£713k Grid Fees Threaten Britain's Datacentre Land Grab

Britain has 315 datacentre projects asking for 73GW of grid capacity, almost 30GW more than the country’s peak electricity demand of 45GW. That gap is the real story behind Ofgem’s proposed datacentre grid access fees: the regulator is no longer treating the connection queue as a neutral waiting list. It is treating it as a scarce public resource being distorted by projects that may never be built.

Ofgem’s plan would force large datacentre developers to put up serious financial security before holding a grid connection slot, according to Guardian World. The proposal targets “speculative” projects that reserve capacity while hospitals, schools, housing developments, clean power schemes and other energy users wait behind them.

Britain’s datacentre boom just hit the hard wall of grid scarcity

The central signal is blunt: cheap grid reservations are over if Ofgem’s consultation becomes policy.

The regulator wants a Data Centre Commitment Fee for large projects accepting a connection offer. The fee would sit between £237,500 and £712,500 per megawatt, equivalent to 2.5% to 7.5% of average project costs, according to Ofgem. Developers could provide security through upfront fees, letters of credit, bonds or cash deposits.

That changes the economics of optionality. A company can still pursue a major AI or cloud facility, but it may need to prove it has capital, procurement plans and commercial maturity before it gets to occupy space in the queue.

“Britain's electricity demand connections queue has more than tripled in size in less than a year, and consumers should not bear the risks created by speculative projects taking up space in the system,” said Eleanor Warburton, Ofgem’s Director for Energy System Design and Development.

This is not just a datacentre problem. Britain’s AI ambitions now depend on electricity infrastructure as much as chips, models and engineers.

The 73GW datacentre queue exposes grid planning failure in plain numbers

The figures are out of proportion with any credible near-term buildout.

Ofgem says 315 datacentres are queueing to connect, representing 73GW of demand. Britain’s peak electricity demand is about 45GW. Ofgem also says demand connection applications rose from 41GW in November 2024 to 125GW by June 2025, with data centres accounting for at least 80GW.

A queue does not equal actual future demand. Some projects will fail to secure financing. Others may be delayed, sold, redesigned or abandoned. But phantom demand still matters because network planners must respond to the queue they can see, not the projects that will ultimately survive.

That creates two failures at once:

• Delay: Viable projects can sit behind weaker applications.
• Distortion: Network operators receive inflated signals about where reinforcement may be needed.
• Consumer risk: Ofgem says consumers should not carry risks created by speculative applications.
• Policy conflict: Grid capacity is also needed for electrification, housing, schools, hospitals and clean power projects.

The UK government’s own “compute roadmap” calls for “at least 6GW of AI-capable datacentre capacity by 2030”. The National Energy System Operator’s conservative forecast for Great Britain’s datacentre capacity by the end of the decade is 3.7GW, with top-end estimates of 6.7GW. Against those numbers, a 73GW datacentre queue looks less like demand forecasting and more like land-grabbing for power rights.

Datacentre grid access fees would turn speculative plans into expensive bets

Ofgem’s proposed datacentre grid access fees work by making a connection slot costly to hold.

The fee would be refunded when the project reaches energisation. If the developer exits the queue early, it would be forfeited. The Guardian reports that providers, including companies such as Amazon and Microsoft, could be required to pledge £350m to reserve 500MW of grid capacity for new facilities.

That is the point. A weakly financed developer may think twice before tying up capital or credit support for a project that exists mainly as an option. A committed developer has a reason to accept the cost if the grid slot is critical to delivery.

Ofgem also wants datacentre-specific milestones. Developers would need to show progress through evidence such as:

• Financial capability: Proof the project can be funded.
• Commercial maturity: Evidence the plan has moved beyond an early concept.
• Procurement activity: Signs the developer is preparing to build, not just reserve capacity.

XOOMAR analysis: the policy does not create new grid capacity. It cleans the queue. That is useful, but it is not a substitute for physical network investment, faster connection works or clearer government prioritisation of strategically important demand.

Hyperscalers, networks, investors and communities will read the same plan differently

Ofgem’s proposal divides the market by balance sheet strength.

This is where the AI story becomes concrete. The UK can announce AI growth zones and compute targets, but datacentres still need grid connections. If those connections are scarce, the state has to choose between first-come, first-served allocation and a harder test of readiness.

For XOOMAR readers following AI risk and infrastructure from other angles, this sits beside questions raised in Escaped AI Agent Hits Hugging Face in OpenAI Security Test and Hospitals Face Legal Heat for Healthcare AI Failures. Software ambition is moving faster than the institutions and infrastructure around it.

Britain’s grid queue problem echoes earlier clean power bottlenecks

Ofgem has already faced a wider backlog of power generation and storage schemes waiting years for grid connections. The government-owned National Energy System Operator used reforms last year to clear hundreds of “zombie” projects from the queue and make room for clean electricity schemes with a stronger chance of meeting government energy targets.

There is evidence that tougher queue discipline can work. Ofgem says wider connection reforms have accelerated around 7.8GW of projects by an average of six years.

The datacentre queue may be harder to fix. There is little consensus within government over how much data capacity the UK will need in the coming years, despite the push for AI growth zones. That uncertainty makes it easier for developers to over-apply and harder for planners to separate real demand from tactical positioning.

The pressure is not confined to Britain. The Uptime Institute identified 250 global projects exceeding 100MW in energy demand announced between 2021 and 2024. It believes half will either not happen or be delayed, yet still expects an “unprecedented and rapid” increase in power required over the next five years.

Ofgem’s shake-up will decide which AI-era datacentres survive the power crunch

The next test is whether Ofgem can set the fee high enough to deter speculation without blocking credible projects that already face heavy development costs.

If the proposal survives consultation, weaker projects should start leaving the queue rather than posting meaningful security against uncertain timelines. Better capitalised operators may gain an advantage. Smaller developers may need partners, stronger offtake commitments or sites in areas with more available grid headroom.

The consultation runs until 16 September 2026. Evidence that would support Ofgem’s thesis would be a thinner queue, faster connection dates for viable projects and clearer network planning signals. Evidence against it would be credible projects being pushed out because the financial security burden is too blunt.

The necessary filter is now visible. The unresolved question is whether Britain can build enough grid capacity quickly enough to support AI growth without squeezing households, public services and other industries competing for the same electrons.

Impact Analysis

• Ofgem’s proposal could force speculative datacentre projects to prove financial commitment before occupying scarce grid capacity.
• The policy aims to free up connection slots for hospitals, schools, housing, clean power schemes and other users.
• Large AI and cloud infrastructure projects may face higher upfront costs to secure electricity access in Britain.