Study questions claims AI will solve the climate crisis

A comprehensive new study has cast doubt on claims by major tech companies and industry advocates that artificial intelligence will solve the climate crisis it's helping to create. The research, funded by climate action groups and conducted by energy analyst Ketan Joshi, examined 154 claims about AI's potential climate benefits and found them largely unsupported by evidence.

The energy consumption problem

The study comes amid growing concerns about AI's environmental impact. As tech giants rush to build massive datacenters to power increasingly energy-hungry AI systems, utilities are turning to fossil fuels to meet the surging demand. This creates a paradox: the very technology being touted as a climate solution is driving increased carbon emissions through its massive energy requirements.

Former Google CEO Eric Schmidt and Microsoft cofounder Bill Gates have both argued that AI's long-term climate benefits could outweigh or offset the emissions from datacenters. The International Energy Agency has also supported similar claims. However, Joshi's research suggests these assertions may amount to greenwashing rather than genuine climate solutions.

Traditional AI vs. generative AI confusion

One of the study's key findings is that most climate benefit claims conflate traditional AI systems with generative AI. Traditional AI includes predictive models and computer vision applications that have been in use for years. Generative AI, which includes systems like ChatGPT, Gemini, and Copilot that create text, images, and other content, is far more energy-intensive.

"This analysis found that the overwhelming majority of AI climate benefit claims relate to 'traditional' forms of AI rather than generative AI," Joshi wrote. Only four of the 154 claims examined related to generative AI systems, despite the fact that most of the projected energy consumption from AI will come from these newer, more demanding systems.

Lack of evidence

The study found that most claims about AI's climate benefits lack proper citations. Only 26 percent of the claims cited published academic papers, while 36 percent had no citations whatsoever. Corporate publications made up 29 percent of cited sources, but these typically didn't include primary evidence or peer-reviewed research to support their claims.

Joshi noted that even the narrower, older forms of AI may have their climate benefits overstated, given the lack of strong, peer-reviewed evidence showing their deployment has actually reduced greenhouse gas emissions in real-world applications.

Industry response

When contacted for comment, the International Energy Agency didn't respond. Microsoft had no comment, while Google defended its methodology. "We stand by our methodology, which is grounded in the best available science," the company stated, adding that it is transparent about the principles guiding its claims.

Growing public skepticism

Joshi told The Register that he was motivated to undertake the research after hearing Google persistently claim its AI deployments would reduce emissions by 5 to 10 percent based on what he characterized as "mind-bogglingly weak evidence."

He suggested that the tech industry's AI push is built more on hype than effectiveness, and that public opinion is increasingly turning against the massive datacenter developments. "There is widespread public skepticism for what all of this is actually for, skepticism of the tech industry and bipartisan hostility towards destructive data centre developments," Joshi said.

Several U.S. states are considering moratoriums on new datacenter projects, and companies like Elon Musk's xAI have faced public backlash and regulatory scrutiny over their energy practices. The study suggests that as awareness grows about the gap between AI's promised climate benefits and its actual environmental impact, the industry's narrative may face increasing challenges.

The research ultimately concludes that the narrative of AI providing gigatonne-scale emissions reductions to offset the harm of generative AI is not supported by the available evidence, calling into question whether the technology can truly be part of the climate solution while simultaneously driving increased fossil fuel consumption.