AI data centers consume orders of magnitude more electricity than traditional facilities. And the demand just went vertical. Every autonomous agent running around the clock is an inference engine that never sleeps — and every inference cycle draws power. The always-on era means always-on energy consumption.
The numbers are staggering. The AI companies have committed over $650 billion in capex for the year, and a massive share of that is going toward power generation and grid infrastructure. At a White House meeting in March 2026, the Energy Secretary said the nation that leads in AI will be the "military superpower" — and the government agreed to bring down regulations so AI companies can build their own power generation facilities.
This is not a future problem. The hyperscalers are monetizing compute capacity the moment it comes online — but they can only install capacity as fast as they can power it. The bottleneck has shifted from chips to energy.
The data center power problem has a hierarchy of solutions. Natural gas is the near-term answer — it's available, scalable, and the infrastructure exists. But gas generation has emissions constraints and fuel cost volatility. Solar and wind are intermittent and require massive battery storage for baseload reliability.
Nuclear is the long-term answer. It's the only energy source that provides clean, reliable, 24/7 baseload power at the scale AI demands. The AI companies know this. The U.S. government knows this. The regulatory environment is shifting to accommodate it. And the companies that generate nuclear power — or supply the fuel for it — are positioned at the center of a structural demand curve that will persist for decades.
We hold positions in the power layer of the AI stack. The specific companies, the thesis behind each, and our assessment of the nuclear timeline are available to members.
The energy constraint is the most underappreciated bottleneck in the AI revolution. We've been positioned for it since early 2025. Members get the full thesis.
See the Power Positions →The energy bottleneck will define the pace of the AI revolution for years. We track the power landscape daily — members get every position and update.
Subscribe — $297/quarter →