What should communities ask before approving a new AI data center? A water-risk checklist

The 10-question checklist

1. How much water will the project use on an average day?

Start with the basic number, but do not stop there. Average-day demand is the baseline for utility planning, but it can hide the times of year when the system is actually under stress.

2. What will peak-day or peak-season water demand be?

This may be the more important number. A facility that looks manageable on an annual average can still create major strain during heat waves, drought restrictions, or summer demand peaks. If the operator cannot provide a credible peak-demand estimate, the review process is incomplete.

3. Will the site use potable water, reclaimed water, or both?

This is one of the fastest ways to distinguish between a weak and a serious proposal. If reclaimed water is available and technically viable, the public should know whether the developer evaluated it and why it was or was not chosen. Amazon's recycled-water expansion shows that this is technically and operationally plausible at scale in at least some markets.

4. What type of cooling system will be used?

This question shapes almost everything else. Communities should ask whether the site will rely primarily on:

• Outside-air cooling
• Evaporative cooling
• Closed-loop systems
• Hybrid systems
• Chilled-water systems

The water, energy, and wastewater profile can look very different depending on that choice.

5. What happens during drought or utility restrictions?

This is where many project reviews get vague. The community should ask for a written drought and curtailment plan that answers:

• Will the facility reduce load?
• Switch water sources?
• Draw on stored water?
• Compete with residential demand?
• Trigger emergency utility upgrades?

A January 2026 HARC white paper estimates Texas data centers will consume 25 billion gallons of water in 2025 and could rise to 29 to 161 billion gallons by 2030, while noting that over 80% of Texas was in drought in April 2025 and 17% was in exceptional drought.

6. What wastewater will the facility generate, and where will it go?

Do not let the conversation stop at withdrawals. A data center that uses evaporative cooling may generate concentrated blowdown that needs to be managed through pretreatment, discharge, reuse, or other pathways. TNFD explicitly warns that wastewater from evaporative cooling can leave behind high concentrations of salts, heavy metals, and other pollutants when mismanaged.

• What is the expected discharge chemistry?
• What pretreatment is required?
• Which utility or receiving water body will take it?
• Has the receiving system assessed cumulative load if more facilities are coming?

7. What local data will be public after approval?

This is one of the most important questions, because it determines whether the community can verify what was promised. A UK government-linked AI and water report recommends mandatory, location-based reporting and more explicit integration of water planning into data-center and AI infrastructure development.

At minimum, public reporting should include:

• Annual and peak water demand
• Source-water type
• Cooling type
• Discharge volumes
• Major changes to operations
• Drought-response measures

8. Has the project been evaluated at the basin level, not just the parcel level?

A single project may appear manageable in isolation. A cluster of projects in the same basin can be a different story. A 2026 PLOS Water paper warns that data-center development can undermine water governance, contribute to unsustainable water use, reduce flexibility in water decision-making, and increase water use across scales through electricity demand.

That means a permit review should ask not only 'Can this site work?' but also 'What happens if five more sites are approved nearby?'

9. Is the company's local story consistent with its public sustainability story?

If the company says it is water positive, using recycled water, or minimizing strain on utilities, the permit process should ask for evidence that those claims apply to this specific site. Broad ESG language is not enough. The local utility agreement, engineering documents, and reporting commitments are what matter.

10. What is the fallback if assumptions fail?

Every review process should ask what happens if:

• Recycled-water supply is interrupted
• Demand runs higher than modeled
• Wastewater treatment is more difficult than expected
• Drought restrictions tighten
• Neighboring projects move forward sooner than expected

If there is no clear fallback plan, the water-risk review is not complete.

The short version: ten things to ask for in writing

1. 1Average daily water demand
2. 2Peak-day or peak-season water demand
3. 3Potable vs reclaimed-water split
4. 4Cooling system description
5. 5Drought contingency plan
6. 6Wastewater and blowdown handling plan
7. 7Pretreatment and discharge details
8. 8Public reporting commitments
9. 9Basin-level cumulative impact assessment
10. 10Evidence that local operations match public sustainability claims

Why this matters for Water Utility Report

Water Utility Report is not a permitting website. But it is a water-intelligence website. That means when readers hear that a major AI project is coming to their region, they should leave with sharper questions — not just stronger opinions.

The most useful water reporting is not the reporting that says 'be worried' or 'don't worry.' It is the reporting that tells people what to verify. That is what this checklist is for.