BREEAM

Breana Wheeler

Sustainable Infrastructure Advocate

The rapid expansion of AI is changing the requirements for the buildings and infrastructure that support it. Data centers are becoming larger, more power-intensive and more geographically dispersed, putting access to electricity and grid capacity at the center of development decisions. The resource equation, however, should not end with power.

Water enters the data center equation both before electricity reaches the facility and after it arrives.

Right now, much of the discussion around data center water use focuses on the water consumed inside the facility to cool increasingly powerful computing equipment. Less attention goes to the water required to generate the electricity those facilities depend on. Because most buildings receive electricity from generation sources relatively close to where they operate, often within roughly 50 miles, the water demands associated with power generation can be closely tied to decisions about where data centers are located.

For data centers that rely on evaporative cooling, water is part of the process of keeping servers operating within acceptable temperature ranges. The U.S. Department of Energy notes that water consumption in cooling-tower systems is directly related to the heat load created by IT equipment and the efficiency of the cooling system.

That means by the time a data center is operational, some of its most consequential water decisions have already been made, embedded in the site's location, cooling architecture and building design.

Water availability belongs in the site-selection conversation

Data center developers already evaluate sites based on factors such as power availability, network connectivity, land, permitting and proximity to customers. Water availability should be considered with the same discipline.

This becomes particularly important as development expands into regions where water resources are already constrained. A 2026 U.S. Geological Survey assessment found that water availability is highly localized across the United States, with 26.7 million Americans living in areas experiencing chronic high or severe water limitation. The same assessment of AI data center development concludes that projects should evaluate both current and future water availability because additional demand can increase competition among users.

The implication for developers is straightforward; a site that has sufficient water today may not have sufficient water for the life of the asset.

That’s why water risk should be evaluated alongside projected IT loads, climate conditions, and utility capacity, before a building is designed around a particular cooling strategy.

Cooling architecture can lock in water demand

There is no single cooling technology that will be right for every data center. Evaporative systems can offer energy-efficiency benefits, while dry cooling can reduce direct water consumption. Liquid cooling is also becoming increasingly important as computing densities rise.

  The question for developers, then, is not simply how much water a data center will use. It is whether the building's design makes sense for the water conditions where it will operate, both today and over its useful life.  

These choices involve tradeoffs between water, energy, cost, reliability and performance. The Department of Energy specifically notes that dry cooling can reduce water use while potentially increasing electricity consumption, underscoring the need to evaluate water and energy together rather than optimizing in isolation.

Technology is also expanding the options available to developers. Microsoft, for example, has developed a data center design that uses a closed loop cooling system and does not require water for cooling, avoiding more than 125 million liters of annual water use per facility under its stated design assumptions.

The right cooling strategy will depend on the water realities of the site, but the overarching lesson is clear: if water availability is a site constraint, it needs to shape the cooling strategy before the design is locked in.

By the time operations begin, the parameters are set

Once a data center is operational, teams can optimize cooling controls, monitor water usage and improve system efficiency. However, operational optimization has limits when the fundamental design does not align with the site's water conditions.

That distinction matters as AI accelerates demand for data center capacity. The International Energy Agency projects global data center electricity consumption to roughly double by 2030, with AI a major driver of growth. As computing demand grows, the physical infrastructure supporting it will need to account more deliberately for the resources required to operate it.

The question for developers, then, is not simply how much water a data center will use.

It is whether the building's design makes sense for the water conditions where it will operate, both today and over its useful life.

That question belongs at the beginning of the development process, when there is still an opportunity to make a different decision.