UK departments at odds on AI datacentre energy demand
TL;DR:
- DSIT projects 6GW of AI-capable datacentre capacity by 2030; DESNZ’s commercial services forecast implies barely a tenth of that.
- DSIT quietly raised its ten-year AI compute emissions estimate by more than a hundredfold after questions from journalists.
- The mismatch is an early sign that the government’s AI growth plan and its carbon budget are not yet built on the same numbers — a problem for any business planning around either.
Two UK government departments responsible for delivering apparently complementary strategies — making Britain an AI superpower and reaching its climate targets — are working from incompatible assumptions about how much electricity AI datacentres will use, according to documents reviewed by The Guardian.
The Department for Science, Innovation and Technology forecasts at least 6GW of AI-capable datacentre capacity by 2030 in its compute roadmap. The Department for Energy Security and Net Zero, asked separately by NGO Foxglove how it had factored AI growth into its emissions plan, referred researchers to a broader projection that the entire commercial services sector will grow by just 528MW between 2025 and 2030. That is about one-tenth of DSIT’s headline AI figure, and slightly less than the energy footprint of a single AI growth zone.
Numbers shifting under scrutiny
DSIT’s published carbon-emissions estimates for the same compute build-out moved sharply during the Guardian’s reporting. Original figures for additional AI computing capacity sat between 0.025 and 0.142 million tonnes of CO₂-equivalent — under 0.05% of projected UK emissions. After questions from the paper, DSIT updated the same numbers to a ten-year range of 34 to 123 MtCO₂, equivalent to roughly 0.9 to 3.4% of projected UK emissions. That is a hundredfold revision.
A DESNZ spokesperson said datacentre emissions were factored into modelling for Carbon Budget 7, due this summer, and pointed to the AI Energy Council. DSIT directed comment to DESNZ.
Why it matters for UK business
The gap echoes a parallel infrastructure constraint reported by Reuters this week, in which UK industrial-site owners are waiting years for the National Grid connections needed to actually power the datacentres ministers want built. Together, the two stories suggest the UK’s AI-growth strategy is currently propped up by aspirational siting and disputed energy maths, not a delivery plan.
For businesses planning AI capacity in the UK — whether building datacentres, contracting compute, or assessing decarbonisation pathways — the practical question is which department’s number ends up in the next carbon budget and the next investment guidance. Until those reconcile, AI-growth policy and net-zero policy are pointed in directions that cannot both be right.