The United Kingdom aims for a low-carbon economy driven by sustainable energy sources while also striving to become a leader in artificial intelligence. However, the government agencies overseeing these goals seem to disagree on key figures. The Department for Science, Innovation and Technology (DSIT) estimates that AI data centers will require 6 gigawatts of power by 2030. In contrast, the Department for Energy Security and Net Zero (DESNZ) projects usage at under one-tenth of that amount. Tim Squirrell, strategy director at the nonprofit Foxglove, commented: ‘The government’s lack of awareness regarding data centers’ ecological effects is concerning, though it might seem absurd.’ Cecilia Rikap, a scholar at University College London, noted: ‘This discrepancy could indicate either departmental shortcomings or unrealistic optimism about AI and major tech firms. Regardless, it reveals how these companies influence both the AI industry and UK policymaking.’ DESNZ manages the nation’s carbon targets and strategy for meeting global climate commitments. In January, Foxglove submitted a request for details on how AI data centers were included in emissions forecasts. DESNZ pointed to general predictions for the commercial services industry’s energy consumption, stating it lacked specific data center estimates. Those predictions suggest the sector’s power use will rise by 528 megawatts from 2025 to 2030, comparable to adding electricity needs for 1.7 million households. This is significantly lower than the power pledged for AI data centers in the UK’s compute strategy. Released by DSIT in 2025, the document outlines an ambitious initiative to enhance the national computing framework through AI facilities. It states: ‘We anticipate the UK will require at least 6 gigawatts of capacity for AI data centers by 2030.’ This would involve several development areas nationwide to draw investments, each needing at least 500 megawatts—nearly matching DESNZ’s total sector growth estimate. The source of the mismatch between the departments remains uncertain. But following an inquiry from the Guardian the next day, DSIT updated its online emissions data for the AI data center industry, increasing it over 100 times. Initially, DSIT’s calculations for extra AI computing emissions ranged from 0.025 million to 0.142 million tonnes of CO2 equivalent, under 0.05% of the UK’s expected output. These were in an attachment to the strategy document, which was taken down earlier this year after scrutiny from Carbon Brief. On Thursday, after the Guardian’s questions, DSIT revised the information. In an online update, it stated: ‘Cumulative UK greenhouse gas emissions from AI computing over 10 years might range from 34 to 123 million tonnes of CO2, representing about 0.9% to 3.4% of the nation’s total projected emissions during that time.’ It continued: ‘Achieving the UK’s grid decarbonization goals would help lower data center emissions to the lower part of this spectrum.’ A DESNZ representative said: ‘Emissions from data centers are included in our models, including for the seventh carbon budget,’ and mentioned: ‘The AI Energy Council is examining ways to encourage investments and promote clean energy for data centers.’ The seventh carbon budget is due this summer. A DSIT representative directed inquiries to DESNZ.
