A 25-year-old built one of the fastest-growing hedge funds on the planet.

He turned $225M into $5.5B in just one year.

But he didn’t invest in AI the way most people think.

Today we’re breaking down Leopold Aschenbrenner and the simple idea driving his portfolio.

Leopold Aschenbrenner was a researcher at OpenAI.

At 22, he was pushed out after raising concerns about how the company was handling advanced AI security.

Soon after, he published a thesis with a bold view: AGI could arrive around 2027.

(AGI = AI that can perform most human-level work across tasks.)

If that timeline is even close, demand for AI explodes.

(Read full report here)

So he built a portfolio around that outcome.

Within months, his fund returned 47% vs 6% for the S&P 500.

The returns look great, but the real story is how he saw the opportunity.

Most people are focused on the latest model.

GPT, Opus, etc.

Leopold focused on what they all rely on.

The physical infrastructure behind AI:

If any of these can’t keep up, AI progress slows.

Training an AI model means running thousands of powerful chips (GPUs) nonstop, processing huge amounts of data to learn patterns.

That takes an enormous amount of electricity, often comparable to what a small city (like Aspen or Bermuda) uses over several days.

That’s a huge bottleneck.

So he built his portfolio around that constraint.

Different companies, same idea: as AI grows, they win.

Software moves fast. New models show up every few months.

Infrastructure doesn’t move like that.

Building data centers, securing power, expanding the grid… that takes years.

And that gap starts to matter.

Because when AI demand moves faster than supply, you get bottlenecks.

And bottlenecks create scarcity.

That’s where the value starts to build.

So the better question is: What does AI need, no matter who wins?

Follow what AI consumes, not just what it produces.

Because in the AI revolution, the companies helping AI run may matter more than the ones building it.