UK government massively underestimated AI datacentre carbon emissions
The Department for Science, Innovation and Technology revised its climate impact estimates this week, increasing projected carbon emissions from AI datacentres by more than 100 times. The new figures show energy use from these facilities could emit up to 123 million tonnes of CO₂ over the next decade - equivalent to the annual emissions of 2.7 million people.
The previous estimate, since deleted, claimed maximum annual emissions would reach 0.142 million tonnes of CO₂. Officials made the revision after independent watchdogs flagged the original calculation as a significant underestimate.
What the revised figures mean for UK policy
The updated estimate places potential AI datacentre emissions between 34 million and 123 million tonnes of CO₂ between 2025 and 2035. This represents 0.9% to 3.4% of the UK's projected total emissions over that period.
The lower end of the range assumes faster improvements in AI hardware efficiency and quicker decarbonisation of Britain's energy grid. The higher end reflects current energy infrastructure constraints.
AI datacentres consume far more electricity than conventional data storage facilities. Most UK electricity still comes from fossil fuels, making the carbon cost substantial.
The policy contradiction
The government has committed to reaching net zero by 2050 under legally binding targets. Simultaneously, it has pursued an aggressive buildout of hyperscale AI datacentres as part of its economic growth strategy.
Tim Squirrell, head of strategy at Foxglove, an independent watchdog, said: "The government doesn't seem to have done even the most basic arithmetic needed to measure the potential new carbon emissions of these datacentres."
Unchecked expansion of AI infrastructure could double the UK's total electricity consumption, according to analysis by Foxglove and Carbon Brief.
The government declined to comment on the record.
For government officials managing AI policy, understanding these environmental trade-offs is essential. Resources like the AI Learning Path for Policy Makers provide frameworks for evaluating AI infrastructure decisions against climate commitments.
Your membership also unlocks:
