May 2026

Why the data-center industry is quietly investing billions into water reuse, recycling, and low-water cooling systems

Artificial intelligence is fueling one of the largest infrastructure booms in modern U.S. history. Across Texas, Arizona, Ohio, Georgia, Iowa, and Northern Virginia, new data centers are rising at extraordinary speed to support AI training, cloud computing, and high-performance workloads.

But beneath the headlines about chips, electricity, and trillion-dollar AI valuations lies another critical resource increasingly shaping the future of digital infrastructure: water.

Data centers require enormous amounts of cooling. As AI systems become denser and hotter, the pressure on cooling systems has intensified — especially in regions already facing drought, groundwater depletion, and public scrutiny over industrial water use.

Over the past 18 months, major technology companies and data-center operators have responded with a significant shift in strategy. Instead of treating water efficiency as a secondary sustainability initiative, corporations are increasingly redesigning facilities around reclaimed water systems, closed-loop cooling, on-site treatment infrastructure, and even fully waterless designs.

The change is no longer theoretical. Between January 2025 and May 2026, companies operating in the U.S. data-center market committed an estimated $570 million to more than $1 billion toward water-related infrastructure and technologies tied directly to data-center operations. That figure is conservative and likely understates the true scale of investment because most companies still do not disclose water-specific capital expenditures.

What is clear, however, is that water strategy has become a core infrastructure issue for the AI economy.

Water Is Becoming a Constraint on AI Infrastructure

For years, discussions about data-center sustainability focused primarily on electricity consumption and carbon emissions. Water received far less attention.

That is changing quickly.

Modern AI systems generate far more heat than traditional cloud-computing workloads. Higher rack densities require more aggressive cooling strategies, and many of the fastest-growing data-center markets are located in regions with rising water stress.

The result is a convergence of pressures:

• AI workloads are increasing thermal density.

• Communities are pushing back against large industrial water users.

• Regulators are tightening water-use expectations.

• Investors are demanding better disclosure.

• Operators are trying to preserve siting flexibility in water-constrained regions.

The industry’s response has been a rapid acceleration of water-related infrastructure investment.

The shift is visible across multiple categories:

• reclaimed-water sourcing agreements

• on-site water recycling and treatment

• closed-loop cooling systems

• direct-to-chip liquid cooling

• non-evaporative cooling designs

• waterless cooling architectures

• watershed replenishment programs

What began as a sustainability discussion is increasingly becoming an operational requirement.

Amazon Is Setting the Pace

Among all major hyperscalers, Amazon Web Services has provided the clearest evidence of direct investment.

By early 2026, AWS reported:

• 13 utility agreements for reclaimed-water use at data centers

• 31 on-site water-treatment facilities

• more than $1 billion in contracted investment tied to treatment upgrades and distribution infrastructure

• plans to expand recycled-water systems from 24 U.S. data centers to more than 120 locations by 2030

The company says those systems are expected to preserve more than 530 million gallons of drinking water annually across the United States.

One of the most significant recent examples emerged in Mississippi, where Amazon partnered with Veolia on a reclaimed-water cooling project intended to reduce dependence on potable municipal supplies.

AWS has also improved operational efficiency metrics. The company reported a global water usage effectiveness (WUE) of 0.15 liters per kilowatt-hour in 2024, alongside major reductions in cooling-water demand across North America.

Importantly, Amazon’s disclosures reveal how the industry’s investments are evolving. Instead of focusing only on internal facility systems, companies are increasingly funding external water infrastructure — including treatment upgrades, distribution systems, and utility integration.

That marks a major change in how data centers interact with local water systems.

The Rise of Waterless Cooling

At the same time reclaimed-water infrastructure is expanding, another major trend is reshaping the industry: the migration toward closed-loop and waterless cooling systems.

Several operators now market low-water cooling as a defining feature of new AI campuses.

Developer Edged has become one of the clearest examples.

Between 2025 and 2026, the company advanced or opened waterless facilities in Illinois, Iowa, Arizona, Ohio, and Georgia. Individual campuses are projected to save between 52 million and 380 million gallons of water annually compared with conventional cooling approaches.

The appeal is obvious.

Traditional evaporative cooling systems can consume large amounts of water, particularly in hot climates. Waterless or near-waterless systems dramatically reduce dependence on local supplies while simplifying permitting and community relations.

Oracle has embraced a similar strategy for its growing AI infrastructure footprint.

The company disclosed that major AI campuses in New Mexico, Michigan, Wisconsin, and Texas are using direct-to-chip, closed-loop, non-evaporative cooling systems that are filled once and then reused continuously.

In Abilene, Texas, Oracle stated that each building would require only about 50,000 gallons of water annually for cooling maintenance — a remarkably low figure for large-scale AI infrastructure.

Microsoft is moving in the same direction.

The company says its newer data-center designs eliminate water evaporation for cooling entirely and can avoid more than 125 million liters of water consumption per facility each year.

