Joining the queue cost a few thousand dollars, the wait was years long, and holding the slot cost nothing more, so developers held it whether or not they had a plan [4]. The word doing the work in Ofgem's proposal is nonrefundable [2]. A large refundable deposit is a treasury problem, solvable by anyone with a credit line. A nonrefundable one is an expense booked against a named site, and coupled with the demand for customers and funding up front [3], it requires the commercial case to exist before the connection does.
The size of the repricing is roughly five orders of magnitude at the top end, from a few thousand dollars to potentially hundreds of millions [4]. Nothing about a developer's queue behaviour survives that unchanged.
There is already evidence about which gates work. Since 2024, applicants have had to show secured land rights, a submitted planning application, and other milestones [8]. Those tests were in force while the queue's total demand climbed from 41GW to 125GW, an increase of 84GW in about seven months [5][1]. Paperwork you can buy from a planning consultant does not deter a free hedge.
The bill for the phantom entries lands on other people's schedules. Because operators must study the combined effect of large proposals on network stability before granting any single connection, each junk application makes the studies for viable projects longer, costlier and more complex, according to Olivier Darmouni of HEC Paris [9]. He also describes the secondary market the wait created: middlemen apply for power in order to flip land at a premium with a connection attached, which he likens to ticket scalping [10]. Neara's Taco Engelaar says nobody really understands what the real grid demand is because of the phantom projects [17]. WIRED reports the jam is both stretching wait times for viable schemes and fouling demand forecasting and expansion planning [18].
The arithmetic shows what forecasters are working with. New data centres account for 73GW of the 125GW queue, about 58 percent [5][6][2]. Ofgem's own comparison, that 73GW is one and a half times last year's UK peak, implies a national peak of roughly 49GW [6][3], which makes the queue about 2.6 times everything the country actually drew at its highest [5].
What is unresolved is the price. Ofgem's Nathan Macwhinnie says connecting viable data centres more quickly is an enabler of the UK's AI ambitions [14], and the regulator says it will weigh industry concerns before settling on a fee, while noting that energy markets do not compare like for like across borders [15]. Industry's read, per WIRED, is that the reforms risk making a country already hampered by expensive power and scarce land one of the world's costliest places to build [12]; Knight Frank's Alex Burgoyne warns against shooting the golden goose [13]. One rule, two outcomes, separated by a number nobody has written down yet. And the proof-of-customer test cuts wider than the scalpers it is aimed at: a developer building capacity ahead of a signed tenant fails it on the same terms [3].
