Blazing New Trails: America’s New Data Center Laws Are Taking Shape, Town by Town and Coast to Coast

A look at how 100 local governments are making new laws and forging their own regulations to respond to constituents’ demands. It may spell bad news for Big Tech.  

Modern Data Centers didn’t exist ten years ago. Not many existed five years ago. 

But the recent onslaught of new proposals across the U.S. caught communities of all sizes in all regions off guard. Their local governments have been scrambling. 

They are working with strained budgets and no roadmaps. Constituents are breathing down their necks while tech companies with unlimited funds are pushing for more, bigger, and more powerful data centers. 

And a look at 100 local laws, passed over the past year, shows towns and counties across America are taking a bolder stand against how data centers can operate in their jurisdictions. The list is still taking shape, but it shows many mean business. They are writing new laws to handle it. They include: 

More refined noise ordinances. 

Huge setback requirements.

Power limits.

Community outreach.

Limits on testing emergency generators.

Annual reports on environmental issues, noise, water consumption and more.

If you want to know how a local government thinks about data centers, look at the “whereas” clauses in the resolutions used to pass their laws.

“It is plain that data centers – as they have developed in the 25 years since (local)… regulations were first added…– are energy-hungry land uses that are generally aesthetically displeasing and provide few public benefits,” city officials in Monterey Park, Calif., stated in their legislation. “Current data show that, overall, buildings consume 74 percent of all electricity in the United States.” 

Here’s what some are doing: 

• Atlanta, Georgia: Ordinance 24-O-1222 defines data centers and prohibits them on property located within 2,640 feet—one-half mile—of a high-capacity transit stop.
• Yorkville, Illinois: Ordinance 2025-53 requires 100-foot landscaped buffers, 500-foot separation from residential or commercial structures, noise studies and monitoring, restricted generator-testing hours, and a prohibition on nuclear power for data centers.
• Fairfax County, Virginia: The county’s zoning amendment establishes building-size limits, equipment screening, noise studies, a 200-foot residential setback for buildings, additional separation for ground equipment, and a one-mile setback from Metro entrances unless modified through a special exception.
• Chandler, Arizona: Ordinance 5033 permits principal-use data centers only through Planned Area Development zoning and requires neighborhood outreach, pre- and post-construction noise studies, continuing noise monitoring, and limits on backup-generator testing.
• zones subject to size, ownership, noise, water, and power-related requirements.
• Henry County, Georgia: ULDC Section 6.6.3 requires conditional-use approval and imposes a 20-acre minimum site, 200-foot buffers, 1,320-foot residential separation, noise limits, open-space requirements, equipment screening, and a binding development agreement.
• Stafford County, Virginia: Ordinance O23-24 establishes 100- to 200-foot setbacks near noncommercial or nonindustrial property, 50-foot vegetated buffers, mechanical-equipment noise mitigation, and post-development noise evaluations for ten years.
• Campbell County, Virginia: The county amended its zoning code to remove data centers as a by-right use in the Industrial-Heavy district and to require a special-use permit and public review.
• Shenandoah County, Virginia: The county’s ordinance requires special-use approval, environmental and utility assessments, water and cooling safeguards, setbacks and screening, generator-testing restrictions, and continuing compliance with specified A- and C-weighted noise limits.
• Lancaster, Pennsylvania: Ordinance 07-2026 allows data centers by special exception in SM zones and requires noise, air-quality, energy, water, heat-mitigation, emergency-response, and decommissioning plans, along with residential setbacks and landscaped buffers.
• Monterey Park, California: Ordinance 2253 defines data centers occupying at least one acre and permits them in the Saturn Park Innovation/Technology Zone only through an approved development agreement.
• Burleigh County, North Dakota: Section 30 requires data centers drawing at least 50 megawatts to obtain a special-use permit in an industrial district and imposes a one-mile sensitive-use setback, noise limits, generator restrictions, financial security, and decommissioning obligations.
• Limerick Township, Pennsylvania: Ordinance 438 regulates data centers as conditional uses, generally requires 1,000-foot residential setbacks unless reduced with sound mitigation, mandates closed-loop cooling and recurring noise studies, and requires financial security equal to 110 percent of decommissioning costs.
• Aurora, Illinois: Ordinance O26-022 subjects data centers to conditional-use review and establishes standards for residential separation, noise, water and energy efficiency, equipment screening, renewable energy or storage, and parking.
• West Rockhill Township, Pennsylvania: Ordinance 257 permits data centers by special exception only in the Planned Industrial district and requires a 35-acre site, large residential setbacks, sound studies, closed-loop cooling, utility-capacity analyses, extensive solar generation, and a funded decommissioning plan.
• Cochise County, Arizona: Ordinance 26-09 allows data centers only through special-use authorization in specified rural, business, and industrial districts and requires review of utilities, water consumption, noise, fire protection, emergency access, screening, and eventual decommissioning.

