Google publishes data on energy consumption of its Gemini chatbot for the first time

Google has published a report on the environmental footprint of its Gemini chatbot, in which it has for the first time disclosed estimates of energy consumption, CO₂ emissions and water use when processing requests.

This is reported by ZDNet.

The report takes into account not only the resources of processors and graphics chips during AI operation, but also "background" consumption - the operation of RAM, cooling systems and idle machines, ready for peak loads. According to Google, the real footprint of Gemini significantly exceeds the simplified figures, but is still lower than previously public forecasts.

Over the past year, the average footprint of one text query has decreased 33 times in energy consumption and 44 times in CO₂ emissions due to algorithm optimization and "speculative decoding" technology.

Previously, OpenAI and Meta provided only general estimates without methodology, which caused criticism from environmentalists and researchers. Experts say the rise of AI services is putting additional strain on data centers' energy and water resources, even if the footprint of a single query is small.

In June of this year, OpenAI CEO Sam Altman published on his blog specific calculations of the energy and water usage of a single ChatGPT query. This was the first public example of assessing the environmental impact of AI, which is often considered "intangible."