Data Center Water Resilience Modeling: Why WUE Is Not Enough

WUE can look efficient while a project still fails under heat, scarcity, discharge limits or reclaimed-water constraints.

Data center water resilience modeling asks a harder question than water use effectiveness. It asks whether the facility can keep operating when water is scarce, heat is high and local systems are under stress.

That distinction now matters in site selection.

Data Center Knowledge's 2026 analysis of water resilience argues that average WUE says little about whether a facility can hold cooling setpoints during peak stress. The article points to the proposed 330 MW Imperial Valley AI data center in California, which was originally presented as relying on reclaimed wastewater before later seeking Colorado River water access. The Imperial Irrigation District denied that request in May 2026, creating a live legal and permitting risk around the water strategy.

That is the lesson. A low WUE number is not the same as a financeable water plan.

WUE measures efficiency, not resilience

WUE is useful. It measures liters of water consumed per kilowatt-hour of IT energy. It helps compare cooling strategies and track operational efficiency.

It does not answer the developer's full diligence question.

A site with strong average WUE can still fail if reclaimed water is unavailable at the required volume, if discharge permits are constrained, if peak cooling demand coincides with drought restrictions or if local politics turn water into an entitlement issue.

Water resilience requires a site-specific model covering supply, quality, discharge, peak stress and competing users. It is closer to power diligence than sustainability reporting. The developer needs to know whether the water strategy survives the worst month, not whether it looks efficient over the year.

Peak demand is the risk developers underwrite

The Potomac River basin shows why averages mislead.

The Interstate Commission on the Potomac River Basin reported in March 2026 that data centers in the Washington Metropolitan Area used about 4 million gallons per day on average in 2025, with peak-day use of about 15 million gallons per day. Data centers represented only 1% of total withdrawals in the area, but 9% of annual consumptive use and up to 12% of summer consumptive use. The commission also projected that average data center water use in the area could rise to about 22 million gallons per day by 2050, with peak use above 80 million gallons per day under baseline assumptions.

Those numbers change the diligence lens. The political and infrastructure risk sits in peak conditions, not average consumption.

A development team should model:

• Average daily use under normal operating conditions.

• Peak-day use during heat events.

• Seasonal overlap with municipal, agricultural and industrial demand.

• Local drought rules and curtailment authority.

• Cooling technology tradeoffs between water and power.

• Discharge capacity, treatment requirements and thermal limits.

If the model stops at WUE, it misses the stress case.

Reclaimed water is not automatically available water

Reclaimed water can be a strong strategy. It is not a checkbox.

Developers need to prove volume, distance, treatment requirements, quality, delivery infrastructure and backup supply. Reclaimed water with high total dissolved solids, biological load or unreliable delivery can require treatment upgrades that change both capex and schedule. A reclaimed source that exists regionally but cannot serve the parcel at peak load is not bankable supply.

The same applies on the discharge side. Cooling systems produce blowdown and other wastewater streams. Sewer capacity, treatment plant limits, permit conditions and thermal discharge restrictions can cap operations even when intake water is available.

The practical diligence question is: can the site take in, treat, use and discharge water under peak conditions without depending on a best-case assumption?

AI can connect water evidence before site control

Water diligence is fragmented. The evidence sits in utility plans, watershed studies, drought rules, public meeting minutes, cooling design assumptions, treatment capacity reports, local ordinances and environmental filings.

AI helps by connecting those inputs into a living water risk model. For a development team, that model should do four things:

1. Extract local water source constraints from utility plans, basin studies and permit records.

2. Compare cooling options against power use, water use, discharge requirements and capex.

3. Flag reclaimed-water assumptions that lack volume, quality or delivery evidence.

4. Monitor public records for changes in water rules, opposition, moratoriums or infrastructure capacity.

An AI-native operating partner like Build can make this useful by tying the water model back to site selection, entitlement strategy and underwriting. The output should be a go, no-go or investigate decision, not a sustainability slide.

Human judgment still owns the water strategy

AI can find weak assumptions. It cannot negotiate water rights, redesign cooling or read the local politics better than the project team on the ground.

Engineers still own the cooling design. Counsel still reviews water rights, discharge permits and entitlement exposure. Development executives still decide whether a site with water risk is worth controlling. Community strategy remains human because water is often where technical feasibility becomes political legitimacy.

The right division of labor is clear. AI keeps the evidence current and forces the hard questions early. Humans make the call.

Water resilience belongs in first-pass site screening

Data center developers already screen power before they screen everything else. Water now deserves the same discipline in certain markets.

That does not mean every data center is a water problem. Some climates, designs and operators can run with low direct water use. It does mean water resilience should be modeled before site control where cooling demand, watershed stress, reclaimed-water dependence or community scrutiny is material.

WUE is a metric. Water resilience is an underwriting question. Developers who confuse the two will find out late, when the site is already expensive to abandon.