Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

Drax and Brookfield plan a 1GW AI datacentre on Britain's biggest emitter

Drax has struck a £20bn deal with Brookfield and Radiant to build a datacentre of more than 1 gigawatt at its North Yorkshire biomass plant. Up to 100MW of compute will run from the first quarter of 2028, the Guardian reported on 3 November.

AI & modelsExplainerGrace OkonkwoPublished: 3 November 20266 min readSources 14
Drax and Brookfield plan a 1GW AI datacentre on Britain's biggest emitter
Drax has struck a £20bn deal with Brookfield and Radiant to build a datacentre of more than 1 gigawatt on the

The site is the UK's single largest emitter of carbon dioxide, according to analysts quoted by the Guardian. It produces almost 5% of the country's electricity by burning woody biomass pellets. Now it is being recast as an AI campus.

The deal, announced on 3 November, brings in the infrastructure investor Brookfield and the AI infrastructure platform Radiant. Drax will contribute land and access to the national grid through existing connections at the site. Direct supply to the datacentre, rather than grid supply, is pencilled in for the early 2030s. That is subject to regulatory approvals the company hopes to clear before 2031. The structure matters because it shifts the plant's economics onto a tenant with a steady, very large appetite for power. It does so while the plant's own subsidy regime is being rewritten in Westminster. That tenant is not a utility or a manufacturer but a compute operator, and compute operators sign long contracts.

The subsidy clock is the real driver

The timing is not accidental. The government has said Drax will play a "much more limited role" in future after sustained criticism of its sustainability claims. From 2027 to 2031 the plant gets access to only half its current £2m a day in subsidies. A datacentre tenant with a 1GW appetite is one way to replace revenue that is being legislated away.

Drax chief executive Will Gardiner said the project would "help secure the long-term future of the site and protect hundreds of jobs". He said it would also be "supporting the north's reindustrialisation and strengthening sovereign AI capability". The company described it as "the next chapter" for the plant. Its market value rose by almost 9% on the London Stock Exchange to almost £3bn, the Guardian reported.

Whether the electrons are green is a separate question. Drax's emissions are not counted against the UK government's inventory. The company argues the equivalent carbon is absorbed by the trees that supply its pellets. Analysts have disputed that accounting for years. The Guardian's report notes the site remains the country's largest single source of CO2 on the books of the people who track it. The announcement landed a week after the government published a draft national policy statement on datacentres. It called them "as fundamental to our economy as the railways and electricity networks on which we have long depended". Ministers added: "We must take control over the development of datacentres in this country and ensure that the country and host communities benefit from them."

Grid capacity is the constraint nobody has solved. The Guardian noted on 1 November that National Grid has secured a site for what will be the UK's largest substation, near a planned datacentre in Teesside. Building the wires is one problem. Finding the generation to put through them is another.

Behind the meter beats shifting the load

An Ember analysis published in October 2026 and reported by pv magazine on 3 November puts numbers on the flexibility question. It covers nine markets across the US, Europe and Southeast Asia. The headline finding: shifting computing workloads around is a weak tool. On-site generation or storage does far more.

In Ireland, datacentres already account for roughly 19% of peak demand. In January 2025, when system demand peaked at 6GW, they drew an estimated 1.1GW. Ember projects that share rising to 24% by 2030, with datacentre demand at 1.6GW against a system peak of 6.5GW. Load shifting could cut that peak by about 100MW, or 1.5%. Across the other markets analysed, the estimated reductions run from 0.1% to 0.6%.

The reason is workload mix. Ember estimates that cloud, enterprise and AI inference, all latency-sensitive, will be about 87% of global datacentre capacity by 2030. Under its assumptions, 35% of AI training is flexible, 5% of inference is, and traditional datacentres offer nothing. That leaves roughly 6.5% of global capacity theoretically shiftable. About 3% is effectively available once an average utilisation rate of roughly 48% is applied.

Bring-your-own-capacity means on-site renewable generation, batteries and backup running alongside the grid. It could deliver nearly five times more relief, with potential reductions of up to 2.7GW in Texas. The mismatch driving all of this is simple. A large datacentre needs 200MW to 500MW of capacity and can be built in 18 to 24 months. Conventional grid expansion takes four to 10 years.

Canary Media made a related argument on 3 November: batteries rather than gas plants should be the default pairing for new datacentre load. The turbine route is already in litigation. Tom's Hardware reported on 2 November that a judge ruled gas turbines at xAI's Mississippi datacentre are not exempt from the Clean Air Act. SpaceX had failed to dismiss a lawsuit over the pollution.

