The United Kingdom's ambitions to become an artificial intelligence superpower are clashing with its net zero goals, as government departments present starkly different forecasts for the energy consumption of AI datacentres.
Conflicting Projections
The Department of Science, Innovation and Technology (DSIT) projects that AI datacentres will require at least 6 gigawatts (GW) of electricity capacity by 2030. In contrast, the Department of Energy Security and Net Zero (DESNZ) appears to estimate less than one-tenth of that figure, based on its broader forecasts for the commercial services sector.
This discrepancy has raised concerns about the coherence of government planning. Tim Squirrell, head of strategy at the NGO Foxglove, described the situation as alarming. Cecilia Rikap, a researcher at University College London, suggested that the misalignment indicates either incompetence or a form of magical thinking regarding big tech's influence over the government.
Environmental Impact Assessments
In January, Foxglove requested an environmental impact assessment from DESNZ, asking how AI datacentres were incorporated into emissions projections. DESNZ responded that it does not hold separate projections for datacentre growth and instead referred to forecasts for the entire commercial services sector. Those forecasts show an increase of 528 megawatts (MW) in energy use between 2025 and 2030, equivalent to powering 1.7 million homes. This is roughly one-tenth of the capacity DSIT has committed to for AI datacentres under its UK compute roadmap.
The compute roadmap, published by DSIT in 2025, outlines plans to build AI datacentres in multiple AI growth zones across the country, each requiring at least 500 MW of electricity. This amount is nearly equal to DESNZ's projected increase for the entire commercial services sector.
Revised Emissions Figures
After the Guardian requested comments from both departments, DSIT revised its emissions estimates for AI datacentres. Originally, the department's projections for additional AI computing capacity were between 0.025 million and 0.142 million tonnes of carbon equivalent (MtCO₂), less than 0.05% of the UK's projected emissions. These figures were removed from the government website earlier this year after Carbon Brief raised questions. Following the Guardian's inquiry, DSIT updated its numbers, stating that cumulative 10-year greenhouse gas emissions from AI compute could range from 34 to 123 MtCO₂, representing 0.9% to 3.4% of the UK's projected total emissions over that period.
DSIT noted that successful grid decarbonisation could help reduce emissions towards the lower end of that range. A spokesperson for DESNZ said datacentre emissions are factored into their modelling for carbon budget 7, which is due to be released this summer. The AI Energy Council is exploring opportunities to attract investment and support clean power for datacentres.
