AI’s 2025 Toll: Carbon Surpasses NYC, Water Use Outpaces Global Bottled Industry
As per the study, the carbon footprint of AI systems in 2025 was 80 million metric tonnes, whereas the water they utilised reached 765 billion litres.
The widespread proliferation of AI systems and technology has caused the emission of as much carbon dioxide as New York City emits, according to a study by Dutch academic Alex de Vries-Gao, the founder of Digiconomist.
The amount of water this tech utilises also surpasses the aggregate sum of global bottled water demand, as per the study, which has been published in academic journal Patterns.
According to De Vries-Gao, this is the first attempt to measure the cumulative environmental impact of AI itself instead of just data centres. The research comes amid the widespread adoption of applications like OpenAI's ChatGPT and Google's Gemini.
The numbers cited by the researcher indicate that greenhouse gases from the use of AI are now equal to over 8% of aviation emissions around the world.
As per the study, the carbon footprint for AI systems in 2025 was 80 million metric tonnes, whereas the water they utilised had reached 765 billion litres. This showed that water use by AI technology was a third higher than the initial estimates of water use by data centres.
AI data centres were found to draw out as much electricity as power-guzzling aluminium smelters, with data centre electricity consumption expected to be more than double by 2030. The International Energy Agency also reported that the biggest AI-focussed data centres being made now will consume as much electricity as two million households within the US.
