Drax site could host 1GW AI datacentre as grid and water questions mount

The plant burns woody biomass pellets and produces almost 5% of the UK's electricity. According to the Guardian, that makes it the single largest emitter of carbon dioxide in the country. Those emissions are not counted against the UK government's books. Drax argues that growing trees absorb an equivalent amount of carbon.

That argument is why the deal is worth watching beyond North Yorkshire. A datacentre on the site would sit next to the country's biggest single source of disputed carbon. The power it draws would be counted somewhere.

Who pays, who powers

Under phase one, Drax provides land and access to the national grid through existing connections. Up to 100MW of compute would run from the first quarter of 2028. Drax hopes to supply power directly by the early 2030s. It says the project could pull in more than £20bn ($26.8bn) in investment. Chief executive Will Gardiner said it would "help secure the long-term future of the site and protect hundreds of jobs". The company's market value rose almost 9% on the London Stock Exchange to almost £3bn, the Guardian reported.

The timing matters. Ministers have promised Drax a "much more limited role" in future. It would get access to only half its current £2m a day in subsidies from 2027 to 2031. A datacentre tenant is a replacement revenue line for a plant whose public support is shrinking.

The grid is the other half of the story

Two days earlier, the Guardian reported that National Grid had agreed to buy 29.5 hectares (73 acres) at Dorman Point on Teesworks land. The site would host what would be Britain's biggest substation, serving a planned datacentre of nearly 500,000 sq metres. Ben Houchen, mayor of the Tees Valley combined authority, told Sky News that up to 3.2GW of capacity there would let the region compete for projects.

"Everyone who has tried to build a datacentre in this country knows the biggest problem is getting enough power to the site quickly," Houchen told Sky News, according to the Guardian.

Water is the quieter constraint. Neither Drax nor its partners have published a figure for the project's cooling demand. That gap is not unique. Recent headlines are dominated by water use, local opposition and cooling design, from Dallas rules to Alberta hearings. Those items cannot be cited as fact here, but they show where scrutiny is landing.

The model side of the ledger

The compute being built has a counterpart in cheaper inference. Gartner analysts Mark Petty and Raymond Paquet wrote that nine in 10 agent workflow steps are routine. Open-weight models handle them "at a quarter of the price of the cheapest closed model", they wrote. That is as little as a twentieth against a frontier model run on every step. Their realistic estimate is five or six times cheaper per task, The Register reported on 3 November. The firm's planning assumption: by 2030, 90% of enterprise operational agent workflow steps will run on open-weight models.

If that holds, the economics of giant single-site datacentres start to depend on how much frontier-class work remains. On 3 November, Tom's Hardware reported that Google's Gemini 4 Carbon scored 59 on Artificial Analysis' Intelligence Index V5. That was one point ahead of Claude Opus 5.5, at $7.76 per task. The same benchmark suite put the best hallucination score at just 54 on a -100 to +100 scale.

Elsewhere the physical side of AI is scaling too. Tutor Intelligence said on 2 November it had raised $75m in Series B led by Andreessen Horowitz. It expects more than 1,000 robots under contract by the end of 2027, according to The Robot Report. Robots do not need cooling towers, but they do need factories.

Drax still needs regulatory approval to supply the site directly before 2031. Until then, the £20bn is a plan, not a power purchase.