The Unproven Promise: Tech Giants' AI Climate Claims Face Scrutiny

In late 2023, Google made a bold assertion: artificial intelligence could slash global greenhouse gas emissions by 5 to 10 percent within seven years. That figure, equivalent to the European Union's annual output, rippled through media and academic papers. Energy researcher Ketan Joshi found it compelling—and deeply suspicious.

His investigation traced the number to a Google-commissioned report by Boston Consulting Group, which itself cited only the firm's "experience with clients" as evidence. Joshi calls the foundation "flimsy." This analysis predated the ChatGPT-fueled infrastructure boom now consuming vast energy. Notably, Google's own 2023 sustainability report later acknowledged that its AI expansion was increasing corporate emissions, even as it continued promoting the BCG statistic to European policymakers.

Joshi's new report, supported by environmental groups, examines over 150 industry claims that AI serves as a net climate benefit. His findings: only a quarter cited academic research; more than a third offered no public evidence whatsoever.

The debate isn't just about evidence, but definition. Many claims conflate decades-old, efficient machine learning used in science with the new, energy-hungry generative AI like ChatGPT driving data center expansion. "The narrative that we need big AI models—and quasi-infinite amounts of energy—tries to sell us the idea that this is the only kind of AI we need," says researcher Sasha Luccioni.

Experts note a critical information gap. Precise data on AI's energy consumption remains scarce, with companies slow to disclose environmental impacts. While efficient AI tools already aid climate work, tech leaders often promote speculative futures powered by models that don't yet exist.

Joshi argues for transparency: if companies want their climate promises believed, they must disclose the actual energy costs of their AI ambitions. Without that basic data, their grand claims about saving the planet remain just that—claims.