Rethinking Water in the AI Data Center Era

The AI data center boom has turned water into a first-order design constraint, placing it alongside power, land, and connectivity in early site and campus planning. The reason is straightforward: many of the most energy-efficient cooling strategies still rely on water (particularly evaporative cooling) at the same time the industry is expanding into regions where water is already politically, hydrologically, or economically constrained. Media coverage has captured this collision course in water-stressed metros such as Phoenix, where large campuses can often be permitted faster than new, long-term water supplies can be secured.

But water use is not a single line item. Increasingly, the conversation is splitting into two distinct questions:

• How much potable water is consumed onsite to cool IT equipment?

• What is the total water footprint of the data center’s electricity supply, i.e. its indirect water use, which can be substantial but is far harder to see, measure, and report?

Recent coverage and industry disclosures map neatly onto what is becoming an emerging playbook for addressing both questions at once:

• Treat and reuse water onsite at industrial scale.

• Substitute reclaimed municipal wastewater for potable water in cooling systems.

• Offset and replenish impacts through measurable, locality-specific watershed projects that deliver community co-benefits.

What follows is a look at how these approaches are being discussed and deployed, and what they imply for operators trying to scale AI factories without triggering the next generation of water moratoriums, rate battles, or public-trust crises.

Is WUE the New PUE?

If PUE became the industry’s shorthand for energy efficiency, WUE [Water Usage Effectiveness] is quickly emerging as the shorthand for water performance. WUE is typically expressed as liters of water used per kilowatt-hour of IT energy. Microsoft, for example, defines WUE as the annual liters of water used for humidification and cooling divided by annual IT kWh.

Why WUE matters:

• It enables comparison of water intensity across sites, regions, and cooling architectures.

• It surfaces the core trade-off the industry can no longer avoid: some low-water cooling approaches increase energy use, while some of the most energy-efficient designs increase water consumption. Reporting highlighted by Reuters underscores this tension, noting that closed-loop or “zero-water” approaches can reduce onsite water use but may materially increase electricity demand.

What to watch out for:

• WUE can improve while local impacts worsen. A site may show better WUE metrics after switching from potable municipal water to a non-potable source that still competes with agricultural, industrial, or ecological needs, or simply by operating in a watershed where any incremental demand is politically toxic.

• WUE does not capture water quality or discharge impacts. Issues such as blowdown handling, discharge chemistry, and thermal effects fall outside a simple volumetric metric. As a result, the conversation is expanding beyond raw water volumes toward broader stewardship and accounting frameworks.

Acknowledging these limitations, the industry’s thinking is converging around a more nuanced hierarchy of water decision-making:

• Avoid — Design to reduce or eliminate onsite water dependence where conditions demand it.

• Substitute — Use non-potable or reclaimed sources where water-based cooling remains the most energy-efficient option.

• Recycle / Recover — Close loops, minimize blowdown, and treat wastewater for reuse.

• Replenish — Invest in credible, catchment-scale projects that restore water to the watershed and deliver verifiable local benefits.

Gradiant Brings Industrial-Scale Water Treatment Into the Data Hall

Gradiant’s May 2025 announcement is notable not for the technology alone, but for how it reframes the data center itself. Rather than treating water as a secondary cooling input, Gradiant positions large AI campuses the way semiconductor fabs and advanced manufacturing facilities already do: as mission-critical industrial water systems with uptime, reliability, and reuse requirements that extend well beyond “just cooling.”

The company disclosed that it secured two contracts from leading technology companies to design and deploy sustainable water solutions at new data center sites in the United States and the Indo-Pacific, regions it explicitly identified as increasingly water-stressed. In announcing the projects, Gradiant emphasized that the pace and scale of AI infrastructure growth are forcing operators to confront water constraints much earlier in the site-selection and design process.

As Gradiant CEO Anurag Bajpayee put it:

The Key Technical Claim: Near-Closed-Loop Water

Gradiant highlights zero liquid discharge (ZLD) as a core capability, pointing to systems that can recover and reuse up to 99% of process water onsite, dramatically reducing freshwater withdrawals in water-constrained regions.

In data center terms, ZLD (or ZLD-like architectures) signals where cooling and water management are headed in the most constrained markets. Traditional cooling towers rely on continuous makeup water and generate blowdown to manage scaling and water chemistry. ZLD-style systems treat that blowdown so it can be reused, leaving a far smaller residual waste stream (often solids or brine) instead of continuous liquid discharge.

In practice, this pushes operators toward a “water plant on campus” model:

• Pretreat incoming sources, including non-potable water.

• Support cooling and humidification loops.

• Treat blowdown and other wastewater streams.

• Reuse water internally.

• Minimize or eliminate routine discharge.

This matters because community opposition to new data center development often coalesces around two fears: “you’ll take our water” and “you’ll dump your waste.” A credible recycle-and-reuse strategy addresses both concerns; particularly as regulators increasingly evaluate large AI campuses as industrial users rather than conventional commercial buildings.

Gradiant also positions its SmartOps AI platform as a water operations tool, using real-time monitoring and analytics to optimize water and wastewater systems, reduce operating costs, and minimize downtime.

