Limitless Podcast - Inside The Portfolio Of The 24 Year-Old Who Made $4.5B AI Investing

Episode Date: March 3, 2026

Leopold Aschenbrenner is a 24-year-old investment prodigy who grew a billion-dollar fund to $5.5 billion by focusing on AI and energy infrastructure. Now he's shifting exposure from traditional chip stocks like NVIDIA to firms such as Bloom Energy, addressing the power needs of AI datacenters. Central to his strategy is the prediction of AGI by 2027, outlined in his essay "Situational Awareness." Tune in as we explore Aschenbrenner's innovative insights and investment strategies.------🌌 LIMITLESS HQ ⬇️NEWSLETTER:    https://limitlessft.substack.com/FOLLOW ON X:   https://x.com/LimitlessFTSPOTIFY:             https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQAPPLE:                 https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890RSS FEED:           https://limitlessft.substack.com/------POLYMARKET | #1 PREDICTION MARKET 🔮https://bankless.cc/polymarket-podcast------TIMESTAMPS0:00 Introduction to Aschenbrenner0:42 Insights from 13F Filings1:43 From Chips to Infrastructure3:45 Exiting NVIDIA Puts5:30 Bloom Energy9:27 Bitcoin Mining12:16 AGI and Market Trends14:21 Bull vs Bear Case19:31 The Future of Leopold's Investments------RESOURCESJosh: https://x.com/JoshKaleEjaaz: https://x.com/cryptopunk7213------Not financial or tax advice. See our investment disclosures here:https://www.bankless.com/disclosures⁠

So there's this guy named Leopold Ashbrenner. He's 24 years old. And last year when we covered him on one of our old episodes, he was only 23. He was managing a billion dollars. And he was responsible for investing in new frontier AI ideas and technologies. Today, that one billion dollars, one year later, is worth five and a half billion dollars. This guy who is younger than both of us by a pretty large margin just went on a generational run that made him basically more money than any other fund in the world of AI. And the fact that AI is the hottest market means, The competition is stiff. So clearly, this guy Leopold, he's doing something different than everyone else. And what's just happened in the last week is his new 13F filings came out so we could actually get a look into what he has been trading. So that's what we're going to do here. We're going to go through the filings. We're going to see what this guy has been up to that has yielded him $5.5 billion of money

to manage now. He did this in 12 months as well. So his fund started at the end of 2024 at a right, $255 million. And then within six months, he had outperformed the S&P 500 by eight times, growing it to $2 billion. And then since our last episode where we spoke about his Q3 fund filings, he's gone up $1.5 billion. So he's on a generational run. And like you said, he's like super young. He's done a pretty major pivot, but it's all in accordance to, I guess, his Bible, which is called situational awareness.

It's a 165 page essay where he basically predicts that we're going to reach AGI by 2027. And in this massive essay, he basically writes a story of how he sees the AI revolution unfolding. And I've got to say, he's pretty much nailed it. He called the GPU infrastructure run, and now he's calling a very important pivot that we're about to get into. Yeah, I think the entire thesis is a pivot from kind of chips to infrastructure. And what we're seeing on screen here, actually, this is very cool. he does, we use Claude to create an artifact that is going to walk us through the entire change log that happened between last year and this year.

Perhaps starting with what he dumped, because these positions that he dumped were pretty large. And these include names like InVita where he dumped $300 million and put options in one quarter, right? Yeah, yeah. So a lot of these names that you're going to see on the positions that he exited are very popular names that a lot of people are invested in right now. So then it begs the question, why has he just exited $1 billion worth of these companies? He's dumped NVIDIA, he's dumped Broadcom, he's dumped TSM, he's dumped a micron. These are all the big AI infrastructure companies. He actually made money on dumping NVIDIA, by the way. He held a $300 million pot position, which means that because Mvideo price went down over the last

