When a Data Center Comes to Town: Nine Questions Every Local Government Should Be Ready to Answer

Much of the public conversation about data centers asks whether a community should want one. That is a legitimate policy question, and it belongs to elected officials and the residents they answer to. This guide is about the question that comes first: whether the jurisdiction is able to answer it.

A jurisdiction that cannot classify a data center proposal, obtain the information needed to evaluate it, route it to the right reviewers, model its fiscal result, and document its findings does not actually choose among the available outcomes. It defaults to delay, improvisation, or approval under rules written for a different use. Those defaults can look like policy choices. They are usually the absence of one.

That is why review capacity is the neutral position, and why every participant has a reason to support it. Applicants benefit from knowing the review path, information requirements, decision criteria, and schedule before investing in a site. Residents benefit from a public record that addresses the actual project rather than analogies, promises, or fears. Staff benefit from defined roles, access to technical support, and findings that can be defended without reconstructing the file later. Elected officials benefit from having a real set of options. And local government attorneys benefit from an ordinance, review sequence, and evidentiary record that match the governing law.

Depending on governing state law and the jurisdiction's authority, four broad outcome types may be available. None is a point on a moral scale. Each simply requires different capacity.

A by-right system can be highly protective if its performance standards are strong. A discretionary system can be unpredictable if its criteria are vague. A moratorium can be careful or merely avoidant. A denial can be well supported or vulnerable. Process quality and policy outcome are separate things, and this guide evaluates only the first.

The threshold problem is classification. Many codes define a data center as an office, a warehouse, a utility, a communications facility, or a generic information processing use. Those labels may be serviceable for a small enterprise server room and inadequate for a campus with a large electric interconnection, extensive cooling equipment, battery storage, fuel storage, and emergency generation. When the label does not fit, the label quietly decides the case: the project is reviewed, approved, or opposed as something it is not.

A durable definition combines a use description with measurable intensity thresholds and anti-avoidance rules. Rather than distinguishing "AI" from "non-AI" facilities, which describes a workload that can change without any land use event, a more stable structure uses three tiers: incidental computing equipment accessory to another principal use; a data center as a principal use; and a large-scale or high-intensity data center subject to additional review when one or more measurable thresholds are met. Candidate thresholds include anticipated electric peak load, nameplate backup generation capacity, aggregate floor area, cooling water demand, or battery capacity. Whatever the jurisdiction selects, the definition should aggregate phases, buildings, co-located users, and parcels under common control, should address conversion of existing buildings, and should prevent an applicant from avoiding a threshold by dividing a campus or filing in stages.

An application may arrive before a jurisdiction has adopted a specific data center definition, forcing the authorized official to classify the use under the law applicable to that application. For an unfamiliar, high-consequence use, the interpretation should be formal, written, and reviewed by the county or municipal attorney before issuance. Counsel's role is not to make the policy choice, but to identify the source and limits of the official's discretion, relevant prior interpretations, applicable appeal and vesting rules, and the record needed to treat similarly situated applicants consistently. Informal answers should be avoided because they can complicate reliance, consistency, and any later code amendment.

Policy snapshot cross-reference. Florida's 2026 statute supplies a measurable threshold and aggregation rules for its utility and water provisions; see the dated policy snapshot above. A statutory utility definition is a useful reference point, not a model zoning definition. Local land use definitions should use triggers connected to the effects the jurisdiction is authorized to regulate.

What survives challenge. A definition is stronger when it uses observable facts, states how multiple buildings and phases are aggregated, gives similarly situated applicants the same notice, and connects each review tier to locally regulated effects.

Siting is a systems question, not only a map exercise. It combines land use compatibility with the location and capacity of electric, water, wastewater, road, fiber, and emergency response systems. The same proximity that makes a site attractive can also concentrate effects: a parcel near transmission may reduce the need for new lines while sitting near residences, schools, agriculture, or constrained infrastructure. A defensible siting decision evaluates both sets of facts in the same record.

The evidence to assemble spans four layers: the plan designation, zoning district, and review path; the surrounding context, including adjacent uses, sensitive receptors, environmental constraints, and the cumulative concentration of similar facilities; the infrastructure systems, including existing and planned transmission, substations, water, wastewater, roads, and fiber, along with which authority actually controls each system's rates and capacity; and the land's opportunity cost for industrial, residential, agricultural, or conservation purposes.

