Measuring the environmental impact of AI inference
Summary
Google details a comprehensive methodology for measuring the energy, emissions, and water impact of Gemini prompts, reporting specific per-prompt resource consumption and efficiency gains.
Key quotes
we estimate the median Gemini Apps text prompt uses 0.24 watt-hours (Wh) of energy, emits 0.03 grams of carbon dioxide equivalent (gCO2e), and consumes 0.26 milliliters (or about five drops) of water
the energy and total carbon footprint of the median Gemini Apps text prompt dropped by 33x and 44x, respectively
The post outlines Google’s ‘full-stack’ approach to AI efficiency, including custom TPU hardware, Mixture-of-Experts (MoE) architectures, and data center optimizations. It argues that comprehensive measurement must include idle machines and data center overhead to avoid underestimating the operational footprint.