For much of this year, the site in Loughton announced as Britain's largest AI supercomputer was still operating as a scaffolding yard. The project arrived with ministerial fanfare during a state visit, a £2bn price tag, Microsoft as anchor tenant and more than 23,000 advanced GPUs promised for a 2027 opening. UK Power Networks has now told the developer, Nscale, that the 90MW connection the site requires will not be available until the early to mid-2030s. Planning permission and political backing were in place. Electricity was not.
The link between an Essex data centre and a hospital trust is structural and indirect, and it is worth stating plainly. The NHS's ambitions in AI-assisted diagnostics, genomics and the national data platform depend on compute, and compute depends on electricity allocated through a connection queue that is badly oversubscribed. Ofgem reported in June that 315 data centres were waiting for a connection, with demand running well above what the network can supply.
The health service is also in that queue by another door. New hospital buildings, the decarbonisation of an ageing estate and the electrification of ambulance fleets all draw on the same distribution networks. A trust finance director planning a new energy centre faces the constraints that have stalled a flagship national project, with far less political leverage.
There is a constructive reading of the episode. Loughton makes reform hard to postpone. A project championed from the top of government could not secure power by enthusiasm, which strengthens the case for a rules-based system of priority. Ofgem has proposed commitment fees and delivery milestones intended to clear speculative applicants from the queue, and ministers have signalled that strategically important demand should be given precedence. Each speculative project removed releases capacity for a hospital as readily as for a server hall.
The episode also speaks to state capacity. Ministers announce national capability, but the infrastructure it depends on answers to regulators, network operators and planning authorities that keep their own timetables. The NHS knows the pattern well. Targets for diagnostic centres, digital records and new hospitals have repeatedly met the gap between a ministerial commitment and the dependencies it rests on.
For NHS leaders, the practical lesson concerns sequencing. Digital strategies are usually drafted by technology teams and energy plans by estates teams, often with little contact between them. Any trust or integrated care board expecting to run imaging analysis or ambient AI tools at scale should ask where the compute physically sits and when its power is assured. The reassurance is that current deployments run on capacity that already exists, so no clinical service is waiting on Loughton. The harder test falls on the ambition for sovereign, UK-resident compute for patient data, which now needs a credible date.
Life sciences and health-tech companies face a similar calculation. Drug discovery and genomic analysis consume vast amounts of compute, and firms choosing where to site UK operations will read the connection regime as closely as the tax regime. Nscale's search for on-site generation, including fuel cells, shows that developers will route around the grid where they can, and more than 100 projects are reported to be considering gas. That sits awkwardly with clean power targets and with the NHS's own net zero commitments, and it is a trade-off ministers will have to own.
The opportunity for the health service is to arrive in that debate with a case. The Department of Health and Social Care and its national bodies could set out what health infrastructure and health data compute will need over the next decade, and ask that it be recognised as strategic demand while the connection rules are still being rewritten. A digital strategy that cannot name the year its power arrives has yet to become a plan.
