leverage, and it is the part almost no one is saying. A 40 percent position, amplified by borrowed money, is large enough to move a price. Copycats piled into the same names because he owned them. The AI companies watched their own share prices climb and raised billions at those prices, then spent the money on the very orders that every earnings call cited as proof of demand. Nebius raised $4 billion in March, near the top, because a soaring share price is a gift you bank while it lasts.
So when Aschenbrenner's July letter insisted the fundamentals were accelerating, he almost certainly believed it, and the fundamentals almost certainly were. But some unknowable portion of that acceleration was his own capital and the hype around it, reflected back at him as confirmation. From inside a feedback loop, the loop looks exactly like being right. He was not only forecasting the boom. He was helping inflate the thing he was forecasting, then reading the inflation as proof of the forecast.
Enter the man who feeds on this
Every blow-up needs a buyer, and there is one firm that has spent twenty-five years making itself the buyer of last resort. When Situational Awareness began to list, Ken Griffin's Citadel moved the way it always moves. Griffin, in London at the time, put his people on the fund's book overnight. Millennium's Izzy Englander circled from Europe. Citadel approached late on a Tuesday and had a signed contract before Thursday's open, buying the bulk of the public equity portfolio at a reported discount of more than 10 percent.
The pattern is decades old. In 2001, Griffin chartered a jet to Houston the day Enron collapsed and hired its traders out of the wreckage. In 2006 he took Amaranth's natural-gas book, alongside JPMorgan, after a $6 billion loss vaporized most of that fund. In 2007 his team bought Sowood's portfolio in a deal struck at half past three in the morning while competitors had gone home. Not every raid worked; the 2021 rescue of Melvin Capital did not save it. But the logic never changes. When everyone else is retreating, only a handful of firms have both the capital and the machinery to absorb billions overnight, and that scarcity is itself the edge. It hands the buyer enormous pricing power over a seller who has run out of time.
Cliff Asness, no lightweight, once described his own distressed call from Citadel in 2007 as hearing the Grim Reaper's scythe at the door. The line captures the doubleness of the thing perfectly. What looks like rescue is also harvest. And this harvest was rich: the stocks Citadel bought from the forced seller rallied once the forced seller was gone, and Griffin's flagship fund posted a blowout month in a July when many rivals bled. The disciplined, patient, deliberately unglamorous giant had eaten the visible young prophet, and booked the gain almost immediately.
The room he did not build
Strip the AI branding away and this is the oldest story on Wall Street, the one about the difference between being right and staying solvent long enough to collect. It has a specific shape here worth sitting with.
There is a kind of firm that arranges its entire existence so this can never happen to it. It takes no outside capital, so no lender and no investor can force its hand. It leverages nothing it cannot hold through a bad week. And it makes certain that no one can see its positions clearly enough to trade against them. Those firms are quiet on purpose, and their quiet is not modesty. It is armor. Aschenbrenner built the precise inverse: maximum visibility, maximum leverage, maximum publicity, a book studied like scripture and financed with money that could be recalled in days against assets he intended to hold for a decade. An investment horizon measured in years, collided with financing that could be challenged in an afternoon.
The essay that made him famous argued that a few hundred people possessed situational awareness of the future and the rest of the world did not. He was probably right about the macro future. He simply lacked situational awareness of the trade directly beneath his own feet, the liquidity mechanics that every seasoned risk manager watches precisely because they, and not the thesis, are what end careers.
He kept the Anthropic stake, the illiquid trophy his lenders could not force him to surrender, and about $10 billion, and he is 24. In his letter to investors he wrote that he takes full responsibility, and promised changes to his risk teams. Maybe this is the making of him. Plenty of great investors have a blow-up in their past. But the lesson is not really about him, and it is not about leverage ratios. It is that conviction is not a risk-management strategy. The market does not pay the person with the boldest forecast. It pays the person who is still at the table when the forecast comes true, and it is ruthless about the difference.
None of this is investment advice. It is reporting and analysis drawn from public accounts of events that are still developing; figures and details reflect reporting as of early August 2026 and may be revised.
