Britain’s AI infrastructure policy is moving from ambition to a more prosaic problem: who gets to hold a place in the electricity queue.
On 29 July, Ofgem opened a consultation on “Curate”, a proposed set of connection reforms for data centres. Its starting point is unusually stark. Ofgem says contracted demand in the queue rose from 41GW in November 2024 to 125GW in June 2025, with data centres accounting for around 73GW; Great Britain’s 2025 peak demand was about 45GW.1
The regulator’s answer is not a moratorium. It proposes that data-centre projects above 40MW secure a returnable financial commitment after accepting a connection offer, and demonstrate progress at defined milestones. The stated purpose is to make it harder for projects without a credible route to build to occupy scarce capacity, while allowing viable projects to advance.1
That is a real change in the terms of competition for grid access. It is also a bounded one. The proposals remain open for consultation until 16 September, with decisions expected later in 2026. They can improve the queue’s filtering mechanism; they do not, on their own, answer the separate questions of generation, network build-out, local planning, water use or public consent.12
Ofgem is proposing a test of seriousness for large projects
Ofgem’s consultation proposes two linked measures for in-scope data centres. First, projects above 40MW would have to secure a returnable commitment of 2.5% to 7.5% of average data-centre capital expenditure from acceptance of a connection offer until energisation. Ofgem estimates that at roughly £237,500 to £712,500 per MW requested; it says the security would be returned if the project proceeds as required and forfeited if it terminates or fails the relevant rules.1
Second, developers would have to show evidence at specified stages that a project is progressing. The consultation lists a credible compute customer or end-user, procurement of long-lead electrical equipment, financial capability and technical readiness among the proposed requirements. A project that does not meet the milestones could lose its queue position.1
These are proposals, not a finding that every project now in the queue is speculative. Ofgem explicitly says Curate’s aim is not to reduce the queue to a pre-set volume. Its stated aim is to deter non-viable applications while imposing the lowest feasible burden on genuine developers.1
The proposal fits a government strategy to reserve capacity for selected AI projects
The consultation sits inside a wider government programme to accelerate AI data centres. The government’s November 2025 AI Growth Zones paper identifies slow planning and access to power as the central barriers to new domestic capacity. It says it intends to remove speculative demand, reallocate released capacity and reserve future capacity for AI Growth Zones and other projects it identifies as strategically important.2
That creates a documented institutional connection. Ofgem’s proposed filtering rules address the regulator’s concern that non-viable queue positions can distort investment signals and delay credible projects. Government policy envisages using released or reserved capacity to support designated strategic projects.12
The connection is not proof that a particular commercial developer will lose out to an AI Growth Zone, nor that prioritisation is inherently fair or efficient. It describes the mechanism the two public documents set out: clearer evidence of project viability, alongside a government plan to allocate capacity by strategic priority.12
A cleaner queue would not remove the underlying political choices
The strongest interpretation of Curate is modest: it attempts to turn a cheap option on grid capacity into a more accountable commitment. If a developer must post security and show an end-user, equipment procurement and readiness, holding a position without a credible project should become less attractive. That is an inference from the consultation’s design, not a measured result.1
It does not follow that the reforms will make every remaining project socially acceptable or environmentally benign. The public argument now includes whether Britain should keep expanding AI data centres at all. The Guardian reports that the Green party has called for a pause on major projects, while the government has rejected a moratorium and says data centres do not automatically receive priority access to electricity or water.3
Those claims are about different decisions. Queue management concerns the credibility of a connection request. Planning consent, environmental effects, water allocation, affordability, location and the pace of physical grid investment remain separate processes. Conflating them would overstate what a connection reform can accomplish.123
Better queue discipline may be necessary, but it may not be sufficient
The testable hypothesis is that a financial commitment and verifiable milestones will improve the quality of data-centre connection requests, releasing capacity that can be used by projects able to build. That is the mechanism Ofgem is consulting on, not an outcome it has demonstrated.1
It could fail if the proposed cost deters viable projects more than non-viable ones, if milestones are poorly calibrated, if constrained network works remain the main delay, or if released capacity is not allocated and built out promptly. The consultation itself does not resolve those implementation questions.1
The test is whether the rule releases capacity without simply raising barriers
The near-term evidence will be concrete. Ofgem’s consultation closes on 16 September 2026 and it expects to make decisions later in the year. Its final decision should show whether the 40MW threshold, the proposed security range and the milestones survive consultation, and how existing projects are treated.1
The broader hypothesis is that a well-designed commitment rule will release capacity from projects that cannot demonstrate a path to energisation while allowing credible projects to proceed sooner. It would be supported by published data on queue exits, the capacity released, connection dates and the distribution of outcomes across developers. It would be weakened if viable projects are disproportionately deterred, if the queue remains congested without meaningful releases, or if capacity is not translated into actual connections.1
None of that outcome is established yet. The available evidence documents a proposed rule and the policy problem it seeks to address, not its performance.
What we should watch next
- 01Ofgem’s final Curate decision after the 16 September 2026 consultation deadline
- 02Government rules on capacity reallocation and reservation for AI Growth Zones
- 03Data on data-centre queue positions, exits, released capacity and connection dates
- 04Code, licence and contractual changes needed to implement Curate
- 05Planning and environmental decisions for individual large data-centre projects
Ofgem is proposing to make a place in the grid queue cost something.
Ofgem’s Curate consultation is a targeted response to a specific bottleneck in Britain’s AI infrastructure plans: data centres account for a large share of contracted demand in a grid-connection queue larger than national peak demand. Its proposed commitment fee and delivery milestones would make developers show both financial seriousness and progress. That is a meaningful change in the governance of scarce capacity, and it aligns with government plans to clear speculative claims and direct capacity toward strategic projects. But it remains a proposal. It cannot substitute for new network capacity, generation, transparent allocation rules, local planning or a settled public answer on the acceptable scale and location of data-centre growth.
The next question is not whether the queue is long, but whether a tougher place-holder rule makes the power system move faster and more fairly.Sources
Threadmap News is an AI-assisted publication operated through a fictional editorial team. Daniel Mercer is the publication’s editorial byline. Named desk members represent specialist research roles within the Threadmap system; they are not separate human contributors or outside authorities.
The latest local technology scout passed its structural validator but was rejected for this edition after editorial review: its core assertions relied mainly on one publisher, and it attached an unrelated Kyiv item to the infrastructure thesis. Fresh source-page research was conducted across three lanes (policy, courts/technology and climate/infrastructure); this Ofgem consultation was selected because it provides a current, primary regulatory record with a bounded mechanism. The Ofgem consultation, the UK government’s AI Growth Zones paper and The Guardian’s independent report were individually checked. Facts are limited to what those pages establish; the article treats the likely effect on queue quality as a hypothesis rather than as an achieved result.