Counterweights. Larimer County, Colorado paused to study suitable zoning before receiving any site-specific application. Loudoun County, Virginia, one of the country's largest and most established data center markets, moved new data centers in several industrial districts from by-right treatment to special-exception review in 2025. Fairfax County, Virginia adopted a permanent amendment in 2024 that pairs tiered size and district rules with measurable requirements for equipment enclosure, residential setbacks, distance from Metro stations, sound studies before and after construction, and building design, with specified relief available by special exception. Permanent standards can also preserve a predictable approval path: Fairfax retains by-right pathways for qualifying facilities in designated industrial districts, and the City of Chandler, Arizona permits primary data center uses in planned area development districts subject to operational standards. These examples show different timing and different tools, not a national best answer.

What survives challenge. Siting rules should rely on mapped or measurable criteria, treat comparable uses consistently, respect state utility and environmental preemption, and explain why the selected districts and procedures correspond to locally regulated effects.

A moratorium is an interim legal tool, not a substantive data center policy, and it is only as neutral as its work plan. A pause with a defined study scope, assigned staff, public milestones, draft standards, hearing dates, and an end date creates a predictable decision framework. A pause without those elements simply moves uncertainty to the expiration date.

Before adoption, the jurisdiction should be able to answer a short list of questions. What legal authority permits the pause, and what procedures apply? What specific planning or regulatory gap requires study? Is there a pending application, permit series, or claimed vested right, and how will each be treated? What applications, conversions, expansions, and public facilities are covered or exempt? What is the start date, end date, extension process, and early termination trigger? What deliverables, hearing dates, staff assignments, outside expertise, and budget are approved with the pause? And what happens if the permanent ordinance is not adopted by expiration?

That last question deserves emphasis because the expiration asymmetry is unforgiving. The legal pause ends on its stated date whether or not the research, drafting, notice, and training are finished. An extension may be lawful in some jurisdictions and vulnerable in others, but an extension is not a substitute for production. Monthly public milestone reports make delay visible while there is still time to correct it. Treat the expiration date as a delivery deadline, not a placeholder.

Counterweights. Larimer County, Colorado adopted its pause before any site-specific application arrived, published the questions it intended to study, and later extended it while drafting continued; the extension ran to a fixed date or until permanent standards were adopted, whichever came first, a useful way to build early termination into the instrument. Boone County, Indiana tied a one-year pause to a comprehensive plan and zoning ordinance that predated modern facilities. The Town of Merrillville, Indiana paired its pause with an official explanation that expressly recognized both possible fiscal benefits and unresolved infrastructure, noise, traffic, and compatibility questions. Montgomery County, Maryland enacted an 18-month pause on accepting certain data center applications, creating a defined period for further study and potential standards. Flagler County, Florida broadened its moratorium's scope during adoption, from large-scale facilities to all data centers, deciding part of the definitional question before permanent standards existed. The same instrument, five different work plans.

What survives challenge. The record should identify a legitimate planning purpose, use a defined and proportionate scope, address pending and vested applications under state law, set a real end date, and show steady work toward the stated deliverables. Moratorium authority and notice requirements are state-specific, and local counsel should verify both before first reading.

A project cannot be evaluated from an investment announcement or a conceptual site plan. The completeness checklist should produce decision-grade information covering the full planned campus, not only the first building: who controls the project and answers for each condition; every building, phase, and parcel through maximum buildout; initial and full-build electric demand and interconnection status; cooling design and water use; generators, batteries, fuel, and air permit status; modeled sound at property lines and sensitive receptors; emergency response and traffic; the fiscal submission by taxable category, with incentive assumptions and project-linked public costs; and an operations and closure plan with a decommissioning cost estimate and financial assurance.

One discipline matters more than any single line item: label every utility statement by its actual stage. An interconnection request proves that a request was made. It does not prove available capacity, a completed study, a construction schedule, or an allocation of upgrade costs. Load forecasting, interconnection, resource planning, procurement, and cost allocation are related but distinct functions, and a public record that collapses them will misinform everyone who relies on it.

Confidential projects and public review. Many data center proposals arrive under code names through economic development channels, with some information lawfully protected for a limited period. Confidentiality statutes should be read record by record; they do not make every zoning application, staff analysis, utility fact, or hearing exhibit confidential. Two obligations run in parallel. Confidentiality may lawfully protect legitimate negotiations and security information. Transparency protects the legitimacy of the public decision. A disciplined record allows both to operate inside their legal boundaries. For every withheld item, the file should state the specific legal authority, the custodian, the expiration or release trigger, and whether a redacted version, range, aggregate, or staff description can lawfully substitute. Applicable public records law, not a private nondisclosure agreement, determines whether a record may be withheld. Counsel should advise the records custodian on each claimed exemption. The jurisdiction should publish a nonexempt project sheet with the applicant of record, the property, the requested approval, the maximum physical buildout, the review criteria, and the hearing schedule, and should release formerly exempt records promptly when protection ends. "Confidential" should never function as an undocumented application status.

