UK Grid Queue Hit 125 GW as Phantom Data Centers Pile In

Between November 2024 and June 2025, the total energy demand of projects sitting in the UK grid connection queue jumped from 41 gigawatts to 125 gigawatts. New data center applications account for 73 gigawatts of that figure — equivalent to one and a half times the peak demand for the entire UK last year. Grid operators must model every application as credible before granting any individual connection, compounding study complexity and cost even when the majority of those projects will never break ground. For ML infrastructure teams tracking where large-scale AI training capacity can realistically be sited over the next five years, the UK's grid queue dysfunction is a direct upstream constraint on available compute.

How the Queue Became a Speculative Instrument

The queue ballooned for a structurally predictable reason: asymmetric incentives. Joining the UK grid connection queue previously cost only a few thousand dollars, while the wait for access stretched to years. That combination turned a grid position into an option with essentially no downside for developers hedging against sustained AI compute demand. Some applicants had no intention of building — they were acquiring grid-connected land positions to flip to larger operators at a premium, with grid access packaged in. As Olivier Darmouni, associate professor of finance at HEC Paris Business School, put it, the dynamic resembles concert ticket scalping: congestion attracts scalpers, whose presence deepens congestion. Grid operators cannot discount speculative applications when assessing network stability — they must treat each one as a live load forecast, making demand projection more expensive and less reliable for genuine infrastructure planning.

The government designated data centers "critical national infrastructure" near the end of 2024, and the queue began its rapid expansion roughly at that same inflection point. A 2024 milestone requirement — developers must demonstrate secured land rights and a submitted planning application — provided some filtration but was insufficient to halt the speculative surge.

Ofgem's Proposed Reform Mechanics

Ofgem's July 2026 proposal introduces three layered filters: a steep nonrefundable deposit, potentially reaching hundreds of millions of dollars for the largest projects; a requirement that developers demonstrate signed customer commitments before queue entry; and proof of construction financing sufficient to complete the build. The industry feedback window closes in September, after which Ofgem will finalize the fee structure.

The mechanism is straightforward in theory: raise the cost of holding a speculative position until the expected value of queue-flipping turns negative. Taco Engelaar, managing director at grid optimization firm Neara, noted that an accurate queue would allow grid operators to target investment in upgrades and expansions with far greater precision, avoiding capital expenditure on network segments that phantom projects would never actually load.

The Goldilocks Calibration Problem

The policy design challenge is explicit and unresolved. Setting the deposit too low leaves speculators economically comfortable; setting it too high risks expelling legitimate development projects at early stages when capital commitment is genuinely uncertain. Daniel Mewton, a partner at Slaughter and May specializing in digital infrastructure, framed it directly: unwillingness to put large sums at risk during the development phase is rational behavior, not a signal of bad-faith intent.

The UK's attractiveness as a data center market compounds the calibration difficulty. High energy costs and scarce land already push developers toward the US and renewable-rich European markets. Ofgem acknowledged that differences between international energy markets make direct cost comparisons difficult, but that recognition does not resolve the competitive tension. Smaller AI-native operators — including firms such as Nscale that have previously pledged billions in UK investment — may find the deposit requirement a harder barrier than hyperscalers for whom the same figure is a rounding error on a campus-scale build. The risk is that reform clears phantom demand while simultaneously concentrating the surviving applicant pool among incumbents with the deepest balance sheets, structurally narrowing the competitive field for AI compute infrastructure in the UK.

Infrastructure Planning as the Real Prize

The deeper argument for Ofgem's approach is multi-year grid planning fidelity. As long as phantom load dominates the queue, neither grid operators nor data center developers can trust demand forecasts enough to commit capital efficiently — a dynamic with direct parallels to how pipeline architecture, not just raw model improvements, now determines where AI capacity gains actually materialize. Removing speculative applications makes the connection queue a functional signal rather than noise, enabling targeted network investment aligned to where compute infrastructure will actually land.

The UK's ability to attract a meaningful share of the hundreds of billions of dollars moving into AI infrastructure over the next decade depends less on the exact deposit figure Ofgem lands on in September than on whether the resulting queue produces reliable, actionable data for grid expansion planning. Getting that signal clean is the prerequisite for everything that follows.