UK government underestimated AI datacentre carbon emissions by factor of 100, new estimates show

The UK government quietly revised its estimate of carbon emissions from AI datacentres, raising the figure from 0.142 million tonnes per year to as much as 123 million tonnes over the next decade. That is roughly equivalent to the emissions of 2.7 million people. The revision came after independent investigations by Foxglove and Carbon Brief showed the original numbers were far too low. The new estimate appears in the government's compute roadmap, which outlines plans for a major AI infrastructure buildout. AI datacentres consume far more electricity than traditional data storage centres, and much of that power still comes from fossil fuels. The lower end of the new range, 34 million tonnes, depends on efficiency gains and faster grid decarbonization. Critics point out that the government has a legally binding net zero target for 2050, yet is pushing a datacentre buildout that could double the country's electricity consumption. The episode raises questions about whether officials are properly accounting for the climate cost of AI expansion.