Britain's data centre gap is a sovereignty question, not just a grid one

The US, by contrast, could be operating over 90GW by the same point, up from roughly 29GW now, according to analysis in The Economist.

The immediate causes are familiar to anyone who has followed Britain’s infrastructure debates: industrial electricity prices running close to four times higher than in the US, and a grid connection process that has become the binding constraint on new builds. Analysts have noted that the mismatch is stark: construction timelines for a data centre are measured in months, while grid connections are running to the better part of a decade.

That bottleneck has been compounded by the UK’s first-come, first-served approach to grid queuing, which has left the system swamped with speculative applications from developers with no immediate intention to build. The national connection queue reportedly tripled in size, from 41GW to 125GW, in the seven months to June 2025. Government reforms are underway, including efforts to remove unserious applications and new AI growth zones offering priority grid access in designated areas, but much of this remains stuck in consultation.

Scotland illustrates how quickly that friction can turn political. The SNP’s national council passed a motion to freeze new data centre approvals, which has since been sent to the Scottish government for consideration. Scottish Greens figures put 24 proposed new sites at up to 6,000MW combined, over one and a half times Scotland’s current peak power use. Action to Protect Rural Scotland has argued that hyperscale projects should face mandatory environmental impact assessments and clearer guidance for local authorities, with its director, Kat Jones, crediting community campaigns with pushing the issue onto the political agenda. The dispute sits alongside a separate case in which a UK council was reportedly encouraged by central government to approve a data centre expansion despite local objections over water, power and design.

The more interesting argument, though, is not about grid capacity at all. It concerns whether Britain needs to build data centres in volume in the first place. Some analysts have argued that capacity and sovereignty are distinct problems, and that growing the number of AI companies operating in Britain does not depend on growing the number of data centres on British soil. On that view, Britain’s genuine leverage in the AI race comes less from server racks than from research strength and a services sector positioned to find high-value applications for AI. Under this logic, bulk compute could largely be outsourced to markets with cheaper power and more available land, such as Finland, while latency-sensitive uses, including financial trading and autonomous vehicles, remain domestic for technical rather than sovereignty reasons.

Underlying both the Scottish freeze motion and the UK’s low public awareness is a communications problem the industry has been candid about internally. At last year’s Datacloud Global Congress, panellists from across the sector agreed that data centre operators have failed to explain their social value in terms the public understands. Gary Connolly of Digital Infrastructure Ireland put it bluntly: nobody cares about PUE, or “please use English” as he called it.

CyrusOne’s Emma Fryer cited research across seven European countries showing that while the large majority of respondents held neutral or positive views of data centres, barely half could correctly define what one was, with the UK recording the lowest comprehension of the countries surveyed despite the size of its market. Rowan Digital Infrastructure’s Martin Romo argued the industry needed to meet residents’ direct questions about water and pollution with direct answers rather than deflection, a gap that echoes what Capacity has found reporting from the ground in the US.

That US contrast is instructive. Capacity’s own reporting from Central Texas found data centre water use in the state could climb from around 0.4% of total consumption today to as much as 2.7% by 2030, though researchers at the Houston Advanced Research Center caution that even that projection is likely already outdated given the pace of new development.

Dr Margaret Cook of HARC has argued that the state-wide figure is the wrong place to look for risk in any case, since the impacts are really local, concentrated in counties surrounding Austin with little water planning or oversight. Kamil Cook of Public Citizen, who works directly with affected communities, describes the scale of local opposition as unlike anything else he has encountered in three years working the issue, driven by drought conditions, a sense of local powerlessness over permitting, and projects that often proceed under NDAs and code names until after land or water rights are already granted.

Ferveret’s Reza Azizian, whose company builds liquid cooling technology aimed at reducing that footprint, points to Virginia as a cautionary case, where he says household electricity prices have nearly doubled amid rising data centre demand.

Regulators and city governments are beginning to respond to this pattern rather than wait for it to harden further. The Global Urban Data Centres Pact, launched at London Climate Action Week with mayors from cities including London, Mumbai, Melbourne and Nairobi, commits signatories to integrating data centres into urban planning, minimising their environmental footprint and reinvesting surplus revenue locally.

C40 Cities, which backs the pact, notes that the global data centre count has grown from around 500,000 in 2021 to roughly eight million today, with the sector now responsible for an estimated 2.5% to 3.7% of global emissions. Melbourne’s Lord Mayor, Nicholas Reece, framed the pact’s premise plainly: local communities should be involved in decisions that affect them.

For operators and investors weighing UK versus US expansion, the picture that emerges is one of a narrowing window rather than a durable advantage. Britain’s grid queue remains the more immediate constraint on deployment timelines in the near term. But the Texas and Scotland cases both show how fast sentiment can shift once communities connect data centres to tangible local costs, and the industry’s own panellists at Datacloud have acknowledged that better public communication, not just faster permitting, will determine how much of that friction is avoidable.

Developers and policymakers who treat the UK’s current low-awareness period as permanent, rather than temporary, risk being caught out in the way that projects in Central Texas and Scotland already have been.

The underlying question for Britain is less whether it can match American data centre capacity than whether it needs to. If the sovereignty argument holds, and AI leverage can be built through research and services rather than server density, then Britain’s grid constraints may matter less for national competitiveness than for the commercial fortunes of the data centre sector itself. That distinction, alongside the growing willingness of both local campaigners and city governments to push back on unplanned expansion, is likely to shape how much new capacity gets built, and where.

What is sovereign AI infrastructure, and why does it matter for telecoms?

Global Urban Data Centres Pact launched to improve industry sustainability