AI may be using the power of New York and as much water as we drink from bottles in a year

Embargoed until: Publicly released: 2025-12-18 03:00

International

CC-0. Story by Dr Joe Milton, Australian Science Media Centre

We know artificial intelligence (AI) is energy and water hungry, but how hungry? A Dutch scientist says that, while figuring out the power and water use of chatbots like ChatGPT is hard, looking at metrics the AI firms are obliged to provide in environmental disclosures can provide an estimate. Those disclosures suggest AI systems may have had a carbon footprint of between 32.6 and 79.7 million tons of CO2 in 2025, equivalent to New York City, while their water footprint could be 312.5–764.6 billion litres, in the range of the global annual consumption of bottled water. If we're to accurately track the environmental impact of AI, data centre operators must be required to improve the accuracy of these estimates, the researcher says.

News release

From: Cell Press

The carbon and water footprints of data centers and what this could mean for artificial intelligence

Sustainability & Solutions: AI systems have an estimated carbon footprint comparable to New York City in 2025 and a water footprint similar to the global annual consumption of bottled water, but they are impossible to accurately measure due to a lack of transparency in the industry

Attachments

Note: Not all attachments are visible to the general public. Research URLs will go live after the embargo ends.

Research Cell Press, Web page The URL will go live after the embargo ends

Journal/
conference: Patterns

Research:Paper

Organisation/s: Institute for Environmental Studies, the Netherlands

Funder: No information provided.

Media Contact/s

Contact details are only visible to registered journalists.