Energy researcher Joshi recently exposed a startling lack of empirical evidence behind Big Tech’s claims that artificial intelligence will solve the climate crisis, revealing that core environmental metrics used by giants like Google rest on “flimsy” data. A new report released Monday, which analyzes over 150 industry assertions, found that nearly 40% of claims regarding AI’s “net climate benefit” provide no public evidence, while only 25% cite peer-reviewed academic research.
The Flimsy Foundation of “Green AI” Metrics
Joshi’s investigation began when he scrutinized a recurring statistic: the claim that AI could mitigate the equivalent of the European Union’s annual emissions. He traced this 5 to 10 percent reduction figure back to a report by Google and the Boston Consulting Group (BCG). Upon further inspection, the source of this massive estimate was a 2021 BCG analysis that cited nothing more than the firm’s “experience with clients.”
Despite the introduction of energy-intensive tools like ChatGPT and Google’s own admission in its 2023 sustainability report that AI infrastructure is driving up corporate emissions, the company continues to utilize these disputed figures. Google spokesperson Mara Harris defended the methodology, claiming it is “grounded in the best available science,” yet the company has not clarified how it applied these standards to the speculative BCG data.
Massive Energy Demand Revives Fossil Fuel Use
The aggressive race to dominate the generative AI market carries heavy environmental consequences. In the United States—the world’s largest data center hub—the surge in energy demand is forcing coal plants to remain operational. Grid operators are now queuing hundreds of gigawatts of new gas power, with approximately 100 gigawatts specifically earmarked to sustain data center expansions.
Industry leaders maintain a “growth-at-all-costs” stance. Former Google CEO Eric Schmidt recently suggested that because the world will likely miss its climate goals, society should bet entirely on AI’s problem-solving capabilities rather than imposing regulatory constraints. Similarly, OpenAI CEO Sam Altman has frequently asserted that AI will eventually “fix” the climate, though these promises lack concrete technical roadmaps.
The Deceptive Conflation of AI Technologies
A critical issue identified in Joshi’s report is the intentional blurring of lines between different types of technology. Tech companies often showcase the benefits of traditional, low-energy machine learning—such as algorithms that predict floods or discover new species—to justify the massive energy consumption of consumer-facing generative AI like Gemini or Claude.
David Rolnick, a computer science professor at McGill University and chair of Climate Change AI, describes the speculation surrounding generative AI’s future benefits as “grotesque.” He emphasizes that a significant mismatch exists between the energy-hungry models Big Tech promotes and the efficient, specialized tools actually delivering climate results today. “The narrative that we need big AI models tries to sell us the idea that this is the only future possible,” adds sustainability researcher Sasha Luccioni.
A Call for Radical Transparency and Efficient Models
Research from Luccioni and Hugging Face’s Yacine Jernite suggests that “bigger” is not always “better.” Their findings indicate that smaller, more efficient models often match the performance of proprietary giants at a fraction of the environmental cost. These experts argue that the current AI arms race primarily benefits companies with the “deepest pockets” who have harvested global data to sell back to the public.
To bridge the information gap, researchers are demanding full disclosure of the climate costs associated with AI development. Joshi argues that if tech companies believe the environmental impact of generative AI is being exaggerated, they should provide specific data. Mandatory disclosure of energy growth—specifically detailing how many terawatt-hours power generative models versus traditional operations—would provide the clarity needed to evaluate whether AI is truly a planetary savior or a climate liability.
