Across the United States, communities are pushing back against the rapid surge in data center construction, citing concerns not only about how these facilities could affect their utility bills but also about their potential impact on local water supplies.

“A large, continuous industrial user, like a data center, fundamentally alters a watershed because its water demand never sleeps,” said Matthias Sprenger, an assistant professor in the Department of Forestry and Environmental Resources at North Carolina State University. 

Sprenger, a hydrologist who studies the effects of land use change on water resources, added that data centers that use open-loop evaporative cooling systems “extract water and evaporate it every day of the year” and that, over time, “this constant loss disrupts the natural water balance.” 

Large-scale data centers consume millions of gallons of water per day to continuously cool active computer servers, drawing the water mostly from local municipal drinking water systems, underground aquifers and nearby surface water bodies. 

Whereas municipalities or industrial users usually treat and return withdrawn water to local streams, many data centers use open-loop evaporative cooling systems that convert water into vapor and release it into the atmosphere. 

That vapor eventually returns as precipitation but it doesn’t necessarily fall back into the local watershed from which it was originally withdrawn, which according to Sprenger, can have negative long-term effects on groundwater and surface water. 

Groundwater is water stored underground, while surface water is found in rivers, lakes and streams. Both are important sources of freshwater and are used for drinking, farming, industry and other daily needs.

The Water Cost of the Data Center Boom 

Open-loop evaporative cooling systems operate 24 hours a day, 7 days a week, and the continuous withdrawals can lower groundwater levels and reduce streamflow, putting pressure on municipal water supplies and aquatic ecosystems.

Those effects can become more severe during periods of drought or increased water demand, when natural water supplies are already depleted and may not be able to replenish as quickly as data centers consume them.

“When a severe drought strikes, data centers represent an inflexible demand,” Sprenger said. “Unlike residents who can stop watering lawns, most data centers cannot easily turn off their cooling systems without shutting down operations.” 

Sprenger said data centers utilizing open-loop cooling systems can also impact water quality. These facilities may be permitted to discharge wastewater that is warm or contains chemical additives into waterways.

As water evaporates during the cooling process, it leaves behind minerals and salts. Data centers must periodically drain this leftover water to protect equipment. This drained water is called blowdown.

Blowdown, when discharged into local waterways, “can elevate salinity levels and carry chemical additives like biocides or anti-scalants used to treat the cooling equipment,” Sprenger said. 

These pollutants can degrade water quality, and, in some cases, increase the treatment burden on municipal water treatment plants, requiring additional processes to remove contaminants before the water can be safely discharged or reused.

Furthermore, discharging warm water directly into streams creates thermal pollution. Thermal pollution occurs when industrial processes alter the natural temperature of local water bodies, disrupting aquatic ecosystems. 

“Depending on the time of the year, even a slight temperature increase in river water can push aquatic organisms past their biological limits and lower oxygen levels, endangering local ecosystems,” Sprenger said. 

Data Centers’ Hidden Water Footprint 

Beyond on-site water use, data centers consume substantial amounts of water indirectly through electricity generation and chip manufacturing — often exceeding their direct operational consumption. 

Data centers rely heavily on electricity from regional power grids, pulling from whatever mix of natural gas, coal and nuclear power plants are being operated by local utilities. These power plants require cooling systems that typically use water. 

A report from Ceres, a nonprofit organization, found that powering data centers in seven states — Virginia, Texas, California, Illinois, Georgia, Ohio and Arizona — used an estimated 3.4 trillion gallons of water withdrawals in 2024. 

The report also found that 78% of the electricity in the states analyzed came from power plants that use water to operate, and 66% of that electricity was generated by power plants in areas facing medium-high to extremely high water stress. 

Compounding existing water demand from on-grid power plants, technology companies are increasingly building private, off-grid power plants dedicated to supplying electricity to U.S. data centers.

Global Energy Monitor, a nonprofit organization that tracks oil and gas developments, found that planned natural gas-fired power projects dedicated to U.S. data centers surpassed 189 gigawatts by mid-2026, nearly doubling in just six months. 

“Energy and water are interconnected,” Sprenger said, noting that the water consumed by power plants generating electricity for data centers places additional strain on water resources.

Sprenger added that the construction of data centers and related infrastructure — like most industrial developments — also creates more impervious surfaces, which can reduce rainwater infiltration into the soil and groundwater and increase faster surface runoff into streams during intense rainfall events. 

Reduced rainwater infiltration limits the amount of water replenishing groundwater supplies, while increased surface runoff can carry pollutants into streams and reduce the amount of rainwater available to replenish local water resources.

Same Data Center, Different Impacts

Not all data centers — or their impacts — are equal. Two identical data centers can have dramatically different effects on water resources simply because they sit in different places, according to Sprenger. 

Where a data center is built matters because differences in climate and electricity sources can lead to very different water footprints, shaping not just direct water use but also the indirect pressures a facility places on surrounding watersheds.

