Just caught wind of something that's been rattling around in my head—this 24-year-old Leopold guy everyone's talking about with the $5.5B fund actually *exited* his entire NVIDIA position. Yeah, you read that right. While the rest of the market is still obsessed with chip stocks, he's already moved on.

Here's the thing that got me thinking: his fund went from $1B to $5.5B in about a year, and the strategy shift is wild. He basically called that GPU valuations would be fully priced in by end of 2025, so he dumped NVIDIA, Broadcom, TSMC, Micron—all the obvious AI infrastructure plays. Made money on NVIDIA put options while exiting, which is pretty clean execution.

But the real move? He pivoted hard into infrastructure nobody's really paying attention to yet. His largest position is Bloom Energy—20% of the whole portfolio, like $855M. This is a fuel cell company that can generate power directly for data centers without touching the grid. Their backlog is $20B. Revenue grew 34% last year with 40% projected for 2026. Think about that for a second.

The insight is dead simple: everyone's focused on whether there are enough GPUs, but the actual constraint is power. The grid was built for humans, not for AI data centers. So he's betting on whoever can solve the energy problem. He also loaded up on CoreWeave (another $300M added), which handles GPU deployment infrastructure, plus he's buying up old Bitcoin mining companies for their land, permits, and grid access—basically taking shortcuts on the regulatory nightmare of building new infrastructure.

Even shorted Infosys because he thinks AI agents will replace cheap outsourced labor. That one's tracking pretty well so far.

What's interesting is how unsexy this all is compared to the NVIDIA stock earnings report cycle everyone obsesses over. No one's writing Medium posts about Bloom Energy three months ago. But that's kind of the point—by the time everyone's talking about it, the move's already happened. He's basically betting that physical infrastructure—energy, manufacturing, permits—is where the real bottleneck is now, not chips.

The concentrated conviction is wild too. Putting a fifth of your portfolio into a fuel cell company most people hadn't heard of? That's either genius or spectacular failure, no middle ground. But if he's right about energy being the constraint, then this makes perfect sense.

Curious if anyone's tracking similar infrastructure plays or if I'm just seeing patterns that aren't there.