Leopold Aschenbrenner might be the most interesting investor alive right now. He graduated valedictorian from Columbia at 19. By 21 he was on OpenAI’s Superalignment team working with Ilya Sutskever. He wrote an internal memo telling leadership their security was “egregiously insufficient” to protect model weights from foreign theft. They fired him. The entire Superalignment team dissolved one month later. Sutskever and Jan Leike both left. Two months after getting fired he published a 165-page manifesto arguing AGI arrives by 2027. The Collison brothers, Nat Friedman, and Daniel Gross seeded his fund. He named it Situational Awareness, after the essay. Four employees. No office. His fiancée is Anthropic’s chief of staff. The “portable energy turbines” point is the key to the whole portfolio. Bloom Energy makes solid oxide fuel cells that deliver 100 MW to a data center in 90 days. The grid takes years. Microsoft and Meta are spending hundreds of billions on GPUs annually. Every quarter those chips sit unpowered in a dark rack, they’re depreciating while generating zero revenue. Bloom solves the only constraint money can’t brute-force: time to power. The Infosys short is equally surgical. $57B company, 317,000 employees, $19B a year from outsourced software services. He’s betting AI coding tools collapse the labor arbitrage that built India’s entire IT export economy. Infosys is already down 36% in the past year. And the Core Scientific / CoreWeave connection the tweet spotted is deeper than it looks. Core Scientific signed 12-year hosting contracts with CoreWeave. Aschenbrenner owns 9.4% of Core Scientific and increased his CoreWeave calls by 672% last quarter. These aren’t separate bets. They’re one integrated supply chain. CoreWeave needs rack space and power. Core Scientific has it. Bloom Energy supplies the fuel cells. The bitcoin miner play ties it all together. After the halving crushed block rewards, miners like Cipher, Bitdeer, Riot, IREN, and Hut 8 realized they’re sitting on the scarcest asset in AI: megawatts with high-density cooling already built. He’s buying the conversion of crypto mining infrastructure into AI compute capacity at a discount to building from scratch. The Intel move is the wildest. He sold every share of common stock down to literally 1 share but kept all 20.2M call options. A purely leveraged bet on the 18A turnaround while freeing hundreds of millions to redeploy into more infrastructure. Long Intel, short NVIDIA. Maximum conviction, maximum asymmetry. His 13F equity positions went from $254M to $5.52B in four quarters. 47% net of fees in the first six months while the S&P did 6%. This tells you everything about what he saw inside OpenAI. He watched how much compute they needed to scale, saw the power bottleneck before the market did, got fired for raising the alarm, then bet his entire fund on the gap between what AI needs and what the grid can deliver.