1. What is the Drax data centre partnership?
Drax Group announced on Tuesday, November 3, 2026, that it has signed a joint development agreement with Brookfield Asset Management and Radiant, Brookfield's neocloud platform, to build a gigawatt-scale artificial intelligence data centre at the Drax Power Station estate in North Yorkshire, England. Brookfield estimates that lifetime investment in the Drax data centre project could exceed £20 billion, or about $27 billion. Drax shares rose more than 8% on the news.
The plan is notable for its scale and for its location. Instead of building on a new site and waiting for a connection to the electricity network, the partners intend to use land and power infrastructure that already exists at a major generating station. That approach is aimed at one of the biggest obstacles facing companies that want to expand AI computing capacity.
The announcement covers an initial build of 100 megawatts and a longer-term ambition of more than one gigawatt. It also sets out a first-phase budget and a target date for commercial operations, though both depend on approvals that have yet to be secured.
Readers should note the distinction between the sterling and dollar figures. The £20 billion amount is the one used in the project's own description, and the dollar equivalent of about $27 billion is a conversion at current exchange rates, so the headline total may move slightly as currencies fluctuate.
2. How will the project be built in phases?
The first phase is a 100-megawatt compute facility with a budget of £2.5 billion, or roughly $3.4 billion. Commercial operations are expected to begin in the first quarter of 2028. That makes phase one the test of whether the concept works at the site, and the opening step in a program intended to grow much larger.
Subsequent phases would add about 500 megawatts of capacity between 2028 and 2031. The stated goal is for the site to exceed one gigawatt in total. To put that in perspective, a gigawatt is roughly the output of a large conventional power plant, and data centres of that size are only now being planned as demand for AI workloads climbs.
Funding for the first phase is expected to draw in part on a Brookfield joint venture with Qatar's sovereign wealth fund, the Qatar Investment Authority (QIA). Large AI infrastructure projects are typically financed by a mix of institutional capital, debt and long-term customer commitments, so the involvement of a sovereign investor is consistent with the capital needs of a project that runs into the billions.
The phase one plan remains subject to planning and regulatory approvals, network approvals, commitments from anchor customers and a final investment decision. Until those conditions are met, the figures describe an intention and a potential, not money that has been spent.
The phasing is a deliberate way of managing risk. By committing to 100 megawatts first, the partners can test construction costs, customer demand and operating performance before releasing the much larger sums attached to later stages. If phase one underperforms, the later phases can be reconsidered; if it succeeds, the project can scale with evidence behind it, which is how most large infrastructure programs are structured.
3. Why is a power station a good site for AI computing?
Electricity supply is now the main constraint on building AI data centres. Training and running large models requires enormous, continuous power, and in many European countries a developer can wait years for a connection to the grid. A site that already has a high-capacity link and its own generation avoids much of that delay.
Drax Power Station provides both. The estate has an existing grid connection and dispatchable generation, meaning electricity that can be produced on demand and not only when weather conditions allow. For a data centre, steady and controllable supply is more valuable than intermittent output, because computing clusters need to run without interruption.
That combination explains why the partners describe the location as a way to move faster than a conventional greenfield project. Land, connections and power sources are already in place, so the main work is building the facility itself and securing the customers who will use it.
There is a further advantage in speed. Because the partners do not have to wait in a long queue for a new connection, they can aim at a first commercial date in early 2028, much sooner than would be realistic for a comparable site starting from nothing. In a market where customers want capacity as soon as possible, a shorter path to operation is a competitive asset.
4. What is a neocloud and where does Radiant fit?
Radiant is described as Brookfield's neocloud platform. A neocloud is a company that rents out large amounts of specialized computing power, usually built around graphics processors, to AI developers and businesses. These firms emerged to meet demand that the largest traditional cloud providers could not satisfy quickly enough, and they typically focus on dense, high-power installations designed for AI.
In this arrangement, Radiant is positioned as the operator that would offer the computing capacity, while Brookfield, one of the world's largest alternative asset managers, brings the capital and development experience. Drax contributes the site and the energy infrastructure. Each partner supplies what the others lack, which is the logic behind a joint development agreement.
The structure also shows how investors are approaching AI infrastructure. Instead of backing only chips or software, large asset managers are financing the physical foundations, including land, power and buildings, and then leasing them to the companies that need the capacity.
For customers, the appeal of such a platform is access to ready capacity without building their own facilities. AI developers face long waits for hardware and power, so a provider that can promise delivery on a defined schedule, at a site with its own supply, offers a practical alternative to constructing or leasing elsewhere.
5. Why are asset managers financing AI infrastructure?
Data centres have become a favored target for infrastructure investors because they can offer long-term, contracted income. A tenant that commits to renting capacity for many years provides predictable revenue, which lenders and institutional backers can use to justify large upfront spending. That structure resembles the financing of other long-lived assets such as toll roads, pipelines and power plants.
The AI wave has changed the scale of those projects. Facilities once measured in tens of megawatts are now planned in hundreds, and the largest proposals aim for a gigawatt or more. At that size, a single campus can require billions of pounds, which explains why the participants include a global alternative asset manager and a sovereign wealth fund instead of only technology firms.
For a company like Drax, the appeal is different. It already owns the site, the connection and the generating plant, so partnering with a specialist developer lets it put those assets to work without taking on the full burden of building and operating a computing business. The risks and the rewards are shared among the partners according to what each contributes.
6. How did the market react?
Drax shares jumped by more than 8% after the agreement was disclosed. The move reflects investor interest in a company whose assets, long associated with power generation, may now be valued for a second purpose. A partnership of this size could add a new source of income for Drax if the project reaches its later phases.
The reaction should be read with care. The headline investment of more than £20 billion is a lifetime figure that stretches over several phases and years, and Drax itself would not be funding all of it. The nearer-term commitment is the £2.5 billion first phase, which is itself conditional on approvals and customer agreements.
Shares of companies tied to AI infrastructure tend to move sharply on announcements of this type, since investors weigh the long-term opportunity against the uncertainty over execution. A rise of more than 8% in a single session shows how strongly the market responded to the possibility of a new revenue stream at an established generating site.
7. What are the risks to the plan?
The principal risk is demand. The investment case rests on sustained growth in AI computing, and the spending required is large enough that a slowdown in demand, or a shift in how customers buy capacity, would affect returns. The first 100 megawatts therefore serves as the proof of concept: if the facility attracts anchor tenants and operates as intended, the case for the larger phases becomes stronger.
There are also practical hurdles. Planning permission, network approvals and a final investment decision still have to be obtained. Large developments can face scrutiny over land use, energy consumption and local impact, and timetables can move if any of those processes take longer than expected.
Finally, the economics of AI infrastructure are being watched closely across markets. Investors are asking how quickly the huge sums being committed will be recovered, which is why firm customer commitments are among the conditions attached to the first phase.
8. What's next for the Drax data centre project?
The next steps are the approvals and commitments listed for phase one: planning and regulatory consents, network approvals, agreements with anchor customers and a final investment decision. Progress on any of these would be the first concrete signal that the plan is moving from announcement to construction.
If those conditions are met, commercial operations at the first 100 megawatts are expected in the first quarter of 2028, followed by roughly 500 megawatts more between 2028 and 2031. Investors in Drax and observers of the UK's AI infrastructure buildout will be watching whether the schedule holds, and whether the site can reach the gigawatt scale the partners have set as their goal.
Further announcements are likely to focus on who the anchor customers are, how the financing is structured and how the planning process is progressing, since each of these determines whether the first phase can proceed on the schedule now set out.