Source

Core Thesis

The binding constraint on who builds AGI first is not GPU supply — it's watts per rack. You can order 100,000 GPUs and get delivery in six months. You cannot add 500 megawatts to the grid in six months. Grid interconnection takes 2-3 years. Bloom Energy's solid oxide fuel cells can power a data center in 55 days.

Leopold Aschenbrenner — the 24-year-old ex-OpenAI researcher who wrote the 165-page "Situational Awareness" essay arguing AGI arrives by 2027 — translated this prediction into a pure energy play. He sold every share of NVIDIA and Broadcom in Q4 2025 and built a portfolio around the power gap.

His fund went from $254M in equity positions (Q4 2024) to $5.5B (Q4 2025). Beat the S&P by 47% in its first six months.

The Logic Chain

1. AGI requires scaling — from GPT-4 to superintelligence requires data centers consuming more electricity than most American cities
2. GPU supply is expanding on a known curve — fabs are building, lead times are shrinking
3. Electricity supply isn't — grid interconnection is 2-3 years, permitting even longer
4. The gap between those two curves is the trade — whoever solves power fastest wins the AI race
5. Fuel cells bridge the gap — Bloom Energy delivered a fully operational data center power system in 55 days, a month ahead of schedule
6. BTC miners have the infrastructure — massive power-connected facilities originally built f...

Takeaway

Status: Active Priority: High Type: Secular thesis Scanner: ai-power Created: 2026-04-15