Rising AI workloads are heating up demand for water-intensive data center cooling.Image: Alamy
With AI driving rapid growth in global data center infrastructure, the water required to cool those facilities is emerging as a significant financial, operational, and reputational risk.
According to a recent Moody’s report, water stress is escalating in key data center markets.
The consulting firm identifies water scarcity as a growing credit risk for data center operators, particularly hyperscale providers supporting AI workloads that require denser compute power and higher thermal loads.
“As cloud computing and hyperscalers build or lease new data centers to meet escalating computing demand, their water consumption may strain resources in some markets,” the report cautions.
A Thirst for New Facilities
Most new large-scale facilities rely on water-based cooling systems, which are more efficient for high-density computing but can significantly increase total water consumption.
Moody’s warns that data center developers who fail to use water-efficient systems may face heightened scrutiny from regulators and investors.
John Medina, senior vice president of Moody’s global project and infrastructure finance, tells Data Center Knowledge that groundwater scarcity poses a risk in fast-growing hyperscale regions, like the southwestern U.S., Chile, and India.
Related:Direct-to-Chip Cooling: Everything Data Center Operators Should Know
“Local regulation, infrastructure constraints, and social backlash can delay projects or increase costs,” he says. “Developers may also relocate large data center projects to other, more business-friendly locations.”
One hyperscale facility can consume millions of liters of water per day – on par with small urban areas – depending on climate and infrastructure design.
The report notes that hyperscalers, including Google, Microsoft, AWS, and Meta, have committed to becoming “water positive” by 2030, meaning they aim to replenish more water than they use.
While these companies have successfully lowered their average water usage effectiveness (WUE) to between 0.2 and 0.5 liters per kilowatt-hour (kWh), the overall water demand from the data center industry continues to rise due to rapid growth in hyperscale and AI-driven infrastructure.
Operators Adapt – But Carefully
According to Sean Farney, vice president of data center strategy at JLL, the industry’s most prominent players are taking a proactive approach to mitigate resource concerns during the site selection and design phases.
“Operators won’t put capital at risk in locations where essential resources like water and power are missing,” he says. “The homework has been done, and if a deficiency arises, they’ll mitigate the risk from day one.”
Related:Eastern European Data Center Uses Gorge for Natural Cooling
The challenge is balancing environmental goals with uptime requirements and financial viability. Technologies like liquid cooling and adiabatic systems promise lower water usage but bring their own set of risks and complexities.
“If the business model assumes three to five nines of uptime, anything you do to enhance sustainability cannot compromise that,” he says. “Astute data center managers are balancing uptime commitments with environmental targets.”
Many operators are moving from traditional evaporative cooling to air-based or hybrid systems, particularly in high-stress geographies.
Others are exploring the reuse of greywater, modular cooling designs, or closed-loop systems. Still, such transitions are often expensive, particularly when retrofitting existing assets.
Farney points out that in certain geographies, a kilowatt of compute comes with a predictable draw on local water systems.
Ram Sri, vice president of sustainable finance for Moody’s Ratings, adds that data centers increasingly compete with local communities and agriculture in water-stressed regions.
“This can trigger strict permitting rules, public opposition and project delays,” Sri says. “As water stress worsens, regulatory scrutiny and reputational risks will intensify, affecting growth and costs.”
Related:Liquid-Cooled SSD Eases AI Data Center Heat Challenges
Water Goals Meet Market Pressures
While the tech giants have set water-positive goals, Moody’s cautions that those targets are being tested by rapid AI-driven expansion.
As physical footprints and workload density rise, total water use grows despite efficiency gains.
The report argues that achieving net water positivity requires not just conservation but also local replenishment, often in arid or overdrawn watersheds.
Investors are beginning to take notice. Some operators are turning to green bonds and sustainable financing to fund water-related upgrades, and several municipalities now require impact assessments before approving new builds.
These financial instruments could become more common as ESG criteria tighten and governments introduce new water usage and disclosure rules.
Next-Gen Cooling Investment
With AI pushing the thermal limits of compute hardware, and local authorities raising the bar on resource management, water is becoming one of the most critical – and contested – inputs to digital infrastructure.
Medina notes operators are investing in hybrid and advanced cooling technologies to cut water use without compromising uptime. However, new waterless technologies lack proven track records and introduce maintenance risks, he said.
“From a credit lens, forward-looking adaptation is positive and delays in modernizing infrastructure could indicate delayed capital investment to meet evolving tenants’ demands for increased sustainability in some regions,” he says.
Farney says that, ultimately, the viability of new data centers may increasingly hinge on local water conditions as much as on power availability or fiber routes.
“At the moment, the focus is all on power availability and scalability,” he says. “If a geography suddenly became risky due to water constraints, both the site selection and capacity management teams would already be mitigating risk by building or load-shifting to alternative locations.”