couple of months, he's likely made money on it. And so it begs the question, why is he doing this? Well, if you read his 165 page essay, he talks about how sometime by the end of 2025 and the start of 2026, he thinks that a lot of the market would have priced in the value of GPUs. And that is coming from companies like Nvidia and Brodcom who kind of like create these chips and then stack these chips ready for AI labs like OpenAI and Anthropic to train their models on. And now he moves his focus onto the major constraint that investors haven't really invested in just yet, energy and infrastructure. The issue that a lot of AI labs face today is,

one, not only do they have too many GPUs, but two, the energy. grid that we have today is built to serve humans and not the massive AI demand that we face today. And that's where his investments are focused on. And I'm getting a kick out of seeing the Nvidia puts being sold and just like exiting Nvidia entirely because when I talk to my friends, when I talk to average people on Wall Street, when you turn on CNBC, Nvidia is the company. Nvidia is the biggest investment. And to see him pivoting away from that, I think is a testament to him just again being ahead of the curve, moving to where things are going and not where they have been, which in his mind is this infrastructure.

It's this moving away from the ships into the infraplay. And that's probably where we can get into what he added. Because these are the stocks that you want to pay attention to. This is what he currently owns. This is what he believes are going to go up. And if it's true, we should see some pretty hefty returns from these. So where do we start with the additions this quarter? Well, thanks to our friend Claude, we've got a really neatly illustrated stack of all of his

investments bucketed into the AI technology stacks. We've got power generation, real estate and facilities, compute and hosting, connectivity, storage and memory, chips and silicon. Actually, I wanted to just round out the last segment, Josh. I noticed that he made a trade on Intel down here. And what he cleverly did was he sold the equity, so he sold the shares that he has, but he still has a massive long position open. So he's kind of freed up liquidity to shove money in other companies.

And the major companies that he's put money into is power generation. particularly this company called Bloom Energy, which until about three months ago, no one had really heard of, but they specifically build power generation turbines that help you generate power to power AI data centers, and he's built a massive position in this to the tune of $855 million.

Maybe it says $876 here, but the farting says $855. Bloom Energy being the number one position is different because it has absolutely nothing to do with the chipsack, and it accounts for 20% of the entire portfolio. And I was doing some research on what they do. And it's actually kind of interesting. They create these oxide fuel cells,

which is a fancy way of generating electricity on site from natural gas. So a lot of times when natural gas is fed into these data centers, it has to go through a turbine, which heats and cools things, and it's a very clunky energy creation process. What these bloom boxes do is they take that natural gas and have a fuel cell that converts it right on site to usable electricity for the data centers. and it's modular, it can be deployed quickly, and there doesn't seem to be a shortage of it.

I believe they're set to create two gigawatts of this just this year. So it's an interesting play on power because I was looking for the Nvidia of energy. Like who is the chipmaker but for energy? And I don't believe there is one. And there's a world in which perhaps Bloom could be that person, that company. So I actually looked into the recent finance as a public company. they have a $20 billion demand backlog. Their revenue was up something like 34% in 2025,

and they're projecting another 40% increase in 26 after that 2025 increase. So there's like no shortage of demand. You mentioned fuel oxide cells. The reason why their gas turbines specifically are super attractive is you don't need to rely on the energy grid for it. So I mentioned earlier, like right now we have an energy grid there's a lot of constraint being put on there because humans need energy and so do AI data centers.

So it tends to drive up the price of energy in the places that those AI data centers are set up. If you use something like Bloom Energy's gas turbines, you don't need to rely on the energy grid at all. You just kind of set it up right next to your AI data center. And hey, Presto, you now have energy at a cost-efficient price that you can now power to train or inference any of your GPUs and your data center. So Broadcom and companies like CoreWeave are going to need this. so hyperscalers in general, as well as AI labs. It kind of reminds me of the, I don't know if you played civilization ever, Josh. I feel like you did.

