Can Britain actually power its AI growth zones?

TL;DR:

• The UK has designated five AI growth zones — areas where the government backs datacentre complexes of 500MW or more — but the plans hinge on power the grid may not be able to supply.
• A Capgemini survey finds 77% of senior energy executives fear AI demand will soon outpace their ability to deliver it.
• “Phantom” power requests, up to a fifth of which never materialise, are distorting forecasting and clogging the connection queue.

AI growth zones are the government’s attempt to marry two goals: reviving former industrial areas and keeping Britain competitive in AI. Applicants had to show they could build 500MW-plus datacentres by 2030, with a realistic route to powering them. On the evidence of the first announcements, that condition is where the policy strains.

The bottleneck is electricity, not ambition

The two most scrutinised sites illustrate the problem. Lanarkshire’s developer described plans amounting to building the UK’s largest onshore windfarm in four years, then quietly conceded the site would connect to the grid. North Tyneside’s Stargate UK, from which OpenAI withdrew, appeared to lack a grid connection entirely. In both cases the government’s own criteria — an allotted connection or a “behind-the-meter” alternative — look to have been waved through.

The national picture is starker. A Capgemini report, AI meets the grid, surveyed 600 senior industry executives: 77% worry demand will soon outstrip supply, 80% expect more volatile demand, and 68% foresee regular shortages. Two-thirds flagged “phantom demand” — developers lodging power requests for sites that never get built, 19% of which evaporate. Britain’s grid is a particularly acute case, with electricity costlier than anywhere else in Europe and connection waits of eight to 10 years.

That queue is where the political trade-off bites. Expediting a hyperscale datacentre means potentially bumping homes and hospitals down the list — a decision growth zones were never designed to force. The same pressure is driving deals such as the National Grid’s £1.4bn bet on AI-power platform Joulent.

Looking forward

There is a flip side: 60% of executives believe AI-driven analytics could cut grid failures and outages by around 10%, and utilities are already using it for optimisation. The technology straining the grid may also help run it. But that is a medium-term prize. In the near term, whether the UK can physically deliver the power its AI strategy assumes — not investor appetite or planning consent — is the constraint that decides which growth zones become real.