There is a PDF.

165 pages, free, been sitting on the internet since June 2024, and I keep coming back to it the way you keep pressing a bruise, not because it feels good but because something in you needs to confirm it’s still there, still real, still saying what you think it’s saying.

Leopold Aschenbrenner wrote it after OpenAI sacked him.

He was twenty-two.

Valedictorian at Columbia at nineteen, then a year on the superalignment team, then out the door in April 2024 over a alleged leak he says he didn’t do, and two months later he published Situational Awareness: The Decade Ahead and the argument in it, boiled right down to the bone, is this: AGI is not a distant theoretical problem for our grandchildren to sort out. It is arriving now, this decade, possibly this half of this decade, and the physical infrastructure to support it, the power stations and data centers and cooling towers and grid upgrades and the whole grinding material reality of running computation at a scale we have never attempted, does not yet exist, and whoever builds it will own what comes next.

He then built a fund around that argument.

Situational Awareness LP started with $254 million in late 2024. Its Q1 2026 13F disclosed $13.68 billion in positions.

It has since, per the Wall Street Journal, crossed $20 billion in assets under management, is up 270 percent in 2026 through May, and up over 1,000 percent since it launched.

From $254 million.

In under two years.

The positions in the fund are not what you’d expect if you were betting on AI in the obvious way. He’s not long Nvidia.

He’s long power producers.

• Bloom Energy, whose fuel cells let data centers generate electricity on-site rather than wait for a grid that can’t cope.
• Bitcoin miners who pivoted their facilities toward AI compute hosting.
• Constellation Energy. The fund held $8.46 billion in put options against the chip makers as of Q1, which is a strange move for someone whose entire thesis depends on AI scaling, and the explanation, the real one, is that he thinks the physical layer, the electrons and the buildings and the cooling, is where the money actually lives, not the semiconductors, and the market hadn’t figured that out yet when he started buying.

The document is free.

165 pages and it probably takes you four hours and by the end of it you will either think this man is completely wrong about everything or you will find yourself googling electricity grid capacity at eleven at night and feeling genuinely uneasy about what comes up.

Those seem to be the only two reactions.