- Data centre projects of 40 MW or more could be required to secure between £237,500 and £712,500 for every megawatt of requested grid capacity.
- The money or equivalent financial security would be returned at energisation but could be forfeited if a project withdraws or fails to meet new development milestones.
- Ofgem has identified around 73 GW of data centre applications across 315 projects, far exceeding Britain’s 45 GW peak electricity demand in 2025.
Ofgem has proposed making large data centre developers put up substantial financial security when accepting grid-connection offers, as it seeks to remove speculative projects from an electricity queue that has more than tripled in size.
Under the regulator’s consultation, projects of 40 MW or more would have to secure between £237,500 and £712,500 per megawatt from acceptance of their connection offer until energisation. A 100 MW development would therefore need to provide security worth between £23.75 million and £71.25 million.
The commitment would be returned if the project connects as planned. It could be forfeited if the developer terminates the project, reduces its requested capacity or fails to comply with the relevant rules.
Ofgem is also proposing data centre specific milestones requiring evidence of a credible computing customer, orders for long-lead equipment such as switchgear and sufficient financial and technical capability.
“Britain’s electricity demand connections queue has more than tripled in size in less than a year, and consumers should not bear the risks created by speculative projects taking up space in the system,” said Ofgem director for energy system design and development Eleanor Warburton.
“The connections system must work for consumers and for the projects that are ready to invest, build and connect. Where speculative projects take up space in the queue, they can delay other schemes and create uncertainty about future network needs. ”
Contracted electricity demand applications increased from 41 GW in November 2024 to 125 GW by June 2025. Ofgem’s formal consultation identifies approximately 73 GW of data centre capacity across 315 projects, compared with Britain’s peak demand of 45 GW last year.
The projects range from 1 MW to more than 1,500 MW. Extra-large and hyperscale developments account for more than 93% of the requested capacity, while 99% is associated with projects above the proposed 40 MW threshold.
Ofgem estimates that building every data centre in the queue would imply capital expenditure of around £693 billion, equivalent to approximately 23% of UK gross domestic product in 2025. That scale provides strong evidence that many applications will never be delivered.
Wheat from the chaff
The present system makes it comparatively inexpensive to obtain and retain a demand connection. Land can also increase sharply in value once a grid offer and planning permission have been secured, creating an incentive for developers to reserve capacity before establishing whether they have a customer or finance.
Ofgem’s preferred mechanism is designed to change that calculation without imposing a permanent charge on genuine investment. Its modelling suggests the most onerous version of the fee could reduce the unlevered return on a hypothetical project connecting in 2035 by 1.7%. The regulator argues that this should not make a viable development uneconomic, particularly if a healthier queue delivers an earlier connection.
The measures would apply to relevant existing and future projects connected at transmission level, alongside distribution-connected developments with a material transmission impact. Responses are due by 16 September, with policy decisions expected later this year.
The proposed fee inevitably favours developers with strong balance sheets. Hyperscalers and established infrastructure funds should be better able to provide tens of millions of pounds in security than smaller independent developers, even where the latter have a credible project.
A developer with ample capital can still warehouse grid capacity, but demonstrating an eventual end-user, equipment procurement and technical progress provides a clearer test of whether construction is genuinely approaching.
The reforms also do not create new physical network capacity. They should improve the quality of the information used by network planners and bring forward connections currently blocked by implausible applications, but transmission reinforcement and additional generation will still be required if even a fraction of the credible data centre pipeline materialises.
Ofgem is addressing the first part of Britain’s AI infrastructure problem: distinguishing real demand from paper demand. The more pressing question is how much data centre load the country should plan to accommodate, where it should be located and who should pay for the dependable power required to serve it.