That framing is not marketing fluff. Water systems in data centers fail in expensive and operationally dangerous ways: through scaling, corrosion, biofouling, and chemistry excursions that can threaten uptime. As AI racks drive higher heat loads and more complex cooling architectures, water chemistry and system reliability become availability issues, not just sustainability metrics. Gradiant further emphasizes specialized chemical formulations designed to reduce overall chemical dosing while improving performance and reliability.

This is an under-discussed point: a site can appear “water-efficient” on paper while still carrying a heavy environmental footprint if its cooling loop depends on aggressive chemical treatment and frequent blowdown. Tight operational control, paired with better chemistry, is essential to making high-reuse systems viable without simply trading one environmental problem for another. Gradiant’s underlying argument is that the modern data center is becoming a water-engineering problem with industrial-grade expectations, and that competitive advantage will increasingly accrue to operators who can deploy high-reuse, high-reliability water plants as standard campus infrastructure.

Reclaimed Municipal Wastewater: The Quiet Scaling Lever

In a January 13, 2026 blog post from Koomey Analytics, Jonathan Koomey highlights two things the data center industry has historically struggled to do well:

• Treat reclaimed water as a strategic option, rather than a one-off pilot or market-specific workaround.

• Quantify adoption across major operators, as rigorously as public data allows.

Not all water use is equal, and Koomey makes this distinction plainly. Some data centers rely on potable municipal water, others on surface or groundwater, and a growing number on recycled sources; most commonly reclaimed municipal wastewater. He argues that, from an environmental standpoint, using reclaimed wastewater for cooling generally has far lower impacts than drawing from potable or natural freshwater sources, and that adoption is quietly expanding.

That framing helps explain why reclaimed water has become the most politically defensible choice in many markets:

• It typically relies on water that would otherwise be discharged after treatment.

• It leverages existing municipal wastewater infrastructure, which is often widely available in large metro areas.

• It reduces direct competition with residential potable supplies—a recurring flashpoint in public hearings and permitting battles.

What the Adoption Snapshot Shows

Because disclosure remains limited, Koomey and co-author Zachary Schmidt assembled evidence from a wide range of sources (including public reports, press releases, mapping tools, utility contracts and billing data, conference materials, and even satellite imagery) to infer where reclaimed water is being used for data center cooling.

Their headline finding is instructive:

As of December 2025, Amazon had the highest number of confirmed sites using reclaimed water for cooling (24), followed by two other operators at 17 and 13 sites, several in the 6–8 range, and the remainder at one site or fewer.

Even if individual rankings change with better reporting, the directional signal is clear: reclaimed water is no longer niche. It is becoming a repeatable siting and design pattern among leading operators.

Koomey also flags an important caveat. Some facilities that do not use reclaimed water may use no onsite water for cooling at all, but that approach is generally less energy efficient than water-based cooling would be.

This is the heart of the policy dilemma now facing operators and regulators alike:

• In drought-strained regions, “no onsite water” may be the only socially acceptable option.

• From a grid and carbon perspective, however, higher-energy cooling approaches can shift impacts elsewhere; potentially increasing indirect water use through electricity generation, along with associated emissions.

As a result, data center operators are increasingly turning to reclaimed municipal wastewater as a “license to operate” tool particularly in regions where evaporative cooling remains the most energy-efficient path, but potable water withdrawals are politically radioactive.

Meta + ION Water: Meta Takes Water Stewardship to a Community Scale

A January 15, 2026 announcement between ION Water and Meta may read at first glance like a feel-good CSR story centered on affordable housing. In practice, it reflects a more deliberate shift in how hyperscalers are beginning to approach water stewardship.

The partnership launches a conservation initiative in the Brazos River Watershed, with Meta sponsoring the deployment of ION’s end-to-end water management platform across properties owned by a national affordable housing provider. The project is expected to conserve roughly 26 million gallons of water over five years, supported by what the companies describe as long-term, verifiable measurement.

Rather than focusing narrowly on the data center fence line, the initiative targets demand reduction within the same watershed where Meta operates, reframing water stewardship as a system-level intervention rather than a site-specific efficiency claim.

As Eric Homberger, Chief Commercial Officer of ION Water, explains:

Finding Water by Eliminating Waste: Leakage as a Hidden Demand Driver

ION Water and Meta frame leakage not as a marginal efficiency issue, but as one of the largest and least visible sources of water demand. According to the release, more than half of the water paid for at some properties can be lost to “invisible leaks,” including running toilets, aging water heaters, and faulty fixtures that go undetected for extended periods.

ION’s platform is designed to surface that hidden demand. By monitoring water consumption at the unit level, the system flags anomalies in real time and directs maintenance teams to specific fixtures, rather than entire buildings. The company says this approach can reduce leak-driven water waste by as much as 60%.

This represents an important evolution in how hyperscalers defend and contextualize their water footprints:

• Instead of focusing solely on their own direct WUE metrics, operators are investing in demand reduction within the same watershed where their data centers operate.

• That shift reframes the narrative from simply managing active water consumption to actively helping stabilize stressed local water systems.

