The continued growth of artificial intelligence across the United States could result in approximately 900,000 tonnes of CO2 being added to the country’s emissions annually, according to a new study published in Environmental Research Letters.
While researchers noted this figure equates to a relatively minor 0.02 per cent increase in total US emissions, the energy impact on specific sectors is significant. The study estimates that energy use in individual industries could increase by up to 12 petajoules annually, equivalent to the electricity consumption of approximately 300,000 US homes.
“While the projected emissions from AI adoption are modest compared to other sectors, they still represent a meaningful increase,” said co-author Anthony R. Harding. “This underscores the importance of integrating energy efficiency and sustainability into AI development and deployment, especially as adoption accelerates across industries.”
The findings provide a snapshot of current adoption, but they stand in contrast to dire warnings about the future environmental costs of AI infrastructure.
Far more drastic impact
Separate research from Cornell University indicates the rapid expansion of AI data centres could have a far more drastic impact by the end of the decade. That study warned AI infrastructure could add up to 44 million tonnes of CO2 annually by 2030, equivalent to placing 10 million extra cars on the road.
The Cornell researchers also found that the industry’s water consumption could reach 1,125 million cubic metres per year, potentially putting net-zero emissions targets out of reach.
While the IOP Publishing study characterises current AI emissions as “relatively modest” compared to other industrial activities, the authors urged industry leaders to incorporate energy efficiency into their strategies immediately to ensure responsible growth as adoption scales.