A data center in Arizona, for example, will likely have a different impact on the regional water supply than a data center in Washington. Because Arizona is hot and dry, the facility may need more water for cooling. 

Arizona also powers its data centers using a mixed electricity grid, much of which is fueled by natural gas, whereas Washington relies heavily on hydroelectric power systems that return water back into local watersheds. 

Local water availability and existing demand must also be considered when evaluating the hydrological impact of data centers, according to Sprenger. Many U.S. cities have experienced rapid population growth, driving increases in water demand. 

As a result, Sprenger said placing “a massive, constant water user” like a data center in areas already facing water stress could further strain local water supplies and create competition among residential, agricultural and industrial users. 

Communities cannot easily look elsewhere for more water when local supplies come under pressure, Sprenger added. “Unlike electricity, which can be transported across state boundaries through the electrical grid, freshwater cannot be easily redistributed between regions.”

That’s also why it is important to account for the “temporal variability” of water supply and demand, according to Sprenger. In other words, water availability can change quickly during droughts, while demand from data centers, industry and growing populations remains relatively steady.

The concern, Sprenger said, is both whether a region can withstand those short-term shortages and whether sustained demand could gradually deplete its water resources over time.

A Thirst That’s Only Growing 

While many local governments promote data centers as drivers of economic growth, Sprenger said they present a major uncertainty for the future of water resources. 

“Predicting the scale, geographic distribution and power demand of future facility buildouts over the next two decades remains difficult,” he said.

However, Sprenger noted that AI workloads require more energy than traditional data storage, meaning “hyperscale data centers could exert rapid higher localized water demands than the current generation.” 

AI use has grown exponentially in recent years, triggering a global construction boom for data centers. In the U.S., major tech companies are projected to spend over $700 billion on AI infrastructure in 2026, nearly doubling what they spent the year before. 

Sprenger warned that the growth of AI infrastructure, combined with the effects of climate change, “is making water supplies less predictable.” 

“We are adding a large, constant and inflexible water demand into a system that is becoming more variable, not less,” he said. 

Climate scientists predict with high confidence that air temperatures will continue to rise, allowing the atmosphere to hold more moisture. These changes are likely to increase evaporation, which can intensify drought conditions.

At the same time, Sprenger said more moisture in the atmosphere could lead to heavier rainfall events that produce rapid runoff into streams rather than infiltrating the soil and replenishing groundwater.

Sprenger added that it’s “impossible” to predict where and when severe multi-year droughts or localized extreme rainfall events will strike across specific river basins over a 10- to 20-year horizon.

That uncertainty forces planners to make long-term water allocation decisions, weighing the needs of data centers against other demands for water while accounting for an increasingly unpredictable supply.

“Any attempts towards a sustainable water use of data centers and the conflict between different water needs in a region will have to account for the added uncertainties in water availability that climate change introduces,” Sprenger said. 

Water-Smart Solutions for Data Centers 

Ultimately, Sprenger said efforts to curb the hydrological impacts of data centers must focus on transitioning facilities away from drinking water to recycled or non-potable water, adopting closed-loop or air-cooling technologies and powering operations with renewables like solar and wind to reduce indirect water consumption.

“Since data centers that use less water in their cooling system generally require more energy, renewable energy could reduce the localized water stress power plants and data centers can cause,” Sprenger said. 

Policymakers and regulators — including active discussions within the North Carolina legislature — are increasingly moving to restrict or eliminate open-loop evaporative cooling in favor of closed-loop or dry-cooling alternatives.

Advanced liquid cooling, closed-loop micro-channel technology and dry-cooling innovations could “drastically reduce direct water consumption in future facilities,” Sprenger said. 

However, “there is technological uncertainty surrounding cooling efficiency,” Sprenger said. It remains unclear whether these technologies can be cost-effectively retrofitted into existing facilities or deployed fast enough to offset the water demands of new construction.

For Sprenger, the larger concern is the speed at which data center development is unfolding. New facilities can be built in a matter of years, he said, while the hydrological consequences of sustained increases in water consumption can unfold over decades.

“By the time we can clearly measure a decline in groundwater levels or baseflow, the underlying infrastructure and its water demand are already locked in,” Sprenger said. 

Sprenger added that data centers have one potential advantage: Unlike most water users, their water demand is measurable and, in principle, manageable — provided there is transparency. 

Unfortunately, according to Sprenger, “a persistent lack of transparency leaves public regulators and communities guessing” about how much water data centers use and what those demands mean for local water supplies.

“What gives me hope is that people are paying more attention to the environmental impact of development projects in their regions and they speak up in the communities to voice their concerns,” Sprenger said. 

Sprenger said he hopes decision-makers will use long-term hydrological data, monitoring and modeling to anticipate a facility’s impacts before it is built, rather than after. 

“If regional water budgets become a standard part of siting decisions rather than an afterthought, this growth does not have to come at the expense of the water communities depend on,” he said.

This post was originally published in College of Natural Resources News.