Britain’s Data-Centre Market Is Becoming a Grid-Allocation Market

Britain’s Data-Centre Market Is Becoming a Grid-Allocation Market

*By Matthieu Gallego — 6 November 2026*

The most important part of Britain’s latest data-centre announcement is not the proposed scale. It is the timing.

On 3 November, Drax, Brookfield and Radiant announced a strategic partnership to develop an AI data-centre campus at Drax Power Station in North Yorkshire. The stated ambition is more than 1 GW, with a first phase targeting up to 100 MW and operations from the first quarter of 2028. Brookfield estimates that total investment could exceed £20 billion, including up to £2.5 billion for the first phase. Yet the announcement also makes the conditions explicit. Planning, regulatory and network approvals are still required, and the project needs a suitable tenant. The first phase is therefore a development target, not commissioned IT capacity.

Five days earlier, the UK government confirmed a fundamental change in the way large electricity demand will be connected. Strategic projects may receive priority, future capacity may be reserved, and post-2030 data-centre connections are intended to align with regional infrastructure targets. Those targets will consider grid capacity, water, land and competing demand.

Taken together, these two developments change the investment question. A power-station estate with land, generation, cooling and grid infrastructure can be an exceptional starting point. It is not an automatic route to live compute. In Britain, the winning projects will increasingly be those that can prove both strategic value and regional deliverability before scarce network capacity is allocated elsewhere.

The decision is now about allocation, not simply access

For an investor, operator or tenant evaluating a British site, the first question should no longer be, “How many megawatts are announced?” It should be, “What exactly can be allocated, by whom, on what date and under which policy route?”

That distinction matters at Selby. Drax says the estate covers about 1,000 acres, has 4 GW of grid access, 2.6 GW of active dispatchable generation and existing cooling systems. These are substantial infrastructure advantages. They reduce some of the problems faced by a greenfield project. They do not prove that 100 MW of data-centre load has a final connection agreement, that the capacity is reserved for a specific tenant, or that the full campus can be delivered on the announced schedule.

The UK government’s response makes this uncertainty more structured. Strategic demand projects can be prioritised when capacity is released from the queue. Government may reserve future capacity for individual projects, or for categories of strategic demand. It may also use powers to transfer data-centre connection contracts to strategic AI developments before 2030. After 2030, regional targets are intended to limit connection volumes.

This creates a new source of value and a new source of risk. A project that can demonstrate readiness, national relevance, a credible tenant and alignment with regional infrastructure may move faster than its position in the ordinary queue suggests. A speculative project with land and an early-stage connection request may move backwards, even if its headline capacity is larger.

What PowerlandMap’s UK evidence shows

PowerlandMap contained 260 verified public UK supply records at the 6 November review point. Eighty-three were classified in the active pipeline. Only 55 carried an explicit IT-capacity figure, representing 3,120.06 MW in total.

Those numbers measure the current documentation in the database. They are not a complete estimate of the UK market, and the 3,120.06 MW should not be added to announced campus or electrical figures. Many records have strong evidence for location, developer or project stage without a comparable IT-capacity value. That is precisely the market problem the new policy will intensify. Capacity claims are plentiful. Comparable evidence of what can be energised, when and at which measurement boundary is much scarcer.

The Selby record is deliberately conservative. PowerlandMap identifies the developer group, location, announcement date and stated phasing. It does not record the “more than 1 GW” ambition or the first 100 MW target as verified IT capacity. The capacity basis has not been established, and the project remains subject to planning, network approvals, tenant commitments and final investment decisions.

That treatment is not pessimism. It is the minimum discipline required when policy is moving toward allocation.

A 100 MW label can describe two different projects

The first sensitivity investors should run is not a complex valuation model. It is a measurement-boundary test.

The Drax announcement describes a first phase of “up to 100 MW” but does not, in the materials reviewed for this article, define whether that figure is facility input or IT load. The difference changes the connection requirement, the equipment scope and the energy bill.

The calculation is transparent. If 100 MW is facility input, IT load equals facility input divided by PUE. If 100 MW is IT load, facility input equals IT load multiplied by PUE. Annual energy equals facility input multiplied by 8,760 hours and 90% utilisation.

At a PUE of 1.20, the two interpretations differ by 20 MW of connection capacity and about 158 GWh of annual electricity. That gap is large enough to affect grid studies, transformer capacity, procurement strategy and operating economics. It also shows why a policy that allocates regional connection volumes cannot treat every “100 MW” announcement as equivalent.

