The rapid growth of AI is not only a story about chips, models and cloud capacity. It is also a story about water. As demand for data centers rises, so does the need to cool the computing equipment inside them, and that cooling can require large volumes of water.
That pressure is already visible in Virginia, which is home to the world’s largest concentration of data centers. According to FT, water usage there jumped by almost two-thirds between 2019 and 2023, moving from 1.13 billion gallons to 1.85 billion gallons.
Why AI growth changes the water equation
AI systems depend on heavy computing infrastructure. As more AI services are built, launched and used, the demand for data centers increases. Those facilities house computing equipment that must be cooled to operate reliably.
Water is one tool used in that cooling process. The source article makes the link direct: the AI boom is fueling demand for data centers, and that demand is driving up water consumption.
This matters because data centers are no longer a background utility in the digital economy. They are becoming a more visible part of the physical infrastructure behind AI. Their footprint includes not just electricity and hardware, but also water use in the places where they operate.
Virginia shows the scale of the rise
The clearest number in the source comes from Virginia. The state is described as home to the world’s largest concentration of data centers, making it a key example of how concentrated digital infrastructure can affect local resource demand.
Between 2019 and 2023, water usage in Virginia rose from 1.13 billion gallons to 1.85 billion gallons. The source describes that increase as almost two-thirds. In plain terms, the amount of water associated with this data center hub became much larger over a relatively short period.
That example helps explain why the issue is gaining attention. AI may feel abstract to users, but the infrastructure behind it is local. Data centers sit in specific communities, draw from specific systems, and create specific operational needs.
Why closed-loop recycling does not solve everything
A natural question is why data centers cannot simply recycle water in a closed-loop system. The source gives a partial answer: many data centers do use closed-loop systems, but not all water can stay inside the loop.
Some of the water is used for humidity control. That water can evaporate, which means it is consumed rather than continuously reused. This is an important distinction because recycling can reduce some demand, but it does not eliminate water consumption in every part of the cooling and operating process.
Humidity control is not a cosmetic feature. In drier regions, air that is not humidified can become a strong conductor of static electricity. For computers, static electricity is usually bad news. That creates a practical reason why facilities may use water in ways that are not fully recoverable.
- Cooling: Water helps manage heat from computing equipment inside data centers.
- Humidity control: Some water is used to keep air conditions safer for computers.
- Evaporation: Water used for humidity can leave the system, limiting the benefit of recycling.
Water stress is already part of the data center footprint
The concern is not limited to one region. The source says many describe the trend as unsustainable and notes that it is playing out worldwide. Two large data center operators have reported water use in areas facing water pressure.
Microsoft said 42% of the water it consumed in 2023 came from “areas with water stress.” Google, which has among the largest data center footprints, said this year that 15% of its freshwater withdrawals came from areas with “high water scarcity.”
Those figures show why the issue is more than a technical cooling question. Water consumption becomes more sensitive when it happens in places where water is already stressed or scarce. Even if a data center is essential to digital services, its resource needs still interact with local environmental limits.
The source does not say that every data center faces the same conditions. It does show that major operators are already measuring exposure to water-stressed or high-scarcity areas. That makes water a visible part of the AI infrastructure debate.
The larger implication for AI infrastructure
The central issue is not whether data centers can exist without water. Based on the source, the more practical question is how much water they consume, where that consumption happens, and whether growth can continue on the current path.
AI demand is increasing pressure on data center capacity. Data centers need cooling. Cooling and humidity management can consume water. In regions that are dry, water-stressed or already scarce, that chain of demand becomes harder to ignore.
For readers following AI, the lesson is straightforward: the technology’s future depends on physical systems. Servers, cooling equipment and water supplies are part of the same story as software. As AI use expands, the resource demands behind it will remain a key measure of whether the boom can scale sustainably.