UK AI datacentre emissions revised up to 123 million tonnes – over 100× earlier estimate

The UK government has drastically revised its estimates for carbon emissions from AI datacentres, projecting up to 123 million tonnes of CO₂ over the next decade—a more than 100-fold increase from previous figures. This raises significant questions for the nation's climate commitments.

The UK’s strategic push into artificial intelligence now projects a substantially higher climate cost. The "compute roadmap," a government initiative to develop a world-class AI ecosystem within the UK, recently received a quiet but significant revision to its expected emissions. This roadmap outlines the government's plan for AI delivery, a sector on which it stakes hopes for economic growth.

The Department for Science, Innovation and Technology (DSIT) now estimates UK AI datacentres could emit up to 123 million tonnes of CO₂ over the next 10 years. This new figure, revealed in an update to the compute roadmap, equates to the annual emissions of 2.7 million people. It starkly contrasts with an earlier government estimate, since withdrawn, which projected a maximum of 0.142 million tonnes of CO₂ per year from AI datacentres.

AI datacentres, which power complex AI models, demand vast amounts of electricity, significantly more than facilities for standard data storage. Most of this energy in the UK continues to come from fossil fuels. The DSIT’s revised projections indicate the planned AI buildout could account for 0.9% to 3.4% of the UK’s total projected emissions between 2025 and 2035. Achieving the lower end of this range hinges on greater efficiency in AI models and hardware, alongside a faster decarbonisation of the nation’s energy grid.

This dramatic increase in projected emissions highlights the substantial energy footprint of rapidly expanding AI infrastructure. The revision follows investigations by independent watchdog Foxglove and the Carbon Brief news site, which had highlighted earlier figures as significant underestimates. Patrick Galey, head of investigations for the Global Witness climate campaign, voiced sharp criticism, stating, "wasting limited carbon resources on AI for the wealthy is a historic idiocy that future generations will not forgive."

These updated figures challenge the UK government's legally binding commitment to reach net-zero emissions by 2050. The scale of the increase warrants scrutiny regarding the environmental trade-offs of rapid technological advancement. Observers will now watch how the UK government addresses these revised emissions estimates and navigates its continued commitment to both AI development and its crucial climate targets.