TL;DR · 30-second read
The Short Version
The investment firm Brookfield wants to build a giant computer complex for artificial intelligence on the grounds of a power station in Yorkshire, England. The full plan could cost $27 billion or more.
The surprise is how small it starts. Only about a tenth is due by early 2028, and nothing is final: it still needs government approvals and a paying customer.
Why care? These computers devour electricity, and Britain has struggled to connect big new users to its power network quickly. This project will test whether it can.
CNBC reported on November 3, 2026 that Drax Group has signed an agreement with Brookfield and Brookfield’s neocloud platform Radiant to jointly develop an AI data center on the Drax Power Station estate in North Yorkshire. A neocloud is a newer cloud provider focused on renting out AI computing power. Brookfield estimates total investment, including semiconductors, could exceed £20 billion ($26.8 billion), starting with £2.5 billion in phase 1.
The project targets up to 100 megawatts (MW) from the first quarter of 2028, another 500 MW between 2028 and 2031, and eventually more than 1 gigawatt (GW, or 1,000 MW) by an unspecified date. Phase 1 still requires planning, regulatory and network approvals and a tenant, and final investment decisions have yet to be made. Drax shares rose 8.8% in early London trading.
Executive Summary
The Drax agreement is one of the larger AI data center ambitions announced for Britain, but its structure matters as much as its size. The price tag covers a multi-phase build, including chips, and only the first 100 MW has a near-term start date. That tranche represents at most about a tenth of the eventual target, while the £2.5 billion first-phase investment is one-eighth of the £20 billion estimate.
The plan arrives as the UK contends with high energy costs and long waits for grid connections. In October, Microsoft was reported to have pulled out of a deal to anchor an Nscale data center just outside London amid connection delays. That makes the Drax project a practical test of whether a power station estate can help an AI development move faster in Britain.
The answer is not yet known. Phase 1 will draw power from the grid and still needs network approval, so the location brings potential advantages without removing the central constraint.
A $27 Billion Headline Built on a 100 MW First Step
The headline estimate is best read as the potential value of a staged program, not as committed spending. Brookfield says total investment could exceed £20 billion, and that figure includes semiconductors. Chips can account for a large part of an AI facility’s cost, so the number should not be compared directly with estimates that cover only land, buildings and electrical systems.
The phasing shows how much is still to be decided. The dated portions add up to 600 MW by 2031, which leaves at least 400 MW of the 1 GW-plus ambition without a timetable. Rough arithmetic puts phase 1 at about £25 million per MW if the £2.5 billion maps to the full 100 MW, against roughly £20 million per MW for the wider plan. Both figures are approximate because the capacity and spending targets are open-ended, and phase 1 is not described as including or excluding chips.
Why Power Access Will Set the Pace
The most important operational detail is that phase 1 will use power from the grid. It therefore depends on the same network that has caused difficulty for other UK projects, and it remains subject to network approvals. With commercial customers targeted from early 2028, the developers have less than a year and a half from the announcement to the start of the operating window, and planning, regulatory and network approvals are all still outstanding.
The mechanism is straightforward. An AI data center cannot earn revenue until it has enough electricity, and a tenant needs confidence that the power will arrive on schedule before signing. Delay in one area can therefore affect the other. The reported Microsoft withdrawal from the Nscale project shows how that risk can play out, with grid connection delays cited as the backdrop to an anchor customer’s departure. Nscale told CNBC it remained “committed” to the project and was “exploring alternative power sources for the site.”
The same dependency applies at Drax, where securing a tenant is also a condition of the first phase. Developers, chip suppliers, lenders and prospective customers all have a stake in the connection timeline, because it will shape when capacity can be sold and when capital starts earning a return. The capital estimate describes the scale of the ambition, while power delivery will largely determine how quickly it can be realized.
A Power Station Estate Is an Advantage, Not a Shortcut
The logic of the site is easy to see. Power stations are built around large electrical infrastructure and have an established relationship with the grid, and scarce connections are a known UK constraint. It is reasonable to read the Drax estate as a bid to start from a stronger position than a greenfield site, meaning undeveloped land with no existing energy use. The 8.8% rise in Drax shares suggests investors saw value in the announcement, although one morning’s trading cannot show which element they valued most.
Still, one agreement does not establish a national trend, and the approach has limits. A connection designed to send electricity out is not automatically cleared to bring large volumes in, which helps explain why network approval remains significant here. Whether later phases will also rely on the grid, or on electricity arranged at the site, is one of the key questions for the project’s economics.
Britain’s Pitch Meets Nordic Competition
Radiant CEO Mahdi Yahya framed the project as “a unique opportunity for the UK’s AI ecosystem to build at scale,” one that would give “British researchers, startups and enterprises access to frontier compute at home.” That is a clear statement of intent, but its substance depends on who ultimately takes the capacity. A single large tenant could fill a first phase, which would leave open how much broad domestic access the project would provide.
The UK also faces competition for this investment. The Nordics benefit from abundant power and available land, and Google said in September that it would invest at least €13 billion ($15.1 billion) in AI infrastructure in Finland, its largest single investment in Europe. The Drax figure is larger, but it is a conditional estimate for a project awaiting final investment decisions. If Drax reaches its 2028 milestone, it would give Britain a meaningful proof point. If approvals or connections slip, the comparison with regions that have more readily available power will become harder to answer.
Background
Brookfield is an asset manager, and Radiant is its neocloud platform. Drax Group is a London-listed energy company that operates Drax Power Station in North Yorkshire, the site of the proposed development.
The U.S. has moved ahead of Europe in developing the very large data centers needed for AI, while Europe has been slower to break ground amid high energy and land costs. Within Europe, the UK has struggled with expensive energy and long grid connection waits, while the Nordics have attracted major commitments thanks to abundant power and available land. Source: Brookfield plans for a $27 billion AI data center at UK power station, CNBC’s report on the planned AI data center development with Drax and Radiant in North Yorkshire.Sources