Recent analysis suggests that the environmental impact of artificial intelligence is less significant than many assume, and in some cases, AI may even deliver ecological and economic benefits. A new study challenges the prevailing belief that rapid AI growth inevitably leads to a significant increase in global greenhouse gas emissions. Instead, the authors argue that AI’s overall contribution to global energy demand is modest, and that its potential to support cleaner technologies could outweigh its direct energy use.
The research, undertaken by teams at the University of Waterloo and the Georgia Institute of Technology, assessed the scale of energy consumption associated with AI within the United States. To do this, the researchers combined detailed economic data with estimates of AI adoption across sectors. By modelling how AI-driven automation might expand over the coming years, they aimed to understand the broader effects on energy demand and emissions. Their approach treated AI as part of the wider economic system, rather than isolating it as a standalone technology, allowing them to capture its indirect influences on productivity and energy use.
A key finding of the study is that, although AI in the United States consumes an amount of electricity comparable to Iceland’s total energy use, this still represents only a tiny fraction of national and global energy consumption. The U.S. economy remains heavily dependent on fossil fuels—around 83 per cent of national energy demand is met by petroleum, coal, and natural gas. Against this backdrop, the additional electricity required to power AI systems barely registers globally. However, the researchers emphasise that this does not mean the impact is uniform. AI-related energy use is concentrated in regions where data centres are located, and these areas may face significant local grid pressures.
Dr Juan Moreno-Cruz of the University of Waterloo notes that local communities hosting data centres could experience a doubling of electricity demand, particularly in areas where power generation still relies heavily on fossil fuels. While the national impact remains minimal, these local effects warrant careful attention, as they may lead to increased emissions and strain on infrastructure. The researchers also note that their work did not examine these local economic and environmental dynamics in detail, highlighting an area for future study.
Despite these concerns, the study offers a more optimistic viewpoint for those worried that AI represents a new and unavoidable source of carbon emissions. The authors argue that AI can play a constructive role in advancing environmental progress. It has the potential to accelerate the development of green technologies, improve energy efficiency, optimise industrial processes, and support climate modelling and mitigation strategies. Rather than viewing AI as inherently damaging, the researchers suggest that its broader benefits outweigh its direct energy requirements when deployed responsibly.
To deepen their understanding, the authors plan to extend their analysis to other countries, recognising that the effects of AI adoption will vary depending on national energy systems and economic structures. Countries with cleaner electricity grids may see even lower environmental impacts from AI growth, while those reliant on coal or gas may face more significant challenges. By comparing different regional contexts, future research could provide a more comprehensive picture of AI’s global environmental footprint.
Overall, the study invites a more measured and evidence-based discussion of AI’s environmental role. While acknowledging that localised impacts matter, it challenges the notion that AI poses a significant threat to climate goals. Instead, it proposes that AI could become an essential tool in the transition to a more sustainable future—provided its growth is managed with foresight and attention to regional conditions.
More information: Anthony Harding et al, Watts and bots: the energy implications of AI adoption, Environmental Research Letters. DOI: 10.1088/1748-9326/ae0e3b
Journal information: Environmental Research Letters Provided by University of Waterloo