It's kind of like moving infrastructure and power generation sources to your little silo or settlement to kind of grow it. And that's exactly what's happening here. Yeah, and I think like the obvious play is that there is absolutely no shortage of energy. So the question is who is able to produce the most of it? And they have this tremendously large backlog. But I think the question is, can they actually create enough to start putting a dent into that backlog. The manufacturing becomes the problem here. And with a lot of these

plays, we're now moving into the world of atoms. We're now moving into the world where manufacturing actually matters. And I would love to, we'll do another deep dive some other time to see if they're actually capable of manufacturing at scale. But that's the biggest one. That's a 20%er. What else, what other positions are noteworthy in this new portfolio? Well, he's added like, I think, $300 million worth of CoreWeave. Can you explain CoreWeave? Why CoreWeave is so important here, please, for the people who aren't familiar. Okay. So if you think, think about it, if you're an AI lab, you need GPUs. But it's one thing buying the GPUs from a company like Nvidia, completely different beast setting up those GPUs in rack, servers,

getting the electrical supplies, getting the technical engineering and expertise to maintain your GPU servers, the cooling systems, all that kind of stuff. So what you could do is you could outsources to a company that is known as like a Neo Cloud. That's what CallWeave is. They had a lot of that for you. Broadcom kind of does something like this, but CoreWeave was a smaller company that specialized in this, particularly during the GPU gaming era and is now pivoted to AI specifically. Leopold got in, oh, went in hard on CoreWeave. He added $500 million at the end of Q3 in that episode that we covered back then. And he's added another $300 million here. So we're looking at what is potentially an $800 million position in CoreWeave. Now, the story goes a little deeper.

And that's 700% increase over the quarter. Literally. But there's another layer to this story where he also owns around 10% of a company called Core Scientific, which is CoreWeave's main supplier to setting up energy grids. So if you think about Betts made an investing, Leopold's probably made the largest levered bet in core GPU infrastructure in Neoclods like CoreWeave, as well as energy supply like Bloom. These are his two biggest positions that he currently holds in his fund right now. And what I find interesting is he's beginning to own enough percentage of these companies to become an activist investor to actually influence the outcome of the companies, which I find fascinating. And as I'm going through this, I'm seeing, I mean, power generation is a clear one, but the most

positions that he's added in particular is around real estate. Yes, yes, exactly. It's so fascinating. Like 10 new positions around real estate in general. And this gets to, I guess, the Bitcoin mining play, because what we're seeing here is a lot of Bitcoin mining companies. And I think this seems bizarre and a little wrong. It's like, okay, well, crypto's not doing super hot.

Bitcoin isn't doing well. why is he buying all of these Bitcoin mining companies? And it's because they own the two critical pieces of infrastructure required for AI buildouts. They have the real estate and they have the power. Because, I mean, what does Bitcoin mining require? A tremendous amount of energy and a lot of space to put those GPU racks. And now that Bitcoin mining is not on its way out, but there are better risk-adjusted returns for that real estate, it seems as if he's betting on the idea that these Bitcoin mining companies are going to either sell their rights and their permits or just pivot to AI data centers in general. To be clear, he does not have any interest in mining crypto with these

companies. He's buying them for their licenses. They have access to the electrical grid. Usually these kinds of things take months or years to attain. That's why, although you see all these major announcements from the likes of Meta, Microsoft and Open AI saying, hey, we've got $1.4 trillion worth of compute partnership signed, but you don't see that translated into the models that they're pushing out just yet. That's what GPUs being pushed out. are always a generation behind. It's because they don't get access to this permitting. Leopold has sidestepped the entire thing and is buying up these smaller companies that already have access to it. And so he's like gutting them completely of their crypto services and just

repurposing them for training AI models specifically and being that provider for them. It's kind of like if you take over a bar that already has a liquor license rather than applying for a liquor license and winning many years to get it. It's just a hack. It's really useful. One of the things I've admired most, I think, about the thesis that he's had and just kind of watching it play out over the last year is how simple and yet effective it is. It's like, okay, like I understand Bitcoin mining and obviously they have permits and obviously they have energy and obviously every AI company wants this. Why would everyone not be buying these things? And I think the simplicity of these ideas is what has kept a lot of people away from investing, but over and over, it continues to be proven right.