What survives challenge. A completeness checklist is strongest when each required item maps to an adopted review criterion, is applied consistently, and is paired with a lawful process for claiming and resolving confidentiality.

Generic industry averages are context, not project findings. A fair review separates construction from operations, the initial phase from full buildout, direct effects from indirect effects, and mitigation commitments from enforceable conditions. It then works topic by topic, pairing every potential concern with the project evidence or mitigation that could answer it, and every projected benefit with a verification method: electricity, water, sound, air, land and visual effects, traffic, emergency services, workforce, and fiscal result.

Water shows why project specific evidence matters more than category level claims. Lawrence Berkeley National Laboratory has found that water use per unit of computing work can vary by more than four orders of magnitude depending on cooling design, climate, server efficiency, grid conditions, and operating choices. The defensible conclusion is not that every data center is water intensive, and not that any is water light. It is that the jurisdiction needs the proposed design and operating case for this project, at this site, under these conditions. The same logic applies to sound, which depends on equipment selection, layout, and testing regimes, and to air, where emergency generators may be subject to federal standards and state or local permits. Staff should identify the actual equipment and the permitting agency rather than assuming that backup means unregulated, or that an air permit by itself establishes land use compatibility.

Benefits deserve the same discipline as burdens. Construction employment should be separated from permanent employment, with the duration of each job category stated. Tax base claims should carry the exemption, abatement, depreciation, and recipient detail developed in Question 08. A projected benefit without a verification method is a hope, not a finding.

What survives challenge. Findings should cite the submitted study, the staff or peer review conclusion, the adopted criterion, and the factual basis. Competing expert evidence should be acknowledged and resolved in the record, not ignored. A body that explains why it credited one analysis over another is in a far stronger position than one that simply counted voices.

The record is the deliverable. The useful product of review is not simply a yes, a no, a pause, or a conditional approval. It is a traceable record showing what was asked, what was submitted, which agency had authority over each issue, how conflicting evidence was treated, what findings were made, and how ongoing obligations will be verified. Everything about team structure and sequence serves that record.

No single department owns the relevant authority, so the central management task is producing one evidence record from many reviewers. The team runs from planning, legal, utilities, engineering, fire, and environmental review through finance, the assessor, economic development, and the clerk, coordinated by one project manager with authority over the master schedule and issues list, and reinforced by independent technical experts whenever existing staff cannot verify a material study. The sequence that keeps them aligned runs from legal intake (vesting, confidentiality, conflicts) through full-campus completeness review, concurrent technical review against one shared questions log, peer review, public release of nonexempt materials with enough time for meaningful review, a staff report organized by adopted criteria, the hearing and findings, and finally a condition handoff to inspection, monitoring, finance, and enforcement.

Two disciplines protect the record. First, an evidence index: a running table that maps every adopted criterion to the applicant evidence, the agency review, the public evidence, the staff conclusion, the finding, and the condition. Publish the nonexempt version with the staff report. Second, consistent status labels. An adopted ordinance, an official record, an applicant-stated projection, a third-party analysis, a reported fact, and a pending action are different kinds of evidence, and the record should never let an applicant projection become an official finding through repetition, or a pending ordinance be described as adopted.

Where the decision is quasi-judicial, process discipline extends to the decision makers themselves. Officials should seek counsel's guidance on prior public statements, ex parte communications, conflicts, disclosure, and recusal before participating, and counsel should prepare the body for the applicable evidentiary standard before the hearing. In Florida quasi-judicial matters, findings commonly require competent, substantial evidence; other states use different standards and procedures.

What survives challenge. Use the jurisdiction's actual evidentiary standard and hearing rules, resolve material contrary evidence on the record, and make sure the findings can be located in the file rather than reconstructed from memory.

Conditions convert important assumptions into measurable obligations. A useful drafting formula: regulated subject, plus measurable threshold, plus measurement method, plus frequency or trigger, plus reporting recipient, plus correction period, plus enforcement consequence, plus responsibility after transfer. A condition missing any element will be hard to administer five years later, and if staff cannot explain how compliance will be determined then, the condition is incomplete now.

