Google's 5-10% AI Emissions Cut Claim Built on 'Flimsy' Source
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google's 5-10% AI emissions reduction claim by 2030 was coauthored by its chief sustainability officer in late 2023 and spread via op-eds and academic papers, but energy researcher Ketan Joshi traced the figure to a Google/BCG paper that cited 'experience with clients' — a basis he called 'flimsy.'
- Joshi's new report, released Monday with support from several environmental organizations, examined more than 150 claims about AI serving as a net climate benefit and found just a quarter were backed by academic research, while more than a third cited no public evidence at all.
- Google's own 2023 sustainability report acknowledged the AI build-out was significantly driving up its corporate emissions, yet the company continued citing the BCG numbers — most recently in a memo to European policymakers.
- Joshi's analysis found that nearly all the claims conflated less energy-intensive traditional AI (like machine learning used in scientific disciplines for decades) with the consumer-focused generative AI — tools like ChatGPT, Claude, and Google Gemini — actually driving the data center buildout.
- David Rolnick, chair of Climate Change AI, told WIRED the amount of speculation about what generative AI 'might do' in the future is 'grotesque,' criticizing Big Tech for relying on 'hypothetical AI that does not exist now.'
- A separate study from Sasha Luccioni and Yacine Jernite of Hugging Face, also published Monday, found that smaller AI models often perform as well as massive proprietary ones on applications, challenging the 'bigger is better' narrative.
- Joshi proposed a simple fix: companies should disclose energy growth broken down by generative AI use, e.g., 'our energy growth this year was six terawatt-hours, and two of them were for generative AI.'
Why it matters: Google cited the 5-10% statistic to European policymakers while simultaneously acknowledging in its 2023 sustainability report that its AI build-out was significantly increasing corporate emissions — a contradiction that matters as nearly 100 gigawatts of new gas power is earmarked for US data centers alone, making the evidentiary standard for AI climate claims a direct climate policy question rather than an academic one.
Ask SkimNews
