National Grid just put $1.75B into a company that builds power plants specifically for AI data centers. Your cloud provider's energy bill just became your problem.

A utility company that has existed since 1990 just made its biggest bet ever.

Not on wind farms. Not on nuclear. On a startup that builds dedicated gas-fired power plants for AI data centers.

National Grid is investing $1.75 billion for a 35% stake in Joulent. The first project: a 2.67 gigawatt gas facility in West Texas, built with Chevron, under a 20-year power purchase agreement with Microsoft.

Read that again. A 20-year contract. For power. To run AI.

The numbers tell you everything. Data center electricity demand grew 17% in 2025. Overall global electricity demand grew 3%. AI is consuming power five times faster than the grid can build it.

Microsoft's president of cloud operations called it "closer coordination between energy and infrastructure." That is corporate speak for: we cannot get enough power from the grid, so we are building our own power plants.

This is the real AI infrastructure crisis. Not chips. Not models. Not talent. Power.

National Grid plans to connect 10 gigawatts of data center demand across the UK and US in the next five years. Joulent has a multi-gigawatt pipeline beyond Kilby. The utility expects returns exceeding its regulated 9-10% because this is riskier. That means someone is paying a premium for AI power.

Audit your infrastructure roadmap. If your AI deployment depends on grid power availability, you are already behind. The enterprises winning the AI race are the ones solving the energy problem first. Compute is infinite if you can power it. Power is finite if you cannot.

SOURCE: https://www.prnewswire.com/news-releases/national-grid-ventures-to-invest-1-75bn-to-accelerate-power-solutions-for-us-data-centers-and-ai-302815750.html
VERIFIED: PR Newswire (National Grid official release), MarketScreener (Reuters), Data Center Knowledge
SIGNAL: AI power demand is now a 20-year infrastructure commitment. Enterprises that treat compute as their bottleneck will lose to those that solve for energy first.