How Communities Can Evaluate AI Data Center Projects

A community cannot evaluate an AI data center from a project label, an acreage figure, or a promise of investment. It needs a defined operator, a site plan, resource forecasts, financial terms, decision authority, and commitments that remain enforceable after ownership or operating conditions change. Until those elements are disclosed, both projected benefits and projected harms are scenarios rather than project facts.

This distinction matters because early land and jurisdiction decisions can narrow later choices. Around Salix, Iowa, roughly 900 acres of farmland were annexed before a data center operator or development plan had been publicly identified. MidAmerican Energy described the site as under consideration for a future large customer, while the land still required later zoning and development steps. The Winnebago Tribe's public teach-in and a separate Woodbury County moratorium illustrate two ways communities can create time for questions. Neither action, by itself, proves that a specific project should be approved or rejected.

A better review is disclosure-first: identify the decision being requested, publish the evidence needed for that decision, allow independent scrutiny, and translate accepted claims into measurable conditions. The following framework turns that principle into a practical process.

Map the decision before debating the project

Start with a one-page decision map. List every stage that could move the proposal forward: land options or purchases, annexation, rezoning, utility study, incentive agreement, development approval, building permits, construction phases, and operating authorization. For each stage, name the public body or private party with authority, the documents it must produce, the notice period, and the opportunity for public response.

This prevents different milestones from being treated as interchangeable. A site review is not an application; an application is not an approval; and an approval is not evidence that construction will occur. In Salix, the utility's affiliate held land options that might be transferred to a future customer. That arrangement signaled serious site preparation, but it did not identify an operator or establish a final campus design.

The map should also show which government controls the parcel. Woodbury County's 12-month pause applied to unincorporated land and therefore did not stop a project inside Salix. Jurisdiction determines which rules, hearings, and officials matter. Communities should resolve that question before relying on a county or city action that may not cover the actual site.

Build a public evidence register

Create a versioned register with three labels: confirmed, proposed, and unknown. Confirmed entries should link to a filing, public record, executed agreement, or named official statement. Proposed entries should identify who supplied the estimate and the assumptions behind it. Unknowns should remain visible instead of being filled with industry averages.

The operator's identity is the first high-value disclosure. Without it, a community cannot examine an operating history or obtain binding commitments about equipment, cooling, construction phases, staffing, or compliance. Generic descriptions of hyperscale or AI facilities can help residents form questions, but they cannot substitute for a project-specific design.

Require revisions to be logged. If maximum electrical demand, water use, building count, or requested incentives change, reviewers should be able to compare versions and see whether earlier analyses still apply. Publish written answers to material questions before a vote, not only oral responses during a hearing.

1. Land: test control, compatibility, and reversibility

The land review should document parcel boundaries, current zoning, ownership or option status, adjoining uses, setbacks, and the reason the site fits the community's long-term plan. If annexation has occurred, publish its notices, timeline, map, and stated purpose. A change in municipal control can shift zoning authority and should be evaluated as a consequential decision in its own right.

Ask what the maximum buildout could place on the acreage, including server buildings, substations, transmission connections, cooling equipment, backup generators, fuel or battery systems, roads, fencing, and stormwater facilities. Do not infer those elements from acreage alone; require them in the site plan.

Conditions should survive a sale or transfer. The approval should state which obligations bind a successor, what happens to unused phases, and who pays for decommissioning or site restoration if construction stops or operations end.

2. Power: separate load from who bears the risk

Require initial and maximum load forecasts, phase dates, planned substations and transmission work, and the generation or network assumptions used in the utility study. The review should distinguish the facility's electricity demand from the infrastructure required to serve it.

A promise that the customer will pay for necessary generation, transmission, or substations is useful but incomplete without the agreement that allocates those costs. Ask who calculates incremental cost, which regulator or public body reviews the arrangement, and how other customers are protected if the campus expands more slowly than forecast or closes early. A large customer may support new investment, but the community needs the contract structure before deciding whether costs and risks are contained.

Set reporting points for actual demand by phase. If later approvals depend on earlier forecasts, material deviations should trigger review rather than automatically carrying forward.

3. Water: review the actual cooling design

Water claims must be tied to the proposed equipment and local system. Ask for normal, peak-day, and peak-month withdrawals; consumptive use; source capacity; discharge plans; drought assumptions; and any infrastructure upgrades. The applicant should explain whether the cooling system relies on evaporation, a closed loop, air cooling, or a combination, and how operations change during hot periods.

