AI’s biggest climate problem may not be data centers

The carbon footprint of artificial intelligence (AI) is more severe than initially perceived, according to a new study published in the Nature Portfolio journal npj Climate Action. While data centers are contributing to greenhouse gas emissions by fueling new fossil-fueled power plants, such as an Amazon-backed gas plant in Texas, AI is also exacerbating the production of fossil fuels.

This productivity enhancement could result in up to 13 times more emissions than data centers generate today. The study, authored by former Microsoft employees Will Alpine and Andrew Lutz, identifies "enabled emissions" – the CO2 produced when oil and gas companies utilize AI to expedite and economize extraction and refinement processes.

Additionally, AI aids in making renewable energy more productive. However, the impact on fossil fuels is significantly higher than on renewables. To maintain relatively stable global emissions, AI-driven improvements in renewable energy productivity would need to be four to five times greater than those achieved in the fossil fuel industry.

This is because even minor advancements in fossil fuel extraction can lead to substantial increases in fossil fuel consumption. The research indicates that AI-induced productivity gains in the fossil fuel sector increase global emissions by 0.47 to 1.8 gigatonnes of CO2 annually, which is equivalent to three times Germany's annual emissions.

The study's authors argue that AI, while useful for advancing renewable energy and grid efficiency, has also intensified the productivity of the fossil fuel industry, making it an economic lever that bolsters the dominance of fossil fuels.