And there's another thesis that he had in that early paper that we mentioned that was 165 pages about some timelines, particularly around AGI, which I thought was interesting, because this guy seemingly right about everything. By 2027? By 2027? Are we actually going to get that? And in order to answer this prediction,

we have a polymarket open, which is predicting whether Open AI announces it has achieved AGI before 2027. And I know the odds were against Leopold when he raised this fund, but like, man, these odds are at 13%. So it seems like a stretch that he's actually going to be getting AGI in 2027,

and maybe the investment thesis is right and the timelines are slightly wrong. I don't know. This seems like kind of insane. Yeah, it's like a, it's a small percentage chance, but I have to say he's been, he was originally criticized for his essay being too outlandish and inaccurate.

A lot of some, like 50% of people thought that it would be happening in the next couple of months. A bunch of people said, now this is going to happen until 2030. He's the only guy who's had the prediction that has been nailed so far. He called out GPUs before there was a GPU rush. Now he's calling out energy.

infrastructure before there's an energy infrastructure rush. So I think he's ahead of it here. Yeah, if you believe he's right, there is a lot to be made by pressing that yes button on Polymarket. And thank you Polymarket for sponsoring this part of the show. But it's not only long positions that he has. He's also short a position. And it's particularly on one company. That company is InfoSys. Infosis is basically a company that specializes in IT outsourcing, particularly in India. And their business model is completely based on having cheaper labor than you could get access to in Western countries like America or Europe or continents like Europe. So the point is, hey, just outsource all your admin IT stuff to us and we got you. And the bet that I think

he's seeing here is he's seen the rise of products like Claude Code and GPT Codex 5.3. And he's realized, wow, these models are actually now good enough to automate not just menial work, but really important IT processes. And he's taken out a massive short on this company. This to me is like one of his more intellectual bets. And it's more kind of current to the trends that we're seeing right now. So it's good to see him basically put his money where his mouth is. It's so fascinating going through this and seeing the kind of DCs play out.

I'm going to have to add some personally. I mean, this may cause me to start diversifying a little bit. And maybe, I mean, there has to be some issues with this, right? It's like going through the bull case, going through the bear case, what is there to criticize and be cautious of when you're getting into a portfolio like this? And I mean, the first thing that comes to mind is that this is a 24 year old. And not to just nothing against that because clearly it hasn't worked against him. But I'm not sure he's had, I mean, he certainly hasn't had the experience that a lot of these other investors have.

Is that an advantage clearly to some extent. But at what point does that advantage kind of crumble apart? If so, that seems to be like one of the other things I'm concerned about. The other thing I'm concerned about is that the fund is kind of a single thesis bet, where if AI infrastructure and spending plateaus at all or the macro environment turns, every position in this portfolio correlates the downside, there's not much hedging against that. So there is some potential for holes, but nothing previously has signaled that this is going anywhere but up only.

Yeah, I mean, if you look at some of the most famous investors of our time, it was never about the largest amount of money that they made in a single year or a single year. quarter, it's always just to have been about consistent returns year upon year, decade upon decade that has compounded. Leopold has had an amazing start, and he's beaten the above, he's above average for any kind of hedge fund based in any sector versus just AI, but he's yet to prove himself on a multi-decade kind of time span. So time will tell.

All I will say is for someone who's got fired from Open AI as, I believe it was an alignment researcher, but for someone who's had like major insight and had the biggest or boldest prediction as to where AI is going, he's been the only one to kind of nail it so far. So he's gone full conviction into his 165-page thesis, which I presume he put a lot of thought into. And it's paying off for him now. Now, will that change in the future, maybe? But what you could treat these filings and these investments as is kind of like a

real-time tracker of where he sees the constraints in this AI race. And I want to emphasize this point for a second. Originally, his fund thesis was GPUs. People are going to need GPUs and people are underpricing that opportunity. Now he's saying that opportunity is not overpriced, but it's market priced in. And now he sees the constraint pivoting

or moving towards energy infrastructure. You've got the likes of Elon Musk that is launching data centers into space. Why? Because there's more energy in the sun. You've got everyone for the likes of Google, meta, broadcom, and video. All of them are kind of investing in data centers or data center infrastructure

that gets access to these electrical grids. And he's just putting money where that demand is. And I think it's a smart move. Yeah, I read this great post by Naval the other day. I forget exactly what it said. But the idea of it was that basically software companies who only rely on software are going to have a very difficult time in the future because it's so easy to actually create and generate custom software.

