AI’s rapid growth is increasing its environmental footprint. Studies suggest AI systems could soon emit as much carbon as a major city and consume vast amounts of water, raising concerns over energy use and transparency.
Artificial intelligence is becoming a major part of daily life. It is used in many areas, such as online searches, chatbots, image creation, and business tools. As its usage increases, so does the environmental impact of AI, a cost that is now becoming more apparent.
A recent report suggests that by the end of 2025, the total carbon emissions from AI systems around the world could be about the same as those from New York City. At the same time, the amount of water used to support AI operations could be comparable to the total water consumed by the global bottled water industry.
AI's Environmental Impact
AI systems require large data centres filled with high-powered computers that perform complex calculations continuously. As AI becomes more widespread, more data centres are built, and existing ones are expanded. This leads to a significant demand for electricity. A large amount of water is also needed, mainly for cooling the servers and for producing electricity in power plants.
Related Articles
However, it is difficult to measure exactly how much energy and water AI uses. Major tech companies like Google, Amazon, and Meta release general energy figures that include all their activities. This makes it hard to isolate AI's impact from other digital services. Some companies do not report their water usage at power plants, stating that it is beyond their direct control.
To improve transparency, researcher Alex de Vries-Gao from the VU Amsterdam Institute for Environmental Studies conducted a detailed study published in the journal Patterns. He used public company reports, sales data, and figures from the International Energy Agency to estimate the environmental impact of AI.
Carbon Emissions and Water Use
First, he calculated the electricity consumption of AI hardware by looking at the sales of high-performance chips, such as those made by NVIDIA, which are commonly used in AI. He then estimated water usage by examining the efficiency of water use in data centres for cooling and the amount of water needed to produce electricity.
Using this data, he made predictions for 2025. His estimates suggest that AI systems could produce between 32.6 and 79.7 million tonnes of carbon dioxide in a single year. Water use could range from around 312 to 765 billion litres. These are huge numbers, especially as AI continues to expand rapidly.
Concerns Over Transparency
The goal of the study is not to halt AI development but to emphasize the importance of transparency and responsible planning. De Vries-Gao argues that tech companies need to be more open about their energy and water usage. Clear reporting would help create better estimates and reduce environmental harm. Without action, the hidden environmental costs of AI could keep increasing, adding more pressure to an already warming planet.