The Accounting Shift: Volumetric Water Benefits (VWB)

The release explicitly positions the project as a model for Volumetric Water Benefits (VWB) initiatives, projects intended to deliver measurable environmental gains while also producing operational and financial benefits for underserved communities.

This framing aligns with a broader stewardship accounting movement promoted by organizations such as the World Resources Institute, which has developed Volumetric Water Benefit Accounting (VWBA) as a standardized method for quantifying and valuing watershed-scale benefits.

Meta is explicit that the project supports its water-positive commitment tied to its Temple, Texas data center community. The company has set a 2030 goal to restore more water than it consumes across its global operations and has increasingly emphasized “water stewardship in our data center communities” as a core pillar of that strategy.

This is a more mature form of “offsetting,” for three reasons: it is in-watershed, metered and operationally grounded through leak detection, and politically constructive, delivering tangible benefits in the form of lower water bills and improved reliability for affordable housing providers and residents.

From Buildings to Cities: Xylem and Amazon Take Leakage Reduction to Municipal Scale

If Meta and ION demonstrate how hyperscalers can drive water savings at the building level, a separate partnership between Xylem and Amazon shows what the same stewardship logic looks like at municipal scale.

In September 2025, Xylem and Amazon announced smart water infrastructure upgrades in Mexico City and Monterrey designed to reduce system-wide leakage and improve water reliability in two of North America’s most water-stressed urban regions. The projects deploy Xylem Vue, a data and analytics platform that uses real-time monitoring, pressure management, and advanced leak detection to cut non-revenue water, i.e. water that is treated and paid for but never reaches end users.

The scale is material. The two cities estimate combined water savings of more than 1.3 billion liters per year, including roughly 800 million liters annually in Mexico City alone, where officials acknowledge that as much as 40% of municipal water can be lost through aging pipes before it ever reaches the tap.

The intervention is conceptually similar to Meta’s work with ION Water, but applied one layer upstream. Instead of unit-level fixtures, the focus is on distribution networks, where pressure fluctuations and undetected leaks quietly drive chronic losses. By dynamically managing pressure and pinpointing leaks in real time, utilities can recover large volumes of water without securing new supplies, often the most politically and economically difficult option.

For hyperscalers, the implication is significant. Rather than defending water use solely through site-level efficiency metrics, Amazon is backing in-watershed, infrastructure-level demand reduction that directly improves water security for the communities where it operates. Xylem points to research suggesting that smart water systems can reduce the cost of urban drought resilience by as much as 20%, reinforcing the idea that conservation can be both an environmental and economic lever.

Taken together with Meta’s partnership with ION Water, the message is becoming clearer: water stewardship is expanding beyond the data center fence line. Hyperscalers are increasingly investing in measurable, verifiable reductions in system-wide water losses not just to offset their own demand, but to stabilize the underlying infrastructure on which their long-term growth depends.

The Friction Isn’t Going Away

Even with these emerging strategies, the friction points are not disappearing. If anything, AI densification will intensify them. At its core, the challenge remains a set of trade-offs between water and energy use. Dry or air-based cooling can ease local water stress, but often at the cost of higher electricity demand; shifting pressure onto constrained grids and, in some regions, increasing indirect water use and emissions if generation remains fossil-heavy.

Disclosure Remains Patchy

Koomey’s methodology, e.g. relying on utility contracts, mapping tools, satellite imagery, and secondary disclosures, is a polite way of underscoring a harder truth: public reporting is still inadequate. As regulatory scrutiny and community resistance harden, operators will be pushed toward more standardized, site-specific disclosure. That likely means greater transparency around WUE, water source types, seasonal variability, discharge handling, and watershed context, not just high-level commitments or aspirational targets.

“Water Positive” Will Face Offset-Level Scrutiny

Claims of being “water positive” are beginning to face the same skepticism that carbon offsets now encounter. The industry’s move toward Volumetric Water Benefits and VWBA reflects that reality. Going forward, assertions that “we replenished X gallons” will increasingly need to demonstrate:

• Locality — Benefits delivered in the same basin,

• Additionality — Outcomes that would not have occurred otherwise.

• Durability — Impacts that persist over time.

• Verification — Metered, auditable results.

The emphasis on real-time measurement and verifiable outcomes in projects such as Meta’s partnership with ION Water, and Amazon’s municipal-scale work with Xylem, signals an awareness that credibility, not volume alone, will define success.

Taken together, these developments point to a clear shift in doctrine. The data center industry is moving beyond a narrow “reduce gallons” mindset toward a more integrated operating model:

• Use non-potable and reclaimed sources wherever feasible.

• Engineer high-reuse, low-discharge water systems onsite.

• Reduce demand and recover losses at building and municipal scale.

• Quantify and verify watershed benefits that matter to local communities.

In a world where AI infrastructure is pushing fastest into regions where water is tightest, sustainable water usage is becoming less about any single technology choice and more about system design: source selection, closed-loop engineering, operational intelligence, and credible stewardship accounting working together.

Operators who can integrate all four will face fewer entitlement hurdles, lower regulatory risk, and over time better access to capital and a stronger license to operate. In water-constrained regions, these approaches are quickly shifting from optional best practices to political and economic necessities.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.