Report: Majority of AI Climate Claims Lack Verifiable Evidence

A report published on February 17 by climate nonprofits Beyond Fossil Fuels and Climate Action Against Disinformation has concluded that nearly three-quarters of recent industry and institutional claims about artificial intelligence (AI) combating climate change are not backed by verifiable evidence.

The analysis, led by energy analyst Ketan Joshi, scrutinized 154 statements from major technology companies and public institutions. According to the findings, approximately 74% of these assertions regarding the climate advantages of AI failed to provide concrete supporting data. Researchers highlighted that much of the supposed environmental benefit derives from previous generations of "leaner" machine learning models, not from highly resource-intensive generative AI systems currently powering data center expansion.

Generative AI refers to advanced systems, such as large language models and image generators, that create novel outputs from data. While earlier forms of machine learning typically required less energy, the new wave of generative models—like Google Gemini and Microsoft Copilot—necessitate significant computational resources, translating into greater energy demand and associated emissions. The report found no documented instances of mainstream, consumer generative AI tools leading to measurable emissions reductions.

The study also reviewed an International Energy Agency (IEA) report on AI's climate impact, ultimately finding an even distribution of academic research, corporate material, and unsupported assertions. It noted that the industry’s widely cited projection—that AI could cut 5-10% of global greenhouse gas emissions by 2030—remains "contested" and lacks strong evidentiary backing.

Ketan Joshi argued that the tech industry's attention to AI climate solutions diverts from the actual environmental impact of rapidly growing data centers. He warned that this framing risks obscuring the reality of rising pollution linked to expanding digital infrastructure, particularly as companies pursue ambitious AI deployment.

Environmental scrutiny of the AI sector is intensifying. Google reported a 48% rise in its greenhouse gas emissions from 2019 to 2023, attributing much of the increase to energy use in its expanded data center network. Similarly, Microsoft disclosed a 29% rise since 2020, largely connected to constructing new facilities for AI workloads.

Research cited in the report shows that energy consumption scales sharply with generative AI adoption. For example, a single query to ChatGPT reportedly uses about ten times the electricity of a conventional Google search. Projections by Goldman Sachs suggest that by 2030, data centers could account for 8% of the United States' total power usage, up from 3% in 2022—a shift primarily credited to the AI boom.

The coalition behind the report, including Stand.earth, Friends of the Earth U.S., and the Green Web Foundation, calls for more transparent reporting on energy use and higher standards for claims about environmental benefits. Advocates urge the industry to better align public statements with independently verifiable data and the sector’s stated climate commitments.

Researchers and climate groups argue that without stricter accountability, the gap between industry rhetoric and the environmental reality of AI growth risks widening further.

Reference: dataconomy.com