The climate impact of artificial intelligence could be more complicated than simply measuring the electricity consumed by data centers. Researchers modeled how AI might improve productivity across both renewable and fossil-fuel energy systems and found that, across 64 scenarios, AI could ultimately increase global carbon emissions by 0.47 to 1.8 gigatonnes per year. The study argues that previous assessments often focused on AI's potential to make renewable energy and electricity grids more efficient while overlooking its ability to make oil, gas, and coal production more productive.
The biggest concern is AI's ability to make fossil-fuel extraction cheaper and more efficient. The International Energy Agency estimates that AI could increase technically recoverable oil and gas reserves by about 5% and reduce the cost of deepwater projects by roughly 10%. Energy companies are already using AI for seismic analysis, well planning, exploration, and production optimization. The Guardian reports that Rystad Energy estimates AI and digitalization could create nearly $500 billion (£370 billion) in cumulative value for fossil-fuel exploration and production companies between 2026 and 2030.
The study found that AI could produce a net climate benefit only if it did not significantly increase productivity in fossil-fuel production. If AI adoption occurs at comparable rates across fossil and renewable energy, renewable-energy productivity would need to improve at least four times faster than fossil-fuel productivity for emissions to break even. The researchers emphasize that this is a structural finding rather than a precise forecast, but they say the relationship remained consistent across their scenarios and sensitivity tests.
The findings raise an important question about whether AI's potential environmental benefits can outweigh its indirect contribution to continued fossil-fuel use. AI could improve renewable generation, optimize electricity grids, and reduce waste, but those gains may be offset if the same technology helps companies extract more oil and gas. The article therefore suggests that the climate impact of AI cannot be judged solely by making data centers more energy-efficient; policymakers and technology companies also need to consider what economic activities AI makes more productive and whether those productivity gains accelerate or slow the transition away from fossil fuels.
