London: The Department for Science, Innovation and Technology (DSIT) estimates that AI datacentres could require up to 6GW of electricity by 2030. In contrast, the Department for Energy Security and Net Zero (DESNZ) projects demand at less than one-tenth of that figure.
Tim Squirrell, head of strategy at the NGO Foxglove, criticised the lack of clarity, warning that the government’s understanding of datacentres’ environmental impact is deeply concerning.
Cecilia Rikap, a researcher at University College London, suggested the discrepancy points either to incompetence or unrealistic assumptions about AI and big tech influence, raising questions about corporate sway over policymaking.
DESNZ is responsible for setting the UK’s carbon budget and delivering plans to meet international climate targets. Earlier this year, Foxglove requested an environmental impact assessment from DESNZ regarding how AI datacentres were included in emissions forecasts.
The department responded that it does not maintain separate projections for datacentres, instead including them within broader “commercial services” energy use estimates.
These forecasts suggest the entire sector’s energy demand will rise by 528MW between 2025 and 2030, equivalent to the consumption of around 1.7 million homes. However, this figure is roughly ten times lower than the electricity demand outlined in DSIT’s 2025 ‘compute roadmap,’ which envisions a nationwide expansion of AI datacentres.
According to DSIT, the UK will need at least 6 GW of AI-capable data centre capacity by 2030, delivered through multiple ‘AI growth zones.’ Each hub is expected to require at least 500MW of power, nearly matching DESNZ’s projected increase for the entire commercial services sector.
The source of the discrepancy between the two departments remains unclear. However, shortly after media inquiries, DSIT revised its published estimates for the emissions generated by AI datacentres, increasing them significantly.
Previously, DSIT had projected additional emissions of just 0.025 to 0.142 million tonnes of carbon equivalent (MtCO₂), representing less than 0.05 percent of the UK’s total emissions. These figures were later removed following scrutiny.
Updated estimates now suggest cumulative emissions from AI computing could range between 34 and 123 MtCO₂ over a decade, accounting for approximately 0.9 percent to 3.4 percent of national emissions. DSIT noted that successful decarbonisation of the UK’s power grid could help keep emissions closer to the lower end of this range.
A DESNZ spokesperson stated that datacentre emissions are included in its modelling, including projections for the upcoming carbon budget framework. The department also highlighted ongoing work by the AI Energy Council to explore clean energy solutions and attract investment into sustainable datacentre infrastructure.
The UK’s next carbon budget framework, known as Carbon Budget 7, is expected to be released later this year.
MOST READ | Tech layoffs rise as Meta, Microsoft double down on AI