The condition library for this use typically covers phasing tied to completed infrastructure; caps on the electric, water, generator, and floor area assumptions that support the findings; sound baselines, commissioning tests, generator testing windows, and complaint-triggered retesting; water metering and drought response; site design and as-built certification; emergency planning and incident reporting; annual reporting where authorized; a material-change trigger; successor responsibility on transfer; and decommissioning milestones with security. Chandler's ordinance, effective January 2023 and introduced above, offers a useful operational example: it pairs a preconstruction sound baseline with post-construction annual testing for five years, limits generator testing windows, and requires notices and a community liaison. It does not establish the correct numeric limit for any other jurisdiction.

Resource intensity should be documented as four distinct things: existing local conditions, applicant projections, external benchmarks, and approval limits. For electricity, water, wastewater, sound, traffic, and emergency services, the record should identify the existing baseline, the project's average and peak demand by phase, the assumptions behind each projection, and, where authorized and feasible, the metric that will be measured after construction. Data from comparable operating facilities can test whether projections are reasonable, but only when normalized for scale, cooling technology, climate, utilization, and buildout. When a projected impact supports a material finding, the approval should state whether the number is a binding limit, a monitoring baseline, or context only. That distinction prevents a projection from becoming either an unenforceable promise or an unintended restriction.

Name the instrument. "Community benefits agreement" should not be used as a catch-all. A zoning condition, a development agreement, a utility agreement, a tax incentive agreement, an infrastructure reimbursement agreement, and a private agreement with community organizations differ in parties, authority, consideration, public process, enforcement rights, and successors. A broad benefit should not excuse failure to satisfy an adopted land use criterion, and an impact-related requirement should not be described as a voluntary gift if it is a condition of approval. For every negotiated commitment, the public file should identify the legal instrument, the approving authority, the trigger, the recipient and amount, and the reporting and enforcement provisions.

There is also a constitutional boundary. Conditions need statutory and code authority, a factual connection to the project, and a feasible enforcement path. When a condition requires land, money, or an improvement as an exaction, counsel should apply the governing nexus and proportionality tests. In Sheetz v. County of El Dorado (2024), the U.S. Supreme Court confirmed that legislatively imposed permit conditions are not categorically exempt from that scrutiny.

What survives challenge. Each condition should cite the finding it implements, and the whole set should read as the operational translation of the record rather than a negotiated afterthought. An unusually long list of customized conditions may itself signal that the underlying code does not fit the use. The durable response is to revisit the code.

The announcement number is not the tax base. A fiscal finding should reconcile five different numbers that are often confused: announced investment; investment physically placed in the jurisdiction; taxable value after exclusions and incentives; gross revenue received by the specific deciding jurisdiction; and net fiscal effect after direct service and capital costs. It should also keep three ledgers separate. The deciding jurisdiction's budget result, the utility and ratepayer result, and broader regional economic activity are different analyses, and a positive answer on one does not establish a positive answer on another. Present each where relevant, and never net one against another.

A fully labeled hypothetical shows why the casual shortcut misleads. Suppose a project announces $5 billion in total investment, and a casual estimate applies a 1 percent local tax rate to produce $50 million per year. Now run the actual mechanics. If $1 billion is land and buildings, taxed at 70 percent of value after abatement, the annual yield at 1 percent is $7 million. If $4 billion is equipment, of which 20 percent remains taxable after exemption, assessed at an average 50 percent factor, the yield is $4 million. Illustrative-year gross property tax revenue is $11 million, not $50 million. If project-linked recurring and annualized capital costs for inspection, emergency capacity, roads, and administration run roughly $3 million per year, an assumption each jurisdiction must build from its own budgets rather than borrow, the illustrative net local effect is about $8 million per year before indirect effects and risk adjustment. That is a meaningful number, and it is a fraction of the headline arithmetic. The hypothetical describes no real project or jurisdiction; a real model must use the applicable tax base, rates, recipients, exemptions, negotiated package, assessment schedule, and phasing.

Location drives the mechanics. Loudoun County, Virginia's 2026 assessment schedule moves data center computer equipment from 60 percent of original cost when newest down through 45, 30, 15, 10, and 5 percent as it ages, while Virginia separately offers a qualifying sales tax exemption for data center hardware and software. Indiana's program can exempt qualifying equipment and energy purchases from sales and use tax for up to 25 years, or up to 50 years above a stated investment threshold, with a local personal property exemption also available. The lesson is to model each tax and each exemption separately, by asset cohort and year, and to run the downside cases: delay, reduced buildout, faster depreciation, greater exemption, and early closure, alongside the announced case.

What survives challenge. A fiscal finding should disclose its assumptions, source dates, recipients, incentives, and sensitivity cases, and should never combine gross regional activity with revenue to the deciding jurisdiction.

