The £20bn Drax data centre: power first, promises later

The most arresting number in Britain's AI race is not a chip count or a model parameter. It is 4GW of grid access at a legacy Yorkshire power station. Brookfield and Qatar Investment Authority's AI company Qai, through their joint venture, have partnered with Drax and the Brookfield-backed compute platform Radiant to develop a data centre platform at Drax Power Station in North Yorkshire. Total investment could potentially exceed £20bn, with an initial £2.5bn for a 100MW first phase targeted for commercial operation in the first quarter of 2028. Capacity is expected to expand to 500MW between 2028 and 2031, with further expansion beyond 2031 across the 1,000-acre site. The site already has 4GW of grid access and 2.6GW of dispatchable generation capacity. Drax shares rose 6% to 863p on the news.

That is the shape of the deal, and it tells you where the UK's compute ambitions are actually being built: not on new generation, but on the copper, connections and dispatchable plant that already exist. Drax Group CEO Will Gardiner said Drax Power Station is one of the few UK sites that can provide large-scale reliable power directly to data centre customers using existing and underutilised generation capacity. He is describing a scarce asset. The first phase will draw power directly from the grid, while the site's 2.6GW of dispatchable generation sits behind it. The project is the first by the Canadian-Qatari joint venture established in December 2025, which is itself a signal: international capital is treating co-located, grid-connected sites with dispatchable capacity as strategic AI infrastructure.

The political question is whether that capital can be converted into steel and servers on the timetable advertised. Phase one progression is subject to planning, as well as receiving the required regulatory and network approvals. That is not a formality. It is the whole test. The government has promised faster power connections and planning access for exactly this kind of project. The Drax scheme is the first serious opportunity to see whether those promises shorten a real timeline or merely decorate an announcement. If the 100MW first phase reaches commercial operation in the first quarter of 2028, the promises will have something to point to. If it slips, the gap between ministerial rhetoric and delivery will be measured in quarters, not press releases.

There is a fair case against this reading, and it deserves to be stated plainly. The project is still subject to planning, regulatory and network approvals. It is too early to judge whether government promises have failed or succeeded. A first phase of 100MW is small relative to the eventual 500MW and the wider 1,000-acre site, and the staged dates — 2028 for phase one, 2028 to 2031 for expansion to 500MW, and beyond 2031 for further growth — are targets, not guarantees. The £20bn figure is described as potentially exceeding that level, not as a committed total. The final split of investment and ownership between Brookfield, Qai, Radiant and Drax has not been set out. On this view, the only honest verdict is that the approvals process has not yet had its say, and treating a target date as a delivery date is exactly the kind of premature judgment that makes sensible infrastructure policy harder.

That objection is correct as far as it goes, but it does not neutralise the argument. It sharpens it. The point is not that the Drax project has failed. The point is that it has become the test case for a government promise that was always going to be judged on delivery rather than design. The UK's AI ambitions are being tied to existing industrial power infrastructure because new-build generation cannot be conjured at the speed the compute buildout demands. That is a rational response to constraint, but it also means the binding limits are planning, regulation and network access — the very areas where government has promised speed. A project that still needs those approvals is not evidence of failure. It is evidence of where success or failure will be decided.

For investors, the signal is already clear: grid-connected sites with dispatchable capacity are becoming strategic AI assets, and Drax has a new role beyond its legacy generation business. For ministers, the signal is less comfortable. The Drax scheme will not be judged by the size of its potential investment or the rise in Drax's share price. It will be judged by whether a 100MW first phase reaches commercial operation in the first quarter of 2028, and whether the path to 500MW between 2028 and 2031 stays open. Until then, the gap between promise and delivery remains exactly that — a gap. The project is the first real chance to close it, and the first real chance to show it cannot be closed on the current timetable. Both outcomes are still live. That is precisely why it matters.