This transition matters because AI infrastructure is expected to become dramatically denser over the next decade. Operators increasingly view low-water cooling as essential to scaling future compute capacity.

Geography Is Shifting Alongside Water Risk

The geography of data-center development is also changing.

Historically, Northern Virginia dominated U.S. data-center growth. While the region remains critical, a growing share of new capacity is being built in so-called “frontier markets” — secondary regions where land, power, and development opportunities remain more available.

Industry research from JLL found that nearly two-thirds of North American data-center capacity under construction at the end of 2025 was located in these frontier markets.

Many of those markets sit in regions where water availability is increasingly sensitive.

In the Southwest and Texas basin, companies are rapidly deploying reuse systems and closed-loop cooling to address water stress and political scrutiny.

Key projects include:

• Oracle’s Project Jupiter in New Mexico

• Oracle’s Abilene AI campus in Texas

• DataBank’s South Dallas expansion

• QTS developments in Texas

• Edged’s Mesa, Arizona facility

Phoenix has become a particularly important case study. According to JLL, nearly all data centers built there over the past decade now rely on closed-loop or air-cooled systems that minimize external water consumption.

Meanwhile, Midwest markets such as Iowa, Illinois, Ohio, Michigan, and Wisconsin are also adopting low-water architectures despite facing less severe drought pressure.

The motivation there is broader: future-proofing infrastructure, supporting higher AI densities, and avoiding future regulatory complications.

The Southeast and Mid-Atlantic regions are seeing another pattern emerge — utility-linked reclaimed-water integration.

Projects in Virginia, Georgia, Mississippi, and Florida increasingly involve coordination between operators and municipal water systems, particularly around non-potable reuse.

Regulators and Investors Are Increasing the Pressure

Corporate sustainability goals alone are not driving the trend.

Regulatory momentum is growing.

In April 2026, the Environmental Protection Agency launched Water Reuse Action Plan 2.0, explicitly identifying AI and industrial growth as reasons to accelerate water reuse across the U.S. economy.

State and local governments are also becoming more aggressive.

Florida lawmakers advanced proposals that could require reclaimed-water use for large-scale data centers under certain permitting conditions. Tucson adopted rules requiring large water users — including proposed data centers — to disclose projected water consumption and conservation strategies before receiving service.

At the same time, investors are pressing hyperscalers for greater transparency.

A Reuters investigation reported that North American data centers used nearly 1 trillion liters of water in 2025, intensifying questions around disclosure, resource competition, and long-term sustainability.

The scrutiny is forcing companies to integrate water planning into infrastructure strategy much earlier in the development process.

Water Efficiency Is Becoming a Competitive Advantage

The operational data suggests these investments are producing measurable results.

Microsoft reported a global WUE of 0.27 in fiscal year 2025 and an 18% reduction in water intensity relative to its 2022 baseline.

QTS said its closed-loop cooling approach saved nearly 1.5 billion gallons of water in 2024 across its portfolio.

Edged’s campuses are projected to conserve hundreds of millions of gallons annually.

AWS reduced cooling-water needs by nearly 1 billion liters in North America during 2024.

These are not marginal improvements.

For operators competing to secure permits, attract enterprise customers, and maintain expansion flexibility in water-sensitive markets, water efficiency is increasingly becoming a strategic differentiator.

In many cases, the industry appears willing to accept greater engineering complexity and higher upfront capital costs in exchange for lower long-term water dependence.

That tradeoff is especially important for AI infrastructure.

Liquid-cooling systems, direct-to-chip architectures, and non-evaporative cooling can require additional equipment, piping, controls, and operational sophistication. But operators increasingly view those costs as necessary to support future compute density while maintaining acceptable environmental and political risk.

The Real Investment Total Is Probably Much Higher

One of the biggest challenges in assessing the market is disclosure.

Outside of AWS and a handful of Oracle announcements, most technology companies do not publish water-specific data-center capex.

Microsoft, Google, Meta, Equinix, Digital Realty, QTS, and many private operators all report extensive water-efficiency programs and infrastructure changes. However, they rarely isolate the water-related share of spending from broader facility construction costs.

That means the visible investment totals likely represent only a floor.

The true amount of capital flowing into water-related data-center infrastructure is almost certainly much larger than public disclosures currently reveal.

What matters most, however, is the directional shift.

Water is no longer treated as a secondary sustainability metric inside the data-center industry.

It is becoming a first-order infrastructure constraint — one that increasingly influences siting decisions, engineering standards, investor relations, permitting, and long-term AI expansion strategy.

The next phase of America’s AI buildout will not depend only on chips and electricity.

It will also depend on who can secure, recycle, conserve, and politically sustain access to water.

This article is based on analysis of corporate disclosures, sustainability reports, regulatory materials, infrastructure announcements, and industry research published between January 2025 and May 2026, including materials from AWS, Microsoft, Oracle, Google, Meta, JLL, Uptime Institute, the EPA, and major data-center operators.