This sensitivity is illustrative, not a forecast for Selby. Actual design PUE, utilisation, redundancy, auxiliary loads and phasing are unknown. Its purpose is to identify the document that matters next: a dated electrical single-line and load schedule linking IT capacity, total facility demand and the contracted connection.

Why the power-station model is strategically attractive

The strongest argument for Selby is not that existing generation makes grid constraints disappear. It is that the site can potentially combine several assets that usually have to be assembled separately.

First, land and industrial infrastructure already exist at scale. Second, the estate has established grid and generation relationships. Third, cooling systems and fibre proximity may reduce development complexity. Fourth, Brookfield and Radiant bring capital and AI-infrastructure capability, while Drax brings the site and energy platform.

This combination could shorten the path between project concept and credible delivery. It could also support phased investment. A 100 MW first phase gives the partners a practical unit for planning, tenant discussions and network studies before committing to the full campus.

The broader UK policy could reinforce that advantage. The government wants to prioritise strategic demand and align future data-centre growth with realistic regional infrastructure. A brownfield energy site with a clear economic-development case may compare favourably with a speculative project in an already congested cluster.

But strategic alignment is not a certificate that can be assumed from the announcement. The government has not yet published the final regional targets or the complete framework for selecting projects. The Drax partners say their initial assessment is that the project aligns with the Data Centre Covenant. That is a sponsor assessment, not a formal allocation decision.

The serious objection: a power site can still be the wrong compute site

The counter-argument deserves equal weight. Reusing a power-station estate may optimise the energy side of the project while leaving other constraints unresolved.

A hyperscale or AI tenant will test latency, fibre diversity, water strategy, construction supply chains, workforce, security and the ability to operate in multiple phases. It will also assess the commercial structure. On-site generation does not automatically translate into a competitive, bankable and low-carbon electricity product for a data centre. The treatment of network charges, balancing, carbon accounting and backup supply can materially change the economics.

There is also a concentration risk. A very large campus may require several tenants or one exceptionally large commitment. If the first phase depends on a tenant before final investment, the project must align its technical design and delivery date with that tenant’s actual deployment plan. The policy route may favour strategic projects, but it will also increase scrutiny of readiness and credibility.

The alternative is a portfolio of smaller sites closer to existing demand centres. Those sites may face tighter power constraints, but they can offer lower latency, established ecosystems and incremental absorption. The power-station model wins when energy and schedule advantages outweigh distance and concentration costs. It loses when the connection route, tenant commitment or network design remains too conditional.

What would change my view

I would treat the Selby project as materially de-risked when five pieces of evidence converge.

1. A planning application defines the physical phases and environmental scope. 2. A network document identifies the connection boundary, contracted demand and energisation milestones. 3. The partners state whether each MW figure is IT load, facility input or another measure. 4. A tenant commitment supports the first-phase design and investment decision. 5. The government or system operator confirms the project’s treatment under strategic prioritisation and future regional allocation.

None of these tests requires the project to be operational before it becomes investable. They require the commercial, planning and electrical stories to describe the same asset.

Britain’s next market map will be a hierarchy of evidence

The UK is not closing the door to data-centre growth. It is replacing an open queue with a more directed system. That should favour projects capable of demonstrating readiness, public value and credible infrastructure use. It should also reduce the value of early-stage capacity claims that cannot survive a regional allocation test.

Selby is an important case because it brings the new model into focus. The project has an unusually strong industrial base, experienced partners and a phased route. It also has unresolved conditions that cannot be collapsed into a single gigawatt number.

For investors and operators, the practical response is to rank British projects by evidence rather than ambition. Land ownership, planning status, connection rights, capacity basis, tenant support and allocation route should sit in one dated decision record. A project should move up the shortlist only when those elements advance together.

Britain’s data-centre opportunity remains substantial. The scarce asset is no longer simply powered land. It is a project that can prove why it deserves power, where the system can accommodate it and how quickly that allocation can become live IT capacity.

Sources and method

Primary sources reviewed: Drax’s 3 November 2026 partnership announcement, the London Stock Exchange RNS notice, and the UK government’s 28 October 2026 response on strategic demand connections. PowerlandMap figures were calculated from verified public UK supply records on 6 November 2026, excluding demo records. Record counts measure database coverage, not market completeness. IT-capacity totals include only records with an explicit IT MW value. The sensitivity table is an illustrative calculation and not a forecast for the Selby project.