The British government's plan to become what it calls an AI superpower has a physical core, and it is measured in megawatts. In February the Department for Science, Innovation and Technology opened the call for AI Growth Zones, inviting local and regional authorities, data centre developers and energy companies to propose sites. The government says it will work with network operators to scale each zone rapidly to 500 MW or more, which it describes as enough to power roughly two million homes. The first zone has already been named: Culham in Oxfordshire, home of the UK Atomic Energy Authority.
The zones are framed as a planning and investment measure. The criteria make clear that they are mainly about electricity.
What the government is asking for
The expression of interest asks for sites with large existing power connections, with a current capacity of 500 MW or more, or a clear plan for how energy capacity can be increased. It also favours locations close to suitable sites for major energy infrastructure, such as nuclear reactors, solar and wind farms, or battery storage. The government says it will put particular focus on deindustrialised areas, including sites with existing access to power or those suitable for new energy infrastructure.
Developers and energy firms are being asked to set out timelines and milestones showing how they will ramp up energy capacity, partnership opportunities with local authorities, and what additional support they need from government. The proposals will inform final site selection and broader policy decisions later this year. The government says interest is already building in Scotland, Wales, the North East and the North West.
Inside a zone, the promise is faster planning permission for AI infrastructure, including data centres, and the energy connections needed to power them.
The action plan behind it
The zones come from the AI Opportunities Action Plan, which the Prime Minister launched on 13 January. The government's announcement listed several commitments with energy implications. It pledged to increase public compute capacity twentyfold by 2030, starting with a new supercomputer. It promised a dedicated AI Energy Council, chaired by the Science and Energy Secretaries, to work with energy companies on the demands and challenges of powering AI, with reference to technologies such as small modular reactors. And it named Culham as the first AI Growth Zone, which the government says will also serve as a testing ground for research on how sustainable energy such as fusion can power AI.
The government's economic case is large. It cites an IMF estimate that, if AI is fully embraced, it could raise productivity by as much as 1.5 percentage points a year, gains it says could be worth up to an average of £47 billion a year to the UK over a decade.
The launch came with investment pledges. Vantage Data Centres, Nscale and Kyndryl committed £14 billion and 13,250 jobs, on top of £25 billion in AI investment announced at the International Investment Summit. Vantage plans to invest more than £12 billion in UK data centres, including one of Europe's largest data centre campuses in Wales. Nscale announced a $2.5 billion investment over three years and a contract to build what the government called the largest UK sovereign AI data centre, in Loughton, Essex, by 2026.
Why power is the constraint
The 500 MW threshold is the revealing number. A single site drawing that much continuously would use about 4.4 TWh of electricity in a year, since 500 MW times 8,760 hours is 4,380 GWh. That is the scale of a large power station's output, concentrated at one location. Few sites in Britain have spare grid capacity of that size, and building new connections at transmission level takes years.
That is why the criteria favour sites with existing large connections. Former industrial sites, closed power stations and steelworks often have high-voltage infrastructure that can be repurposed. One regional leader quoted in the government release made the point directly, saying the region has the land, the power and experience moving large projects at Teesworks. The deindustrialised areas the government wants to target are, in many cases, the places where grid infrastructure already exists because heavy industry once used it.
The proximity criterion points the same way. Sites near planned nuclear, wind, solar or storage can, in principle, be supplied without long new transmission lines. Where generation already outstrips local demand, as with wind in parts of Scotland, placing large loads nearby could make better use of existing output as well as shortening connection times.
The questions it raises
The plan leaves several issues open. Connection queues in Great Britain are long, and the government has not said how AI Growth Zones will be prioritised relative to other demand or generation projects waiting for a connection. Giving data centres faster access could push other projects further back.
There is also the question of what power the zones will use. The government links the zones to its clean energy mission and to technologies such as small modular reactors and fusion. Those are long-term options. In the near term, a 500 MW site will draw on whatever mix the grid supplies at the time, including gas-fired generation when wind and solar are low.
Costs are a third issue. Network reinforcement for large new loads is paid for through a mix of connection charges and network tariffs. If zones require substantial new grid investment, how that cost is shared between data centre operators and other bill payers will become a political question, as it already has in parts of the United States and in Ireland.
What to watch
The first test is site selection later this year. The number of zones, their locations, and whether they cluster near existing connections or near new generation will show how seriously the government is treating the energy side. The second is the AI Energy Council's work, which could set the terms for how data centres are connected and supplied.
The third is Culham. A data centre at a fusion research campus is a striking symbol, but fusion will not power it this decade. How the first AI Growth Zone is actually supplied, and how fast, will show whether the policy can deliver megawatts as quickly as it has delivered announcements.