Just been diving into Leopold Aschenbrenner's portfolio moves and honestly, this 24-year-old's investment thesis is pretty wild. His fund went from $1B to $5.5B in about a year, and the strategy shift he's making is way more interesting than most people realize.

So here's the thing that caught my attention: he completely exited Nvidia and other chip stocks. Yeah, you read that right. While everyone's still obsessed with GPU plays, he's already moved on. His take is that by early 2026, the market had fully priced in GPU value. He even sold $300M in Nvidia put options and walked away from the whole semiconductor bet. I get why that seems crazy given how Nvidia stock after hours price movements have dominated market conversations, but his logic is sound—if GPUs are already fully valued, why stay in the game?

The real move though? He's shifted everything toward infrastructure and energy. And I mean everything. His biggest single position is Bloom Energy at 20% of the portfolio—about $855M. Most people hadn't even heard of this company three months ago. They make oxide fuel cells that convert natural gas directly into electricity for data centers. No grid dependency. No waiting months for power permits. Just install it next to your AI cluster and you're done.

Their numbers are insane: $20B order backlog, 34% revenue growth in 2025, projecting 40% for 2026. Supply can't keep up with demand. This is the actual bottleneck he's betting on.

But here's where it gets clever. He also pumped $300M into CoreWeave, this AI cloud infrastructure provider, and started buying up Bitcoin mining companies. At first that seemed random—crypto's been dead, right? But the play is brilliant: these mining companies own land, power permits, and grid access. Instead of spending years getting licenses, he's just acquiring companies that already have them, then converting the infrastructure for AI data centers. It's like buying a bar that already has its liquor license instead of waiting years for your own.

He's also shorting Infosys hard, betting that AI agents will replace outsourced IT labor. Given what we're seeing with Claude and GPT models handling complex tasks, that's not a crazy bet.

The core thesis here is that software is becoming commoditized, so the real value is in physical infrastructure—energy, manufacturing, permits, real estate. AI labs have plenty of GPUs now. What they need is power. Elon's literally launching data centers into space because the sun provides more energy. Google, Meta, everyone's fighting for grid access.

Leopold's essentially saying: forget the chip cycle, the bottleneck is now power generation capacity. And he's putting his money where his mouth is—concentrated, high-conviction bets on companies most people haven't even looked at yet.

Whether his 2027 AGI prediction comes true or not, the infrastructure thesis seems directional. If you're thinking about where capital flows next, this portfolio is worth studying. His track record of being ahead of trends is pretty solid so far.