AI Boom Could Add Nearly a Million Tons of CO₂ Annually, Researchers Warn
Summary
A recent study in Environmental Research Letters reveals that increased AI adoption in the U.S. could raise annual carbon emissions by approximately 900,000 tons, underscoring the urgent need for energy-efficient AI technologies to mitigate environmental impact.
Key Insights
What does 900,000 tons of CO₂ emissions represent in the context of total U.S. emissions?
The projected 896,000 to 900,000 tonnes of additional CO₂ emissions from AI adoption across the U.S. economy equals approximately 0.02% of total national emissions, a relatively modest increase compared to other sectors.
What is a petajoule and how does it relate to AI's energy use?
A petajoule (PJ) is a unit of energy equal to 10^15 joules, roughly the annual electricity consumption of 300,000 U.S. homes. The study estimates AI adoption could increase energy use in individual industries by up to 12 petajoules annually.
Sources:
[1]