Zahlen, bitte! 12.75 billion cubic meters of water used in companies
Although water consumption in Germany's economy is declining – AI data centers could lead to problems in areas with water shortages.
Water is essential for life. Not only does everyone need water to live, water is also vital for a functioning economy. In 2022, companies in Germany used a total of 12.75 billion cubic meters of water. By comparison, Lake Constance, the largest natural freshwater reservoir in Europe, has a water content of 48 billion cubic meters.
By far the largest use is for cooling systems. 82.9 percent, or 10.57 billion cubic meters, were used for this purpose, followed by 1.76 billion for the production of goods and 0.42 billion for irrigation.
The frontrunners here are energy suppliers, which used 6.59 billion cubic meters of water for this purpose. However, water consumption fell in this area: the decommissioning of the last nuclear power plants reduced water use by 2.2 billion cubic meters, which is 16.7 percent less than in 2019.
However, water demand is on the rise due to the rapidly growing number of data centers, especially for resource-intensive AI computing. At Microsoft alone, water consumption is expected to increase by a third in 2022 due to ChatGPT.
It is estimated that the growth of AI data centers will increase global water consumption to between 4.2 and 6.6 billion cubic meters by 2027. The enormous computational effort involved in AI calculations requires powerful hardware such as graphics processors or tensor processors for machine learning. As the largest location for data centers in the EU, Germany has a special role to play here. However, measuring the water consumption of data centers and their impact on the environment is not so easy, as little research has been carried out into this aspect to date.
(Image:Â CC BY-SA 3.0, Victor Grigas)
Two evaluation metrics have been established to measure efficiency: power usage effectiveness (PUE) is a metric for energy efficiency by comparing the ratio of total energy consumption to the energy actually used by the IT equipment. The equivalent for water efficiency is Water Usage Effectiveness (WUE), which compares the ratio of total water consumption to energy consumption.
The research report published by the Gesellschaft für Informatik e. V. in June 2025: “Impact of AI, data centers and semiconductors on water availability and quality” (PDF) criticizes the fact that the key figures do not fully cover water consumption: “The widely used Power Usage Effectiveness (PUE) value allows conclusions to be drawn about the energy efficiency of data centers, but does not adequately reflect either water consumption or material use over the entire life cycle. The Water Usage Effectiveness (WUE) is even less common and also does not adequately reflect the total water consumption of data centers.” The study focused on the entire life cycle of the hardware in data centers in addition to the operational process.
AI data centers exacerbate the problem of water scarcity, especially in regions with water stress, where the water supply cannot adequately meet demand. The US state of Virginia, which now has the largest concentration of data centers in the world due to tax benefits, saw its water consumption grow from 4.28 million cubic meters in 2019 to more than 7 million cubic meters, prompting critics to raise their voices.
Videos by heise
According to the study, the water consumption of AI data centers has only recently become an issue. As many operators of AI centers do not report their water consumption, the studies investigating this issue usually rely on estimates. The study came to the conclusion that many factors influence water consumption by AI, including location factors, cooling methods, and the type of energy generation. They suggest that water consumption should be considered in everything from software design and hardware operation to cooling methods and location factors.
On the one hand, this would meet the needs of critics of such infrastructure projects, and, on the other, it would protect against climate change, for example, if drought exacerbates a region's water shortage.
(mawi)
