Google’s recent claim that each Gemini AI query uses just “five drops of water” is now under scrutiny, as AI experts question the accuracy of the figure. According to The Verge, researchers argue that Google’s calculation significantly downplays the environmental impact of AI operations. Among the critics is Shaolei Ren, a University of California Riverside professor and co-author of the report Making AI Less “Thirsty”, who previously found that training OpenAI’s GPT-3 consumed around 700,000 liters of water and that a ChatGPT session with 20–50 messages could use nearly a pint of water. Ren and other experts say Google’s figures only include water used for cooling data centers, ignoring the water consumption from power plants generating the required electricity. Sustainability researcher Alex de Vries-Gao described Google’s figure as “just the tip of the iceberg,” emphasizing the larger, hidden environmental cost.
The criticism does not stop at water usage. Experts also dispute Google’s reporting of AI carbon emissions, alleging that it relies on market-based metrics that exclude electricity certificates and carbon credits. This accounting method, they argue, hides the true CO2 emissions produced by AI models. The debate underscores a growing call for transparency in the environmental impact of artificial intelligence, as companies like Google, Microsoft, and OpenAI race to build ever more powerful systems. With AI’s rapid adoption across industries, experts warn that underreporting its environmental toll risks delaying critical measures to make the technology more sustainable.