The same site is running eight converted jet engine turbines, Tom's Hardware reported on 3 November, citing the New York Times. Permits allow a 25% higher limit for volatile organic compounds. The Southern Environmental Law Center has challenged Mississippi's approval, arguing xAI has not shown the turbines meet standards for pollutants including formaldehyde.

The queue is enormous and the politics are turning

Dominion Energy, Virginia's primary electric utility, said in its Q3 earnings report that it has 54.9GW of datacentre capacity in various stages of contracting, a 13% increase since the end of last year. More than 12GW is signed under electric service agreements. CFO Steven Ridge said on the call, reported by DatacenterDynamics on 3 November, that the company had added more than 6GW of contracts since the end of last year.

That pipeline is meeting a different mood. Tom's Hardware reported on 2 November on a database of more than 10,000 public comments in which over 90% opposed datacentres. Water was the top concern and only 21% mentioned AI. In Ohio, Time reported on 29 October that the datacentre revolt is upending the midterms.

Google's Minnesota project won approval only after the company dropped an $80M tax break request and scaled the 200-acre scheme down, Tom's Hardware reported on 2 November. The vote followed two delayed council votes and local opposition. Tom's Hardware also reported on 3 November that OpenAI has discussed rebranding datacentres as "AI compute campuses". It reportedly sees the 2028 presidential election becoming an AI issue.

The contract side is moving too. Anthropic's Australian subsidiary has signed non-definitive lease documentation with Australian Data Centres, 85% owned by Dexus, for the first stage of a hyperscale campus in Western Downs, Queensland, DatacenterDynamics reported on 3 November. Dexus CEO Ross Du Vernet told investors ADC has "launched a fundraising process to introduce additional capital to the business, of which Dexus is yet to determine the extent to which it will participate."

On the supply chain, Tom's Hardware reported on 3 November that Nvidia has told datacentre customers to cap business with Chinese customers at 49% of the total. Sales staff and regional teams passed on the guidance, not any published document. The same outlet reported on 3 November that Envision Energy has completed what it calls the first commercial 800V DC power deployment at its Galaxy AI datacentre in Ulanqab, Inner Mongolia, a campus planned to reach 3GW.

One number worth holding onto while reading all of this. Ember's analysis implies that the cheapest megawatt for a datacentre operator may be the one it never asks the grid for. Drax is betting £20bn that this is true even when the meter sits on top of a biomass plant.

Comments 0

Sources

14
  1. 01UK's most polluting power plant could power country's largest AI datacentreEN
  2. 02Behind-the-meter generation could deliver five times more grid relief than data center load shiftingEN
  3. 03Dominion Energy reports 54.9GW of data center demand in various stages of contractingEN
  4. 04Anthropic signs lease with Australian Data Centres in Western Downs, QueenslandEN
  5. 05SpaceX loses bid to dismiss lawsuit over xAI data center pollutionEN
  6. 06Elon Musk's xAI installs eight converted jet engines to power Mississippi data centerEN
  7. 07Nvidia tells data center operators to limit business from Chinese customers with new global restrictions, report claimsEN
  8. 08China deploys 'world's first' 800-volt DC power system for massive AI data centerEN
  9. 09Over 90% of 10k public comments oppose data centers; water tops concernsEN
  10. 10Google's Minnesota data center approved after it drops $80M tax break askEN
  11. 11OpenAI considers rebranding data centers to 'AI compute campuses'EN
  12. 12Why batteries, not gas plants, should be data centers' first choiceEN
  13. 13National Grid to site biggest substation in UK near planned Teesside datacentreEN
  14. 14Ohio's Data-Center Revolt Is Upending the MidtermsEN

All figures and quotations in this text come from the sources listed below.

Content prepared by the editorial team with AI assistance.

Grace Okonkwo

Grace Okonkwo

AI, models and technology

Grace Okonkwo covers AI, models and technology for FLASH24, working from primary sources such as model cards, API documentation and benchmark papers rather than vendor summaries. She checks training data provenance, evaluation conditions and reported scores against the underlying datasets before any figure reaches print. She interviews researchers and engineers directly, tracks release calendars from major labs, and compares successive model versions on the same tests. Her own self-hosting, home-network and documentation-reading habits feed straight into that desk, since she tests tools on her own hardware first. She does not publish benchmark claims without a reproducible method.

Newsroom →

Comments

0
  1. No comments yet — be the first.

Write a comment

Comments are public. We do not publish abuse, spam or advertising.