Artificial intelligence (AI) systems are consuming water at levels that could rival the total global bottled water usage, according to a recent study published in the journal Patterns.
The research estimates that the electricity needed to power AI systems could indirectly require between 312.5 and 764.6 billion liters of water annually. These figures highlight a growing environmental concern as AI technologies expand rapidly across industries.
Researcher Alex de Vries-Gao, who specializes in the environmental impact of digital technologies, analyzed data from tech giants including Google, Apple and Meta. He found that much of the water consumed by AI systems is not just used for cooling data centers directly but also in the production of electricity that powers them.
He argues that the indirect water footprint has been significantly underestimated in current reporting.
Water use by AI systems may far exceed official estimates
The International Energy Agency estimated that all data centers used about 560 billion liters of water in 2023. De Vries-Gao’s analysis suggests that this figure may be three to four times lower than the actual amount when indirect water use is taken into account.
He said that the volume of water linked to AI energy demand alone is comparable to the annual consumption of bottled water worldwide.
🚨 New research warns AI systems may consume 312–765 billion liters of water yearly — potentially rivaling the entire world’s bottled water use.
Much of this isn’t just cooling data centers, but water used to generate the electricity AI needs. pic.twitter.com/GqT2RRXJN7
— Tom Marvolo Riddle (@tom_riddle2025) December 19, 2025
The study also examines carbon emissions. AI systems could account for between 32.6 and 79.7 million metric tons of carbon dioxide per year. For comparison, New York City emitted approximately 52.2 million metric tons in 2023.
De Vries-Gao said that AI’s environmental footprint could soon match that of major cities or highly energy-intensive industries.
Environmental impact of AI systems consuming water
Since the launch of ChatGPT in 2022, demand for AI tools has surged. By 2024, AI-related activities made up an estimated 15 to 20 percent of electricity usage in data centers globally. De Vries-Gao projects that AI systems may require as much as 23 gigawatts of power, a demand equal to the average electricity use of the United Kingdom.
Despite these findings, many major tech companies do not report how much energy or water their AI systems are consuming. Google, for example, declined to include indirect water use in a recent report on its Gemini AI model, stating that it does not control water usage at external power plants.
De Vries-Gao challenges this position, pointing out that companies are still responsible for the environmental consequences of the electricity they purchase, similar to how carbon emissions are reported under the Greenhouse Gas Protocol.
He is calling for stricter reporting requirements that mandate full transparency from companies about energy use, emissions and water consumption, with a separate breakdown for AI-related operations. He believes that without clear data, it will be impossible to understand or address the true cost of AI on the environment.
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
