In the past three years, more than 160 data centres have been built in parts of the U.S. already grappling with limited water resources — up 70% from the previous three-year period. Yet many utilities don’t factor data centres into their long-term water planning.
Generative Artificial Intelligence (AI), edge computing and blockchain mining are fueling explosive compute demands and a growing dependence on water for cooling. High-performance chips powering machine learning and inference workloads generate more heat than ever, and increasingly, the industry is sending that heat into liquid cooling systems, sometimes directly at the chip (direct to chip liquid cooling) or immersing servers in a dielectric fluid (immersion cooling) as physical limits of air cooling are exceeded.
Putting that in perspective, heat rejection increases fivefold when you move from air-cooled to water-cooled systems at the server level. We’ve gone from air to liquid, and now we’re back to liquid at a scale we haven’t seen before.
Given the new cooling calculus, it’s of little surprise that some communities view data centres as water hogs, whereas others were unaware of the role of water in the cooling equation. That trust gap often widens when data centre operators enter a market quietly.
When companies are upfront and show how they're strengthening local infrastructure instead of just extracting from it, the conversation shifts, and communities notice.
Water utilities need to start the conversation before developers select a site, finalise the cooling design, or push the project into execution
Even as some hyperscalers park their sustainability goals in the back seat during the AI arms race, they haven’t tossed them out. Many are rethinking how to meet those goals under tighter water constraints. Microsoft and others have started exploring ways to reject heat without using large volumes of water — part of a broader shift away from evaporative cooling and current server design to implement liquid cooling technologies.
Cooling demands are already changing. Many high-performance chips now require liquid cooling directly at the server rack, and that pushes heat rejection levels far higher than previous generations of server infrastructure.
Some hyperscalers are experimenting with air-based alternatives to avoid evaporative water use. Others are doubling down on liquid cooling but pushing for cleaner, more closed-loop models.
The technologies aren’t standing still, and neither are the demands.
That’s why water utilities can’t afford to wait for formal notifications or permit applications to start planning. Utilities need to start the conversation before developers select a site, finalise the cooling design, or push the project into full-speed execution to address the challenges.
Utilities bring critical context to these builds: they know the terrain and understand the constraints and seasonal stressors. They also know what local systems can and can’t support.
These are real partnership opportunities. However, they only materialise when the water utility is part of the build discussion. If the utilities don’t step in early, developers will solve the water problem without them by switching locations, methods or both.
How the future direction of data center and water evolve will depend on early engagement of key participants to secure permitting, promote sustainable practices and explore options like using recycled water or renewable energy to avoid conflicts with local communities concerned about environmental impacts. Furthermore, it will be important to partner with data centers to get their cost contribution to new infrastructure and offer incentives to attract development and collaboratively execute development of future data centers. Otherwise, the opportunity may be lost.