The operating period may last decades while equipment, tenants, owners, cooling systems, and load change. The approval must define which changes are routine and which reopen review, and approval is therefore the start of a compliance file, not the end of one.

The post-approval program runs from a preconstruction meeting and final condition matrix through as-built certification, commissioning tests, annual certifications of ownership, equipment, water use, and incidents, a public complaint channel, inspection authority, successor acknowledgement before transfer, and closure requirements ending in site restoration and release of financial assurance. A material-change screen keeps the file honest: a proposal should return for the authorized level of review when it changes maximum or staged electric demand, campus area or phasing, cooling technology or water use, generation or storage equipment, modeled sound sources, emergency assumptions, infrastructure responsibility, the responsible owner, or closure timing and security. A change is not automatically material because equipment is newer, and it does not escape review merely because the building envelope stays the same.

Two jurisdictions illustrate the bookends. Chandler's ordinance shows continuing verification, with post-construction sound testing on a defined schedule. Susquehanna County, Pennsylvania's adopted ordinance expressly requires decommissioning funds and liability insurance, a reminder that closure risk belongs in the original application and approval record, while salvage values and cost estimates should be revisited as conditions change.

What survives challenge. Ongoing obligations should be authorized, objective, recorded where appropriate, assigned to a responsible department, funded, and enforceable against successors.

The first 72 hours after a serious inquiry. The first few days often determine whether departments work from the same project and whether informal statements outpace the record. Within 24 hours: assign a project manager and counsel contact, open one project file, record what was asked and what confidentiality was claimed, preserve the rules in effect on the inquiry date without implying that the inquiry creates rights, and instruct staff not to characterize approval likelihood, utility capacity, tax benefit, or vested rights before review. Within 48 hours: convene the interdisciplinary team, map who has authority over what, and issue one coordinated information request. Within 72 hours: give the applicant a written review path with required submissions and the confidentiality procedure, open the evidence and issue logs, and set the next checkpoint with named deliverables.

A 90-day sequence for jurisdictions without an active application. Days 1 to 15: authority, definition, vesting, pending application, and confidentiality review with counsel. Days 16 to 30: infrastructure contacts, entitled site inventory, reviewer roster, and outside expert contracts. Days 31 to 50: the application schedule, evidence index, fiscal assumptions sheet, and public project sheet template. Days 51 to 70: definition and siting options, a condition library, the inspection handoff, and a moratorium contingency plan. Days 71 to 90: a tabletop exercise using a hypothetical campus, corrections, training, and a public adoption calendar.

Peer exchange. Written resources cannot capture every local circumstance. When a jurisdiction faces its first major data center proposal, a structured conversation with counterpart staff in both established and emerging markets can help surface practical questions about staffing, utility coordination, public engagement, and post-approval administration. State and national associations can help identify appropriate peers. Peer experience should inform the local inquiry, not substitute for the jurisdiction's own law, evidence, and findings. The National Association of Counties also maintains policy and research resources on data centers (naco.org).

The workload described in this guide is substantial, and it lands hardest on small and mid-sized departments. Structured, AI-assisted review systems can carry a defined share of it: comparing an application against a published completeness checklist; extracting applicant assumptions into a staff-controlled evidence matrix; tracking inconsistent numbers across plan sets and revisions; maintaining deadlines, assignments, and condition status; generating first-pass public summaries from already public records; and monitoring dated reporting obligations while flagging exceptions for human review.

The limits matter as much as the uses. AI should not decide credibility, resolve a legal exemption, make a land use finding, replace engineering judgment, or determine approval. Every extracted fact needs a source link and human validation, and confidential information needs approved access and retention controls. The technology's proper role here is the same as the guide's: making the review faster and more traceable without predetermining its direction. In Hernando County, Florida, SwiftGov has been used to accelerate routine single-family zoning review (case study); a complex data center review is not comparable to routine zoning work, and the realistic gains there are coordination, completeness, and a traceable record rather than speed alone.

Data centers are the immediate subject, but the administrative lesson is broader. New infrastructure-heavy uses often arrive before codes, staffing, and interagency procedures are ready, and the next one will not announce itself either. A reusable review system separates the stable questions from the volatile project and policy facts, producing faster and more defensible decisions without predetermining their direction.

Depending on state law and its own authority, a community may welcome a data center, approve it with conditions, pause to write rules, or decide that a proposal does not fit. Each of these, where available, is a legitimate outcome of a lawful process, and this guide takes no position among them. Neutrality does not mean withholding judgment. It means having the capacity and the record to make the chosen judgment on purpose, and to explain it to the applicant who invested in the site, the residents who live beside it, and the court that may one day read the file.