The environmental footprint of artificial intelligence may be much larger than the electricity and emissions associated with data centers alone. The research, published in npj Climate Action, examines how AI is being adopted by the fossil-fuel industry and argues that AI can increase the industry's ability to discover, extract and produce oil and gas. In other words, AI isn't only consuming energy—it can also be used to expand activities that generate additional emissions.
The researchers focus on AI's ability to improve exploration, drilling, reservoir analysis and operational efficiency. These applications can help companies identify promising deposits, optimize extraction and reduce the time and cost required to produce hydrocarbons. Greater efficiency sounds environmentally positive on its own, but the paper points to a potential rebound effect: if AI makes fossil-fuel production more profitable, companies may produce and sell more oil and gas rather than simply producing the same amount more efficiently.
This creates an important contrast with AI's potential contribution to clean energy. AI can help optimize electricity grids, forecast renewable generation, improve battery systems and accelerate scientific research. But those benefits have to be compared with AI applications that make fossil-fuel extraction easier. The paper's argument is that the latter effect may currently be substantially larger. That means evaluating AI's climate impact solely by measuring data-center electricity consumption can miss an important part of the picture: what AI enables its users to do.
The issue also fits into the broader environmental debate surrounding the AI boom. New research on AI data centers finds that their impacts depend heavily on local electricity mixes, water availability, land use and infrastructure constraints, while efficiency improvements can reduce—but not eliminate—those effects. The larger lesson is that AI's environmental balance cannot be determined by asking only “How much energy does AI consume?” We also need to ask “What economic activities does AI make cheaper, larger or more profitable?” If AI accelerates both renewable-energy deployment and fossil-fuel extraction, its ultimate climate impact will depend heavily on which applications scale fastest and how governments regulate the resulting incentives.
