Particle.news

Download on the App Store

Google Publishes Per-Query Energy, Water and CO₂ Metrics for Gemini Text Queries

Google confines the metrics to text queries, leaving the broader AI footprint unresolved.

Image
Image

Overview

  • Google reports median figures per Gemini text request of 0.24 Wh electricity, 0.26 ml water and 0.03 g CO₂-equivalent.
  • The company issued a detailed measurement methodology that counts chip utilization, memory, idle capacity and data-center cooling, which it proposes as an industry standard.
  • Google says median energy per text request fell about 33× over the past year and the CO₂ footprint about 44×.
  • The disclosure excludes image and video requests and does not include the energy required to train the models.
  • Comparisons highlight differing scopes, with OpenAI citing about 0.34 Wh per query and researchers and agencies warning of rising overall electricity and water demand for AI data centers.