Not a power problem. A timing problem.

Britain wants 6GW of AI-capable capacity by 2030; the grid can connect about 5GW of it on current forecasts, and the big engine OEMs are sold out until 2028. Around 140 data centre projects are queuing for 50GW of connection capacity. Every month a built data hall waits for energisation is revenue lost and contracts at risk.

The queue behind those numbers is a story of its own: The queue is the product.

The operators winning right now are the ones treating power as something you build, not something you apply for: behind-the-meter generation as a bridge until the grid arrives, or as permanent primary supply with the grid as backup. An array of containerised units scales with the build-out, hall by hall, and can be redeployed when the grid connection finally lands.

And because a generating engine is also a heat source, its waste heat belongs in the cooling loop: absorption chillers turning exhaust into chilled water, lifting total efficiency towards 95% and taking load off the electrical plant.

Megawatts in months is a manufacturing problem. We happen to run the factory. If you're doing this maths on a live project, talk to an engineer.