AI's climate impact extends beyond data centers

The carbon footprint of artificial intelligence (AI) may be far larger than previously understood. A new study suggests that AI's role in enhancing the productivity of the fossil fuel industry could generate up to 13 times more greenhouse gas emissions than the data centers required to run AI systems.

The study, published in the journal npj Climate Action, focuses on "enabled emissions." These are emissions created when oil and gas companies utilize AI to accelerate and reduce the cost of extracting, refining, and producing fossil fuels.

While AI can also improve the efficiency of renewable energy production, the research found that its impact on emissions from fossil fuels far outweighs the benefits for renewables. To maintain global emissions at current levels, renewable energy productivity would need to improve substantially more than fossil fuel productivity.

The study estimates that productivity gains from AI could increase global emissions by 0.47 to 1.8 gigatons of CO2 annually. Emissions from productivity gains in the fossil fuel sector are calculated to be 3.3 to 13.3 times greater than those from data centers today. The research was conducted by authors who previously worked on AI platforms and have since founded the Enabled Emissions Campaign.