The UK’s race to support artificial intelligence infrastructure is running into a power problem that is partly real and partly speculative. Data center developers are crowding into the queue for grid connections, but many of those projects may never be built.
That has created a practical challenge for Ofgem, the UK’s energy regulator: clear out the phantom data centers slowing the system down without making the country too expensive for serious builders.
Why the Queue Became a Bottleneck
Data center developers are competing for a share of the hundreds of billions of dollars flowing into the artificial intelligence industry. To operate the compute needed for sophisticated AI models, they need large and reliable electricity connections.
In the UK, the queue to join the power grid has become crowded with proposals that may not turn into real facilities. The problem is not only that viable projects face already years-long waits. It is also that grid operators must plan around demand that may never materialize.
The surge accelerated toward the end of 2024, around the time the government designated data centers “critical national infrastructure.” According to Ofgem, between November 2024 and June 2025, the total energy demand of projects in the connection queue rose from 41 gigawatts to 125 gigawatts.
New data centers account for 73 gigawatts of that total, which the source describes as equivalent to one and a half times the peak demand for the entire UK last year. The government has said much of the demand appears to be a mirage.
What Ofgem Wants to Change
In July, Ofgem laid out a proposal designed to push speculative data center projects out of the queue. The plans are set to be finalized after an industry feedback process that ends in September.
Under the proposal, developers would face a set of tougher hurdles before holding a place in the grid connection process:
- A steep, nonrefundable deposit that could reach hundreds of millions of dollars for the very largest data centers.
- A requirement to line up customers before moving ahead.
- A requirement to prove they have the funding needed to complete their builds.
The UK has already required all developers since 2024 to show progress on key milestones, including securing land rights, submitting a planning application, and meeting other indicators that suggest they are likely to follow through.
The latest proposal follows the same basic logic: make it harder to reserve grid capacity without a credible plan. Alex Burgoyne, head of data centers at real estate consultancy Knight Frank, said the proposal is expected to reduce speculative applications: “It’s going to pour water on a lot of the speculative data center applications.”
Why Phantom Projects Slow Real Ones
The queue problem comes from incentives. When access to the grid takes years and joining the queue previously cost only a few thousand dollars, developers had a reason to apply even without a firm development plan.
That choice can function as a hedge on future demand for compute. If demand stays high, a place in the queue becomes valuable. If the project never happens, the developer has not risked much.
But the cost falls on the wider system. Grid operators have to assess the combined effect of large infrastructure proposals on network stability before approving individual connections. That means speculative applications can make the technical studies more difficult, more costly, and slower.
Olivier Darmouni, associate professor of finance at HEC Paris Business School, said grid operators must evaluate each project seriously: “They have to treat every project as serious when they try to study whether the system can handle them.” He added that speculative projects are damaging because they increase complexity, expense, and delay.
The queue can also attract middleman developers. According to the source, some apply for power with the aim of later selling land at a premium to data center companies, bundled with grid access. Darmouni compared that dynamic to “scalping for concert tickets.”
The Risk of Going Too Far
Ofgem’s challenge is a balancing act. Fees need to be high enough to deter speculative projects, but not so high that legitimate data center developers decide the UK is no longer worth the risk.
Industry experts already view the UK as a difficult market because of the high cost of energy and the limited supply of land. The US is described as a highly desirable market, while some corners of Europe can offer plentiful, renewable energy.
Daniel Newton, a partner at the law firm Slaughter and May who specializes in digital infrastructure, described the dilemma directly: “If you don’t set [the fee] high enough, you don’t deter speculation. If you set it too high, you risk killing viable projects, because people are unwilling to put large sums of money at risk at the development stage.”
The proposal could also reshape which companies can afford to build in the UK. Very large data center companies may be able to handle steep deposits, while smaller AI-specific operators could struggle. The source notes that some newer companies, including Nscale, have previously pledged to spend billions of dollars in the UK.
Burgoyne warned that the reforms may have broader market effects: “There are new players arriving all the time.” He added: “It might have the unintended consequence that it stifles growth in the sector.”
What Better Grid Planning Could Unlock
Despite those risks, removing phantom data centers could help the UK make more rational infrastructure decisions. If grid operators have a clearer picture of real demand, they can target upgrades and expansions more effectively.
Taco Engelaar, managing director at grid optimization company Neara, said uncertainty from phantom projects makes it hard to understand what real demand will be. He also said an accurate queue would help focus investment on the right grid upgrades and locations.
Ofgem says it will take industry concerns into account before settling on a fee. Nathan Macwhinnie, deputy director of strategic planning and connections at Ofgem, told WIRED that data centers are important to the UK’s AI ambitions and future economic growth, and that helping viable data centers connect faster supports that goal.
The reforms may not immediately make the UK a stronger magnet for AI infrastructure capital. But if they separate credible projects from placeholders, they could give the country a better foundation for deciding where data centers should be built, how large they should be, and what power supply they should rely on.