A comparison can reveal the right questions without predicting local use. Reporting on Microsoft's West Des Moines facilities found withdrawals of about 11.5 million gallons in July 2022, roughly 6 percent of the local water district's monthly use. The Iowa water case does not establish what a Salix facility—or any other proposal—would consume because the operator and cooling design may differ. It shows why reviewers need peak-period figures rather than an annual average or a generic claim about climate.

Make the forecast testable through metering and periodic public reporting. Define what happens when use exceeds the approved range or local supply conditions change.

4. Taxes: calculate net public value

Gross investment and taxable benefit are not the same. Request a year-by-year schedule showing assessed value, exemptions, abatements, rebates, infrastructure contributions, and expected revenue for each affected public body. Place projected public-service costs beside those receipts, including roads, emergency response, utility oversight, and administrative work.

Every incentive should state its duration, performance conditions, reporting method, and remedy if promised investment or operations do not appear. Financial protection may include phased incentives, repayment provisions, or commitments linked to verified milestones. The relevant question is not whether the project is described as large, but whether the disclosed agreement produces durable net value under realistic buildout scenarios.

5. Jobs: split temporary work from permanent employment

Employment figures should never combine construction labor with ongoing operations. Require separate ranges, expected timing, occupation types, and the party responsible for each estimate. Permanent staffing depends on campus scale, automation, security, maintenance, and the operator's model, so an unidentified operator cannot support a precise promise.

Communities can also ask which commitments are local and measurable: apprenticeship participation, local procurement, training partnerships, or reporting on filled positions. These terms should be evaluated alongside the project's labor demand rather than presented as a substitute for a credible permanent-job forecast.

6. Governance: design participation around documents

One public hearing is not a complete review process. Provide enough time for residents and affected tribal, municipal, county, utility, landowner, and labor interests to read the application and its revisions. Make technical appendices accessible, identify conflicts of interest, disclose proposed incentives before the decision, and commission independent analysis where the approving body lacks expertise.

The Salix debate shows why legal authority and public trust are different tests. Local reporting documented a large public response after annexation and concerns about notice, while officials emphasized that no development plan was yet in place. A defensible process explains what was decided, what remains open, and when participation can still change an outcome.

Approval conditions should name an enforcing body, inspection rights, reporting intervals, and consequences for noncompliance. Aspirational statements about responsible development are not governance unless someone can verify and enforce them.

7. Community evidence: turn predictions into a record

Create a public dashboard or recurring report for the commitments that matter after approval: buildout status, electricity demand, water withdrawals, property-line noise tests, backup-system testing, tax and incentive performance, construction employment, permanent staffing, complaints, violations, and corrective actions. Report against the figures used during approval so residents can see variance, not just current totals.

Community evidence should include clear measurement methods. Noise, for example, depends on equipment placement, distance, terrain, barriers, and operating conditions. A condition should specify the property line, time period, test method, limit, and response to an exceedance. Similar precision is needed for water, phased capacity, and financial commitments.

The record should remain available if the property or operator changes hands. This converts a one-time decision into continuing accountability and gives officials evidence for later phases.

Use approval gates instead of a single yes-or-no vote

A phased project can be matched with phased authority. Preliminary review may establish whether a site is eligible for a complete application. Rezoning can be conditioned on defined uses and performance limits. Construction of a later phase can depend on verified compliance in the earlier one. Major changes in operator, cooling design, maximum load, incentive terms, or site layout should reopen the relevant review.

This approach neither assumes that every data center causes the same impacts nor asks a community to accept an undefined project. It lets evidence mature before commitments become difficult to reverse.

Practical implementation checklist

Before advancing a proposal, confirm that the public record includes:

• A named operator and a clear description of the utility, developer, landowner, and government roles.
• A decision map covering land control, annexation, zoning, utility agreements, incentives, permits, phases, and operating oversight.
• A site plan showing maximum buildout, related electrical and cooling infrastructure, setbacks, emergency systems, and successor obligations.
• Initial and maximum power forecasts, infrastructure cost allocation, regulatory review, and protection if demand falls short.
• Project-specific normal and peak water figures, the cooling design, drought assumptions, metering, and exceedance rules.
• A net fiscal schedule that includes exemptions, rebates, public costs, milestones, reporting, and remedies.
• Separate construction and permanent job ranges with assumptions and measurable local commitments.
• Accessible documents, adequate response time, independent technical review, written answers, and disclosed conflicts or incentives.
• Enforceable limits for noise, water, phased capacity, emergency systems, and decommissioning.
• A durable public reporting process that compares actual results with approved claims.

If essential entries remain unknown, the appropriate result is not necessarily rejection. It is a pause at the current gate, a precise disclosure request, and a decision only after the evidence matches the scale of the commitment.