A queue built on intentions as well as projects

The phrase “phantom data centre” does not mean that electricity demand has disappeared. It describes a practical planning problem: applications for large grid connections can be made long before a project has secured a credible customer, financing, equipment, planning consent or a final commitment to build. Some applications will mature into operating facilities; others may never do so. Yet both can occupy positions in a constrained connections process.

That distinction has become significant in Great Britain, where demand for large electricity connections has risen rapidly alongside the push to expand cloud computing and artificial intelligence infrastructure. Ofgem’s analysis says total contracted demand offers in the queue rose from 41 GW in November 2024 to 125 GW by June 2025. It estimates that data centres account for roughly 73 GW across about 315 projects. These figures represent requested or contracted capacity, rather than a forecast that every listed facility will be built and run at full load.

The scale matters because network companies and the National Energy System Operator must use connection requests to plan substations, transmission lines and other reinforcements. A queue dominated by projects with uncertain prospects can create a misleading impression of future demand. That risks either delaying credible schemes while capacity is reserved, or encouraging investment plans based on demand that later fails to materialise. In both cases, consumers may ultimately face unnecessary costs through network charges.

Why data centres are at the centre of the issue

Data centres are not unusual in needing large amounts of power, but their development model can make the connections process especially difficult. A developer may seek land and grid capacity before signing a computing customer, while a prospective customer may hesitate to commit until the site has a firm power date. The result is a circular problem: obtaining a connection can be necessary to prove a project’s commercial case, but regulators want commercial evidence before allowing a project to retain scarce capacity.

The sector’s diversity adds another complication. Some facilities are built and operated by their eventual users. Others are developed for lease or sale to cloud providers, colocation operators or enterprise customers. A regulatory test that requires a fully signed end-user contract too early could exclude viable projects using the latter model. Conversely, a system with few meaningful checks can reward those that apply early and widely, even if they are not best placed to deliver.

The challenge is therefore not simply to reduce the number of applications. It is to differentiate credible future demand from speculative reservation without entrenching the largest incumbents. New entrants and regional developers may have less capital available at an early stage than the biggest technology groups, despite having viable proposals.

Ofgem’s proposed filter

On 29 July 2026, Ofgem opened a consultation on a package intended to “curate” the data-centre queue. The central proposal is a returnable commitment fee for projects above 40 MW. Developers would have to secure the financial commitment after accepting a connection offer and keep it in place until energisation. It would be returned if the project progresses under the rules, but could be forfeited if the project terminates, reduces capacity or fails the applicable requirements.

Ofgem is consulting on a fee between £237,500 and £712,500 per MW of requested capacity, equivalent to 2.5% to 7.5% of estimated average project capital expenditure. At that range, a 100 MW project would need to secure £23.8 million to £71.3 million. The sum is deliberately large enough to make casual capacity reservation unattractive, but it is not a payment for electricity or a final regulatory decision. The consultation remains open until 16 September 2026.

The regulator also proposes sector-specific milestones. Developers would need to provide evidence of a credible computing customer or end-user, procurement of long-lead electrical equipment such as switchgear, and the financial and technical ability to deliver the facility. Ofgem envisages separate evidence pathways for self-operated sites and facilities intended for lease or sale.

These measures reflect an important regulatory judgement: queue access has economic value even before physical power flows. The fee and milestones are meant to make developers demonstrate that they are prepared to bear a portion of the cost of holding that scarce option.

Clearing the queue will not create power by itself

Removing non-viable applications could produce more realistic network plans and allow ready projects to advance. It cannot, however, substitute for building infrastructure. Grid constraints arise from the time needed to construct transmission assets, upgrade local networks, secure equipment and coordinate new electricity supply. Ofgem itself identifies physical network and generation build as a major source of delay for well-developed demand projects.

That is why the proposed reforms sit alongside a broader programme described as Curate, Plan and Connect. Curate concerns the composition of the queue. Plan concerns where demand should be encouraged and which projects may be treated as strategically important. Connect concerns the practical arrangements for building and operating new connections, including potential self-build of high-voltage assets and more flexible connection terms.

Flexible connections could become particularly relevant for large computing sites. The government is considering whether very large users, including data centres, should provide a minimum degree of demand flexibility, allowing usage to be curtailed or limited during periods of network stress. Such arrangements could permit earlier access in some locations and make better use of existing assets. But they also raise operational questions. Data-centre customers often expect very high availability, so flexibility needs clear contractual limits, adequate backup arrangements and a realistic commercial value.

From first come, first served to strategic allocation

The government’s parallel proposals go further than queue management. They would allow capacity released by departing projects to be reallocated to government-identified strategic demand, reserve some future capacity for such projects and give them priority in future transmission queue-design exercises. AI Growth Zones are among the potential beneficiaries.

This approach offers a route to align digital infrastructure with regional network plans rather than allowing projects to cluster solely where land, fibre links or commercial demand appear strongest. In principle, guiding development towards areas with available capacity could lower reinforcement costs and spread investment more widely.

It also changes the nature of the connections system. A purely chronological queue treats an accepted application as the key entitlement. A strategically aligned system gives government a stronger role in deciding which forms of demand should move first. That may be justified where electricity infrastructure is scarce and national objectives include industrial growth, decarbonisation and digital capacity. It also requires transparent criteria. Manufacturers, housing-related infrastructure, electric-vehicle charging, hospitals and other large users may reasonably ask how their claims will be assessed alongside data centres.

The test is credible growth, not a smaller queue

The UK’s phantom data-centre problem is fundamentally a problem of uncertainty meeting a slow-to-expand network. The proposed reforms acknowledge that a long list of nominal projects is not the same as a reliable investment pipeline. They seek to improve the signal that the queue sends to planners, network companies and investors.

Success should not be measured only by how many applications exit. A healthy outcome would be one in which serious projects receive clearer and earlier connection paths, network investment is based on more dependable evidence, and the rules do not turn access to electricity into a privilege available only to the best-capitalised developers.

The consultation offers the first detailed test of that balance. Its eventual design will shape not only the pace of British data-centre expansion, but also who bears the risk when an increasingly digital economy asks the grid to plan for demand that may still be only a business plan.

Sources