Big Tech’s AI Climate Claims Lack Proof
▼ Summary
– Google claimed AI could reduce global greenhouse gas emissions by 5-10% by 2030, a statistic widely circulated in media and academic papers.
– Energy researcher Ketan Joshi traced this claim to a Google-BCG paper that cited only “experience with clients” as evidence, which he found to be a flimsy basis.
– Despite the AI build-out increasing Google’s own corporate emissions, the company continues to promote the BCG-derived statistic to policymakers.
– The rapid development of AI is driving significant energy demand, leading to extended coal plant use and plans for massive new gas power capacity for data centers.
– Tech executives, including from Google and OpenAI, argue AI’s potential climate benefits justify its energy costs, framing it as a solution to environmental problems.
The bold promises made by major technology firms regarding artificial intelligence’s potential to solve the climate crisis are increasingly under scrutiny, with experts questioning the evidence behind the optimistic projections. A prominent example centers on Google’s widely circulated claim that AI could slash global greenhouse gas emissions by 5 to 10 percent by 2030. This assertion, which suggests eliminating an amount of pollution equivalent to the European Union’s annual output, has been repeated in press articles and policy discussions, yet its origins appear surprisingly fragile.
Energy researcher Ketan Joshi investigated the source of this statistic and found it traced back to a paper co-authored by Google and the Boston Consulting Group (BCG). That document, in turn, referenced a 2021 BCG analysis. The foundational basis for the massive emissions reduction estimate was not a peer-reviewed scientific model but rather the consulting firm’s “experience with clients.” Joshi described this as a flimsy justification, especially given that the analysis predated the explosive, energy-intensive growth in AI infrastructure triggered by tools like ChatGPT.
The timing of these claims is particularly striking. Just months after promoting the 5 to 10 percent figure, Google’s own 2023 sustainability report acknowledged a sobering reality: the company’s corporate emissions were rising significantly, driven precisely by its AI expansion. Despite this internal data, Google has continued to promote the BCG-derived numbers, even using them in a recent memo to European policymakers. This dual narrative, publicly championing AI as a climate savior while privately documenting its growing carbon footprint, highlights a significant contradiction.
Google maintains that its methodology is sound. A company spokesperson stated, “We stand by our methodology, which is grounded in the best available science,” and pointed to publicly available principles for calculating emissions reductions. However, the company did not clarify how these standards were specifically applied to validate the BCG figures. BCG itself did not respond to inquiries about the analysis.
This situation unfolds against a backdrop of frantic competition in AI development, a race with profound climate implications. In the United States, the world’s largest data center market, the surge in demand for computing power is directly impacting energy grids. The push for more AI capacity is leading utilities to keep coal plants operational and plan for hundreds of gigawatts of new natural gas power, with a substantial portion dedicated solely to feeding data centers.
Nevertheless, tech leaders consistently argue that this massive energy investment is justified by AI’s future environmental benefits. At high-profile forums like Climate Week in New York, discussions have centered on AI as an “environmental force for good.” Former Google CEO Eric Schmidt has suggested that since the world is lagging on climate targets, the focus should shift to betting on AI as the solution. Similarly, OpenAI’s Sam Altman has publicly promised that AI will ultimately “fix” the climate. These visionary statements, however, continue to outpace the concrete, verifiable evidence needed to support such transformative claims.
(Source: Wired)
