Summary
- FOI-derived data shows relevant permissions falling in South Oxfordshire after Culham’s AI Growth Zone designation, while North Tyneside recorded an increase.
- The figures do not prove the policy has failed because several zones are new and planning data captures only part of the development pipeline.
- Power, occupier commitments, planning capacity, construction, and policy certainty still have to align before AI Growth Zone status converts into operating capacity.
The UK’s AI Growth Zones have yet to produce a consistent acceleration in relevant planning activity, according to newly analysed local-authority data that shows sharply different development patterns across designated areas.
Freedom of Information data analysed by Search Acumen shows just one relevant planning permission granted in South Oxfordshire during 2025, the year Culham became the UK’s first AI Growth Zone. That compares with five in 2024 and eight in 2023.
North Tyneside moved in the opposite direction, with ten relevant permissions in 2025 compared with four in 2024. However, several of the projects were already moving through the system before Blyth and Cobalt Park received AI Growth Zone designation.
The figures should not be read as a direct count of new operating data centres. Planning datasets capture only one part of the development process, while projects may spend substantial periods securing land, power, finance, and environmental work before a full application appears.
The figures are an imperfect measure of programme performance, but that is precisely why they are useful. They show how little a policy designation tells the market about what is actually ready to build.
AI Growth Zones were created to concentrate support for large AI infrastructure developments in places where land, electricity, and planning could be coordinated more effectively. Government guidance says applicants should demonstrate access to at least 500MW by 2030, a minimum 100 acres of available land, and a credible route to full planning consent.
Five zones have been announced covering Culham, Blyth and Cobalt Park in the North East, Anglesey and Trawsfynydd in North Wales, Bridgend in South Wales, and Lanarkshire in Scotland.
Designation is only the first stage. Developers still have to control a site, secure or progress a power connection, design the facility, complete environmental work, fund planning applications, attract customers, and commit construction capital.
That makes planning statistics a lagging and incomplete indicator. A developer can spend months or years on grid and commercial work before a full application appears. Equally, a large number of minor permissions inside a zone does not necessarily mean substantial AI data centre capacity is being delivered.
DataCentral has already reported on governance uncertainty surrounding the programme, while changes to England’s planning framework have since strengthened national support for data centre development.
The new figures add a delivery test to that policy discussion. Culham was selected partly because the end of the Joint European Torus fusion programme made substantial electrical infrastructure available. If a site with an unusually strong power proposition still takes time to generate applications, grid access alone is clearly not sufficient to create a development market overnight.
There are signs of activity elsewhere. Nebius has agreed to deploy AI infrastructure at Vantage Data Centers’ Newport campus, providing a commercial capacity commitment within the South Wales Growth Zone. QTS is also progressing its large Cambois development within the North East zone.
Those examples reinforce the need to distinguish between policy-labelled investment and projects that were already viable for independent commercial reasons. A zone can improve a development environment without being the original cause of every scheme inside it.
The same distinction will matter as government assesses whether to designate further locations. If the objective is faster delivery, success should ultimately be measured in power secured, planning permissions granted, financing committed, construction started, and IT capacity brought into operation — not simply the number of regions carrying an AI Growth Zone label.
Search Acumen has also called for more standardised planning datasets, arguing that reliance on FOI requests exposes how difficult it remains to track the programme consistently.
That is a practical weakness for developers and investors as well as policymakers. Large infrastructure projects are priced against time and risk. If market participants cannot see what is entering planning, which grid connections are credible, or how authorities are handling applications, they have less information with which to compare one zone against another.
The latest data does not establish that AI Growth Zones have failed. Several are too new for that conclusion. It does show that designation has not yet produced a simple, uniform jump in visible planning activity — and that the difficult work still sits between a policy announcement and an energised data hall.
