There is no indication that you are on the verge of what the government has referred to as Britain's showpiece AI development as you drive past Chapelhall in North Lanarkshire. The site draws perhaps 25 megawatts from the grid today. DataVita's public material describes something else entirely: energy parks generating more than a gigawatt, wind and solar infrastructure rivalling the country's largest onshore windfarm, power equivalent to a small nuclear reactor. The planning applications on file cover a few square kilometres. The ambition described requires many times that. Somewhere between the press release and the land registry, the project stopped being a description of what exists and became a description of what its backers hope will eventually be approved, financed and built.

This is not a tale of a single data center in Scotland. It is a story about a habit, now visible across British AI policy, of announcing transformation before securing the infrastructure that transformation depends on. The government designates a site or a programme as nationally significant. Developers describe capability in the present tense that properly belongs in the conditional. Scrutiny, when it comes, arrives after the announcement rather than before it. Correspondence obtained from Scottish officials shows they knew about the power problem privately while endorsing the scheme publicly. That is not fraud. It is something more familiar and more troubling: a policy culture where the announcement has become the deliverable, and the engineering is left to catch up.

The NHS should recognise this pattern because it is currently living a version of it. Ambient voice technology is being rolled out across trusts on the strength of pilot results and vendor assurance, with the MHRA's new National Commission still working out how agentic AI systems should be regulated and who is accountable when they fail. The Federated Data Platform, developed in collaboration with Palantir, was marketed on a similar premise: rather than requiring proof beforehand, the infrastructure and administration would develop in tandem with adoption. In both cases the technology may well work. The question is whether the underlying capacity, in compute, integration, and clinical governance, exists at the scale being promised, or whether it exists mainly in the same rhetorical register as DataVita's energy parks.

The comparison matters because the NHS estate is not a greenfield site in Lanarkshire. It is a live system already operating close to failure in workforce and financial terms, with successive reviews from Ockenden to the Lampard Inquiry pointing to how thin margins for error already are. A datacenter that is unable to secure its power supply simply experiences delays, which is detrimental to investors and a government eager to demonstrate the effectiveness of development zones. An AI system deployed into maternity units or acute wards on the assumption that governance will catch up carries a different order of risk. It makes sense that Sir Jim Mackey's responsibility effort has concentrated on trust performance in relation to contracts. It has said less about whether the same discipline applies to vendors and national programmes whose claims about readiness are, at this stage, no more independently verified than DataVita's.

There is a version of this argument that overreaches, and it is worth resisting. NHS AI adoption is not simply Lanarkshire with stethoscopes. The comparison is structural rather than literal: what connects the two is not the technology but the sequencing, the tendency to let political timetables set the pace of infrastructure verification rather than the other way round. Where that tendency exists, the correction is the same in both cases. Before the rollout, rather than after, someone has to enquire as to what is truly constructed, what is only allowed, and what is merely an artist's rendition of a turbine.

Britain does not obviously have a shortage of ambition in artificial intelligence. It has a shortage of institutions willing to check the ambition against the concrete, the grid connection, the integration test, before the announcement goes out. Lanarkshire will eventually reveal, one way or another, whether that gap can be closed with political will alone. The NHS, given what is already at stake in it, cannot afford to wait for that answer before asking the same question of itself.