What Ofgem published

On 29 July 2026 Ofgem opened a consultation titled Curate: Demand Connections Reform, proposing two measures aimed, in its own words, at "deterring speculative and non-viable projects and facilitating viable developments". The document's executive summary states the position it is responding to: "Between November 2024 and June 2025, contracted demand in the connection queue increased from 41 GW to 125 GW. Of this, around 73 GW are data centres. For a sense of scale, peak demand in Great Britain in 2025 was 45 GW."

Paragraph 2.7 breaks the movement down by network level. The 41 GW was 17 GW transmission and 24 GW distribution. The 125 GW was 97 GW transmission and 29 GW distribution. Almost the whole increase landed on the transmission network.

The remainder of the queue, Ofgem states, is composed mainly of hydrogen electrolysers, industrial demand and transport. It also records that it is working with the National Energy System Operator and the network companies to collect this data regularly, which is a statement that the present understanding of the complete demand queue is incomplete.

The two bases

Ofgem's press release of 29 July 2026 describes the same movement and attributes it differently. It says demand connection applications "have surged from 41 gigawatts (GW) to 125GW", and that the rise "is driven largely by data centre projects, accounting for at least 80GWs". Later in the same release it refers to "around 73GW referenced in Ofgem's consultation".

Both figures are Ofgem's and neither is wrong on its own base. The consultation's 73 GW is a share of the 125 GW total queue. The press release's 80 GW is attributed to the increase, which is 84 GW. A reader who carries one figure into a sentence built for the other has changed the claim: 73 GW is 58 per cent of the total queue, while 80 GW is 95 per cent of the rise. Those are different statements about how concentrated the growth is.

This is the same failure mode this publication documented in Australia's data centre forecasts, where one AEMO projection is published as almost 10 per cent of underlying demand and as 12 per cent of grid-supplied consumption. See Two Denominators, One Data Centre Forecast. The pattern is not that a regulator has erred. It is that a single quantity acquires more than one published share, and the share travels without its base.

It travels quickly. Two legal briefings published in August 2026, from Slaughter and May and from Addleshaw Goddard, each read the consultation directly and each restate the 73 GW as data centres' share of the increase rather than of the total. The consultation does not say that. Anyone quoting a data centre share of Britain's connection queue should name the document and the base it was taken from.

A queue is not a forecast

The reason the base matters is that neither figure describes demand that will arrive. A connection queue counts contracted offers, and Ofgem is consulting precisely because it holds that a material part of the queue will not be built. The consultation puts the point in terms most forecasters would hedge: it argues that even if the proportion of non-viable and speculative applications in this sector were lower than in others, "which there is no evidence to suggest is the case", the impacts would still be considerable.

Two of Ofgem's own comparisons make the scale legible. Peak demand in Great Britain in 2025 was 45 GW, against a queue of 125 GW. And the data centre pipeline alone represents around £693 billion of implied total capital expenditure, which Ofgem puts at around 23 per cent of UK GDP in 2025, on an assumed average of £9.5 million per MW.

Neither comparison proves any individual project unviable. Together they establish the thing a queue figure cannot establish on its own: that the number is an upper bound on interest, not a projection of load. That distinction is what a fee is being designed to enforce.

Where the capacity sits

Table 1 of the consultation distributes the 73 GW across five size bands. The distribution is the reason the proposed fee has a threshold rather than applying to every applicant.

Project counts and total MW are Ofgem's, as are the capacity shares. The share-of-projects column is computed here from Ofgem's counts over its stated total of 315, and the five counts sum to 315 and the five MW figures to 72,978, so the table reconciles to the headline 73 GW.

Ofgem states that it arrived at its proposed 40 MW threshold by examining this distribution and finding that 99 per cent of the total capacity in the queue sits in projects above 40 MW. The forty hyper-scale projects carry 43.0 per cent of the capacity on 12.7 per cent of the count.

Nine gigawatts changed technology

One paragraph of the consultation records a movement that is not visible in any total.

Between May 2024 and August 2025, on Ofgem's analysis, at least 9 GW of capacity already holding a transmission connection request modified that request from a battery technology to a data centre. Ofgem records that stakeholders point to an incentive structure that makes battery and other technologies increasingly likely to convert at the next application window.

This matters beyond Britain because it is a measurement problem before it is a policy one. A queue position that was counted as storage is now counted as demand, without a new application and without anything being built. Any series that reads a connection queue as a pipeline of storage is exposed to reclassification of this kind, and the reclassification is only visible where a regulator publishes it.

What is proposed, and when it is decided

Curate proposes two measures. The first is a data centre commitment fee for projects above 40 MW: a returnable financial commitment secured from the point a connection offer is accepted until energisation, proposed at between 2.5 and 7.5 per cent of average data centre capital expenditure, which Ofgem states as approximately £237,500 to £712,500 per MW of requested capacity. On Ofgem's worked example, a 100 MW project would secure between £23.8 million and £71.3 million at the point of accepting an offer. The commitment is returned if the project proceeds and forfeited if it terminates or fails to comply.

The second is a set of data centre queue management milestones, requiring developers to evidence progress, including financial capability, commercial maturity and procurement activity, to retain a queue position.

Responses are due by 16 September 2026. Ofgem states it will publish non-confidential responses alongside a decision on next steps. Until that decision, nothing here is settled policy.

Limits of this reading

• The consultation is open. No decision, no final fee value and no confirmed threshold exists at the date of publication.
• Every figure is Ofgem's own analysis of a queue that changes continuously. The 41 GW and 125 GW readings are dated November 2024 and June 2025 and were already more than a year old when the consultation opened.
• Ofgem states it is still working with the National Energy System Operator and network companies to improve its understanding of the complete demand queue, so the non-data-centre remainder is the least characterised part of these totals.
• Consultation responses have not been read and are not characterised here.
• Ofgem's November 2025 "Demand connections update" was not used. Its published PDF carries no extractable text layer, and no figure in this report is drawn from it.

What to watch

Three things resolve this. The first is whether Ofgem's decision restates the queue on one base, or leaves a total-queue share and a share-of-the-increase in circulation as it does now. The second is the fee value set within the proposed range, since the range spans a factor of three and the deterrent effect is not linear in it. The third is the next published queue reading: the test of a measure designed to remove speculative applications is whether the queue falls, and by how much of the 73 GW.