A 25 year old built a $45 billion AI hedge fund and posted 400%+ returns. Then it all collapsed in one week. The AI industry feels like an action movie with a new plot twist that happens every single day. And today we have a rather interesting story. It broke a couple hours ago and I'm going to break it down so that no matter your background, whether you're technical or not, you can understand what's happening. This 25 year old got fired from OpenAI. He then wrote a viral essay which turned into a $45 billion hedge fund that posted over 400% returns. This week, several distinct things happened that started the collapse of his hedge fund and we're going to jump into the story. Three things that I'm going to talk about in this video that happened in the market this week, in addition to the collapse of his hedge fund, tell you a lot of what's happening in AI right now. Let's start by talking about who this 25 year old hedge fund manager is, Leo Aschenbrenner. He is extremely intelligent. He graduated from Columbia University at the age of 19 and was valedictorian for his class. He then at 22 joined OpenAI. He joined OpenAI's super alignment team which was pretty much tasked with figuring out how to control AI that's smarter than humans. In really simple terms, they were tasked with figuring out when you build an AI model and it's smarter than you, how do you make sure that it still does what you want instead of running off to do its own thing, which is really interesting because with the hack of OpenAI and Hugging Face, we're seeing this in play. In 2024, he wrote a memo to the board of directors at OpenAI pretty much talking about Chinese espionage and how he thought that OpenAI's security was not adequate enough. OpenAI thought he was leaking confidential information and fired him that year. Two months after being fired, he published a 165 page essay on situational awareness and this went viral. The essay pretty much said that AGI, artificial general intelligence, is going to be here by 2027 and the companies that are building the infrastructure to support it are going to be the winners of this next generation. So what is AGI exactly? The tools that we are very familiar with, whether that is Claude, ChatGPT, they're considered narrow AI because they are very good at completing very specific tasks, whether it's writing an essay or coding. But they still can't do everything that a human can do and that's essentially what AGI is. It's an AI that can do everything that a human can do. It would be able to reason, adapt, and learn and solve problems that it's never seen before. Given the virality of his essay, he turned it into a hedge fund that was investing in a lot of AI infrastructure. Now, let's understand what a hedge fund is at a high level. In its simplest terms, a hedge fund is a pool of money that can invest in multiple different things with little restrictions, but they oftentimes have sophisticated and very well thought out strategies. They can invest in stocks, commodities, currencies, private companies, oftentimes in ways that a normal investor can't. One of the key differences from normal investing is the use of leverage. Pretty much borrowed money that helps you amplify your bet. Leo's strategy made bets on companies like CoreWeave and Bloom Energy. They shorted software stocks like Adobe and made investments in private companies like Anthropic. From late 2024 to 2026, this strategy worked really well. He started with 225 million and that grew at its peak to $45 billion. Here are three things that happened leading up to the collapse and then we'll go into the details of it. The earnings releases of major tech companies revealed that the AI spend and ROI is still unclear. For example, you had Alphabet announcing 200 billion in AI CapEx and got punished 7% by investors because they are no longer rewarding spend and wanted to see the ROI. Second, you have Chinese AI companies proving that frontier AI does not need to be expensive. For example, in a video that we talked about last week with Kimi K3. Third, you have Meta signaling that it has excess GPU capacity and could start selling it. If the biggest buyers of AI infrastructure become sellers, the scarcity narrative that drove valuations higher over the last few years is over. Stepping into this week, following those three events, Leo found himself overexposed. The market was telling him that his thesis was under pressure while being at four times leverage. Both sides of his book moved at the wrong way at once and I'm going to explain what that means. Essentially, he had two bets running simultaneously. Long bets where he owned stocks expecting them to go up and short bets where he was expecting that other stocks would go down. Normally, these can offset each other and reduce risk, but when both sides are moving against you at once, they can magnify losses. The four times leverage made this event so much worse. Imagine you have $100 to invest and you borrow $300 from banks to invest more on that same strategy or bet that you're making. If your investment drops 25%, you have lost $100 which was the money that you had in the bank to make as part of this investment. You still owe that $300 back to the bank and this is what triggers a margin call where the people that lent you the money say they want to be paid back now. These are called prime brokers and in this situation, it was JP Morgan, Goldman Sachs, and Bank of America all saying they wanted their money back now. They could not meet that request, so as a result, they had to sell their entire public stock portfolio and they ended up selling it to Citadel, a large hedge fund owned by Ken Griffin. Despite selling their entire public stock portfolio, they still kept some of their positions like their privately held ownership stake in Anthropic. To be clear, he didn't abandon his thesis, he just could not afford the leverage to wait it out. For the past two years, the market has been rewarding AI infrastructure bets and in the recent weeks, it seems like that sentiment is changing. Leo Aschenbrenner may still be right about AGI arriving in 2027 and his thesis can still survive. The sentiment around AI is changing in real time and what happened to Leo Aschenbrenner's fund today is an example of that. It feels like the era of rewarding belief is coming to a close and the era of demanding returns is quickly beginning.
