UK AI Data Centre Emissions Estimate Jumps from 0.14 to 123 Million Tonnes CO2

Context The UK’s Department for Science, Innovation and Technology (DSIT) initially projected AI data centre emissions to peak at just 0.142 million tonnes of CO2 in a single year. This earlier figure served as the basis for initial assessments of the UK’s AI compute strategy and its climate impact. The original estimate has since been removed from official documentation, following scrutiny.

This revised forecast presents a significant increase in the anticipated carbon output from AI infrastructure. The update challenges the UK's climate targets and necessitates a re-evaluation of energy infrastructure planning assumptions across the country.

Key Facts The UK government now projects AI data centers could emit up to 123 million tonnes of CO2 between 2025 and 2035. This updated figure marks a dramatic increase from earlier calculations. The Telegraph characterized the discrepancy starkly, reporting the original estimate was off by 136,000 percent.

Tim Squirrell, Strategy Director at the non-profit organization Foxglove, stated that the situation is “far graver than initially thought.” He criticized the government, noting it appears to have lacked “basic calculations” before committing to the rapid expansion of hyperscale data centers. Hyperscale data centers are large-scale facilities designed to support massive data processing and storage needs, often for cloud computing or artificial intelligence applications.

This revised projection fundamentally shifts the anticipated environmental footprint of AI development. It raises significant questions regarding the thoroughness of the initial impact assessments conducted by the government.

What It Means The substantial revision in emissions figures introduces new complexities for grid infrastructure planning and renewable energy procurement. Energy planners, local councils responsible for granting planning permissions, and grid operators must now reassess their long-term investment strategies based on these considerably higher projections. Data centers, especially hyperscale facilities, demand significant electricity, placing considerable strain on both local and national energy grids.

Future planning approvals for new energy-intensive facilities, particularly data centers, will likely face heightened scrutiny. Carbon offset requirements, currently less stringent, may also tighten as DSIT works on more accurate assessments. Companies building or operating AI infrastructure in the UK will need to integrate these evolving regulatory demands into their operational and financial models. The government's actions to reconcile its AI ambitions with its legally binding net-zero 2050 obligation will be a critical area to monitor moving forward.