How much energy does Google’s AI use? We did the math
Summary
Google shares its methodology for measuring energy, emissions, and water impact of AI inference, estimating median Gemini prompt energy use at 0.24 Wh, below prior estimates.
Key quotes
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 of water — substantially lower than many public estimates.
Over a recent 12-month period, the energy and total carbon footprint of the median Gemini Apps text prompt dropped by 33x and 44x, respectively, while delivering higher-quality responses.
Measuring the footprint of AI serving workloads isn’t simple... Many current AI energy consumption calculations only include active machine consumption, overlooking several critical factors.
Our comprehensive methodology’s estimates account for *all* critical elements of serving AI globally, including idle machines, CPU, RAM, data center overhead, and water consumption.
Google’s article details a technical paper outlining its methodology for assessing the environmental impact of AI inference, specifically for its Gemini model. Unlike previous estimates, Google’s approach accounts for idle machine energy, CPU/RAM usage, and data center overhead, resulting in a more comprehensive assessment. The company reports significant efficiency improvements, reducing the median text prompt’s energy and carbon footprint by 33x and 44x, respectively, over the past year.