Google Pulls Back the Curtain on AI’s Energy Footprint

Google has taken an unpre

Google has taken an unprecedented step in the debate over artificial intelligence’s environmental impact by revealing detailed figures on how much energy its Gemini AI consumes per query. The company estimates that the “median” text prompt requires 0.24 watt-hours of electricity—about the same as running a microwave for one second, watching TV for nine seconds, or using five drops of water.

While the number may sound small, the disclosure is significant. With billions of prompts processed daily across AI systems worldwide, even modest per-query energy costs can translate into major ecological footprints. Google’s announcement marks one of the first times a leading AI company has provided hard numbers on energy use at scale.

Transparency with Caveats

The data, published alongside a technical paper, highlights important gains: since May 2024, Gemini’s per-prompt energy use has fallen 33-fold, and its carbon footprint has dropped 44-fold. Google credits improvements across hardware, model design, and data center efficiencies.

But the headline number comes with limits. The figure represents only text prompts, not more computationally demanding tasks like image or video generation. It also excludes the energy-intensive training phase of large models, which typically requires far more electricity and water than day-to-day usage.

“On its own, 0.24 watt-hours seems trivial, but scaled to billions of interactions, the footprint is anything but,” said Skye King, a behavioral scientist and AI consultant. “We need to think carefully about how this shapes our everyday behaviors.”

Expert Reactions

The release has drawn praise for transparency while also sparking calls for industry-wide standards.

• Casey Crownhart, climate reporter at MIT Technology Review, called it “the most transparent look yet behind the curtain” from a major AI company.

  PublicistBot Opportunity

  Turn one article into an ongoing visibility engine.

  PublicistBot can keep your business visible with monthly articles, search-friendly distribution, affiliate-aware sharing, and optional AI Marketer video content.

  Activate PublicistBotAdd AI Marketer Videos

• Mairin Loewen, of Urban Climate Leadership, noted the data accounts not only for AI chips but for supporting infrastructure like CPUs and cooling systems—a fuller view than previously available.

• Amy Wright, a tech and policy advisor, cautioned that without standardized reporting frameworks, “these numbers risk becoming more of a marketing headline than a sustainability benchmark.”

Others see the disclosure as a benchmark for corporate decision-making. “AI adoption is happening in boardrooms without clear data on environmental costs,” said developer Artem Bredikhin. “Now we finally have a baseline to work with.”

The Bigger Picture

Google’s Senior VP James Manyika emphasized that the company is still “just at the start” of improving AI efficiency, pointing to investments in renewable energy, geothermal projects, and next-generation nuclear power.

For sustainability advocates, the move is a milestone. Greenpeace IT manager Clíodhna Kirk called it “first glimmers of transparency” from Big Tech on climate impact. Hugging Face researcher Dr. Sasha Luccioni welcomed the disclosure but stressed it is “not a replacement or proxy for standardized comparisons,” urging Google and others to join collaborative efforts like the AI Energy Score project.

Why It Matters

AI is now a core part of search, productivity, and creative tools used by millions daily. Knowing the energy cost of each interaction helps policymakers, businesses, and individuals weigh trade-offs between innovation and sustainability.

As Amy Wright put it, the challenge isn’t celebrating smaller footprints per prompt—it’s ensuring efficiency and transparency keep pace with adoption.