Officials underestimated impact of AI datacentres on UK carbon emissions (www.theguardian.com)

UK officials significantly underestimated the climate impact of artificial intelligence, increasing their projections for AI-related carbon emissions by over 100 times. Recent estimates indicate that AI datacentres could generate up to 123 million tonnes of CO₂ over the next decade, equivalent to emissions produced by 2.7 million people. This stark revision replaces a prior, now-retracted estimate of just 142,000 tonnes in a single year, raising concerns about the environmental cost of the UK’s ambitious AI infrastructure plans. The implications for the AI and ML community are profound, as the energy demanded by AI datacentres is vastly higher than traditional data storage facilities, largely relying on fossil fuels. With the potential emissions accounting for 0.9% to 3.4% of the UK’s total emissions from 2025 to 2035, this raises urgent questions about the sustainability of AI advancements. Critically, experts emphasize the need for improved efficiency in AI technologies and a faster transition to decarbonized energy sources if the AI sector wishes to align with the UK’s legally binding net-zero targets for 2050. Failure to address these challenges could lead to a historic failure to balance technological progress with environmental responsibility.