And I think what this pivot is reflective of, not only specifically around building the architecture, but just bets on the physical work. world and bets on actual manufacturing and bets on factories and bets on energy and infrastructure, things that can't be built using AI, things that actually need human horsepower, like things that need to be built, things that need permits, things that need legislation. This is like the hard physical energy and infrastructure required. And I think that tracks to kind of where the future is going. Energy is the one thing that no one could get enough of. And I mean, that power generation play, the real estate play, it's all around one thing. It's just powering the future. We saw through the

last earning season, $650 billion committed to CAPEX from just a couple companies like Google, Amazon, Nvidia, or not Nvidia, but whoever else is going to be paying Nvidia. And I think it's a testament to just how much money is going to be thrown at solving this problem. And this portfolio is ready to capture all of that upside. It's really pleasing to see someone make trades at the right time, in the right places. He has made some what you would consider risky bets. Like, again, not many people have heard about Bloom Energy, unless you're like deeply within the energy infrastructure place, but they're kind of like a second, maybe considered first-tier energy company, but specifically used for their portable energy. So like he kind of put these two puzzle pieces

together and thought, yeah, electrical grids aren't going to support this. So let me invest in this company. And he just goes in full conviction, right? We're talking about almost like a fifth of his entire portfolio in this one position. So extremely concentrated, high risk, high conviction. But if it does pay off, that's what ends up in these kind of like 4.5 to 5x returns from just a year and a half ago. Man, and we just got to respect it, man. Like one billion to five and a half in a year. No, I'm not salty at all. Josh, I'm not salty at all that I'm older than him and having... I will say, my portfolio has not done as well as his. So perhaps we should be taking our own advice and start copy trading because, like, clearly this guy is seeing something with the clarity

that we are lacking, or at least the conviction that we have not been able to place in our actual trades. But that is the update on Leopold's portfolio. I mean, it's, it's an amazing feat that he's managed to do. And it seems directionally correct again. Like this new pivot to hardware to infrastructure energy, I'm super bullish. I'm not sure there's really any other things to be said other than if you agree with the portfolio. I mean, not financial advice. This is just this dude's portfolio, but it seems like it is promising and it is poised to do very well this year. And yeah, that's the update on Leopold. Yeah. I'm curious what Aweller is. I think you guys have been extremely active in comment sections on platforms like Spotify, Apple, as well as YouTube.

Please keep the comments incoming. I'm curious whether you think our investment analysis here is professional grade, Leopold Aschenbrenner grade, or if you think we're completely incorrect and we're missing an obvious story. No, you know what I want their number one top stock pick of the year. Yes, actually. Like Leopold's throwing it down on Blume Energy. I want to know what is your blue energy? Like what's the thing that we're missing that we need to know. about that is going to 5x again this year.

Yeah, that's how long. Help Josh and I, like, make better portfolio returns in this 24-year-old kid. And to the rest of the community. Yes, please. We're scrolling through the comments. Everyone wins if you pick a winner. So pick a winner, drop it in the comments. Share it with your friends, along with this episode.

Subscribe to our sub-sag newsletter, which is coming out again, I think, the day you're listening to this. Uh-huh. EJaz is writing the piece for this week. It's very cool. Every Friday we release a newsletter updating everything that happened this week. And we're now doing a little more episodes, possibly four episodes. It's possibly four episodes per week now.

So we're going to keep this train going. Thank you so much for your support. Thank you so much always for sharing, for giving us five-star ratings. And yeah, we'll see you guys in the next episode. See you guys.