In a substation control room somewhere on the edge of the national grid, an engineer is looking at a queue that no amount of overtime will clear. It is not a queue of vehicles or patients or planning applications. It is a queue of electrons, promised many times over to data centres that do not yet exist, for workloads that barely existed eighteen months ago. Between November 2024 and June 2025, requests for grid connections from data centre operators rose from 41 gigawatts to 125 gigawatts, a figure that dwarfs the roughly 45 gigawatts of peak demand the entire country currently draws. The connection queue grew by 460 per cent in that same window. Some projects are now being told to expect a wait of fifteen years.

This is usually reported as a story about the technology sector's ambitions running ahead of Britain's ability to power them. That framing is correct but incomplete. The NHS is drawing from the same grid, and its own digital ambitions, repeatedly described by ministers as central to productivity gains and waiting list recovery, are not insulated from this bottleneck simply because they carry a different label.

The Federated Data Platform, ambient voice technology in outpatient clinics, and the slow rollout of agentic AI tools in diagnostics and triage all depend on data infrastructure that must draw power from somewhere. NHS trusts commissioning new data centres to support electronic patient records, or hospitals seeking grid capacity for additional MRI suites and imaging equipment, are competing for connections in the same national queue as commercial hyperscalers. There is no separate lane for public health infrastructure. A trust modernising its estate is, in grid terms, indistinguishable from a private operator building server halls in Slough or Newport.

This should concern NHS leaders more than it currently appears to. The government has spent the past year insisting that AI adoption in the health service is not optional, that it is the mechanism by which waiting lists will fall and clinical productivity will improve without proportionate increases in staffing. Sir Jim Mackey's accountability agenda and the broader push toward a leaner, digitally enabled NHS both assume that the infrastructure to support this transformation will simply be there when needed. The grid data suggests otherwise. If commercial demand for data centre capacity continues to expand at anything close to its recent rate, NHS trusts seeking connections for their own digital infrastructure could find themselves queued behind private sector applicants with deeper pockets and more commercially urgent timelines.

The relevance here is structural rather than acute. No hospital is going to lose power tomorrow because of hyperscaler demand, and the government has not yet had to make an explicit choice between a data centre and a diagnostic suite. But the absence of any ring-fenced grid capacity for public sector health infrastructure means that choice is being deferred rather than avoided. Energy policy and health policy are converging in a way that neither department appears to have fully priced in, and the DHSC's digital strategy currently reads as though grid capacity were a background assumption rather than a live constraint.

For policymakers, the practical question is whether health infrastructure should be treated as a distinct category within grid connection planning, given the political weight already attached to AI-driven NHS reform. For NHS leaders, it is a question of whether digital transformation plans have accounted for a five, ten or fifteen year lag in the very power supply those plans depend on. For life sciences and health-tech firms partnering with the NHS, it is a reminder that the infrastructure underpinning their products is now a matter of national industrial competition, not simply local planning permission.

Britain's AI energy crisis has, so far, been discussed almost entirely in the language of economic opportunity and commercial ambition. The NHS's stake in that same grid queue suggests the conversation needs to widen before the choices being deferred become choices that are made by default.