Leopold Aschenbrenner’s Situational Awareness raised money with the premise that AI is much more important than the markets understand and that they can make the right bets to profit from it. The hedge fund started well and got to $45 billion at its peak. Then came the end of July and a few days of AI stocks crashing, which pushed the hedge fund too close to a margin call, so they ended up selling their public portfolio to Citadel.
The fund is still up and they have chips on the table to make more bets and recover from this, but what this should have taught everyone is that proper risk management is really important in this bet. The expected value in this game is better for those who stay in it longer.
Let’s take a few steps back and understand the fundamentals. Situational Awareness started as a blog post, https://situational-awareness.ai/, written by former FTX and OpenAI employee Leopold Aschenbrenner. He laid out this idea that AI has much more potential than the markets understand. Based on these ideas, he created a hedge fund, Situational Awareness, which tries to find the best bets based on an understanding of the industry and rumors running around in Silicon Valley.
The timing was good and his investments did well. He grew the fund from roughly $225 million in seed capital to more than $15 billion in assets under management in less than two years, with the fund returning more than 1,000% since launch. He got praised on X for his ability to pick the winners.
As many hedge funds do, he also used leverage to increase his profits. This must have boosted returns well, but the bad thing is that these AI companies have been so volatile over the past year that investing with big multipliers makes it likely that a margin call will happen.
On July 24th, Leopold posted a letter to investors saying that now was an amazing time to buy more when the shares had dropped for a few days. He asked for more money before August 1st but didn’t survive that long.
The lesson here for me is to not play so risky that I get liquidated before I get to see the full AI bull run. I have high conviction that S&P 500 companies will make huge amounts of money from AI over the next 10 years, but it’s possible that there will be dips that might last even a year. Surviving these dips is crucial.
I don’t really have any leverage trading at all. I have pretty much everything invested, so in a way I’m all-in, but I’ll survive easily in the worst case: two years without touching this money if necessary. It’s very unlikely that I’ll lose my job, so I’m expecting to be able to invest a little bit more each month. If a dip happens, I’m going to find a way to invest more. It might be using a little bit of leverage, but so little that I’m not going to get liquidated unless everything goes completely wrong.
I’m also not fully invested in the main AI companies because it’s possible that companies who don’t invest that much in AI are able to leverage it much better. For example, OpenAI spends billions developing AI that might get commoditized by other companies. I’m bullish on hardware because, at least right now, the direction is that bigger models perform better, meaning that even if the weights are available for anyone, they still need a lot of hardware to run them.
I understand the temptation to use a lot of leverage in this opportunity to squeeze out every penny, but I’d much rather take great profits than take the risk of not getting anything. I’m not going to make generational wealth out of this bull run, but enough to fund my own projects, which hopefully become my magnum opus.