About ten percent of local governments that passed a new law require annual reporting on what a data center actually does after it opens: how much power it consumes, how much water it draws or how much noise reaches its neighbors.

The rest of the regulatory weight sits overwhelmingly at the front of the project. Developers must clear setbacks, describe cooling systems, secure utility letters, submit traffic and acoustic studies, plant trees, screen generators and show that fire departments can reach the site. Those rules matter. 

But most are snapshots taken when a locality’s leverage is strongest—before the permit is issued and before the certificate of occupancy releases a facility into decades of operation.

That imbalance says more about the state of data center regulation than any single setback. Local governments are learning to regulate one of the largest new industrial loads in America. 

For now, most still govern what is easy to measure once: distance, acreage, square footage and a single decibel figure. Far fewer govern what is difficult to measure repeatedly but matters to the public over time: sustained low-frequency noise, peak electrical demand, cooling-water consumption and whether the facility is operating as promised.

The outliers point toward the next version of the data center ordinance. They replace labels with operating thresholds, promises with verification and one-time approvals with life-cycle obligations.

A locality regulates the pressure it feels

The data center debate is often reduced to a contest over which place is “strictest.” That misses what local governments are actually doing.

Each place tends to regulate the externality it feels and the lever it controls.

Rural and exurban communities usually have abundant industrial land. Their scarce asset is separation from homes, farms, churches and schools—and their long-term fear is being left with a vast, obsolete building. Their ordinances therefore lean on large setbacks, minimum acreage, berms, vegetative screening and decommissioning bonds. These are fixed, visible standards that a small planning staff can enforce without running a permanent environmental monitoring program.

Bradley County, Tennessee, is the extreme version. Its rules require 3,000 feet between a data center and residences, churches, schools and other public gathering places. That is less a conventional buffer than a near-ban across much of a settled landscape. 

Menomonie, Wisconsin, created a dedicated data center industrial district with 400-foot front, side and rear setbacks. Chadds Ford Township, Pennsylvania, combined a 25-acre minimum with setbacks that can reach 750 feet, water controls, a mitigation fund and decommissioning security equal to 150 percent of estimated gross cost.

Cities with constrained water supplies or control over their own utilities reach for different tools. Longmont, Colorado, capped hyperscale load at the lower of 100 megawatts or 5 percent of the municipal power authority’s nameplate capacity. San Angelo, Texas, requires closed-loop cooling, a separate cooling-water meter and limits on initial fill, annual makeup water and full refills. Tucson tied siting to water, grid behavior and open-space mitigation, including a restriction on potable water for cooling and a two-hour battery-backup requirement.

Dense cities regulate geography for a different reason. Atlanta excludes new data centers from land within half a mile of high-capacity transit stops. The concern is not simply whether a server hall can be heard from a backyard. It is whether scarce, transit-rich land is being consumed by a low-employment use that could otherwise hold housing, offices or other job-dense development.

So the meaningful split is not rural versus urban, or permissive versus hostile. It is land, water, power and opportunity cost. Local governments are regulating whatever the buildout makes scarce—and whatever they can reach through zoning, a utility meter or a development agreement.

The permit is not the operation

Nearly every document in the review addressed zoning. About three-quarters addressed buffers or screening, 72 percent addressed noise and roughly two-thirds addressed setbacks. Those provisions shape where a project can go and what it must look like when plans are approved.

Only about 10 percent required annual operational reporting.

That is the life-cycle gap.

A data center can satisfy every preconstruction condition and still run differently than its application suggested. Server density can rise. Cooling equipment can work harder. Generator testing can become more frequent. Water use can climb during heat or drought. An acoustic model can prove wrong once transformers, fans and chillers operate together. Without a reporting duty, verification test or permit-reopener, the locality may not learn about the change until residents complain.

There are practical reasons for the gap. A government’s leverage peaks before approval and becomes harder to reassert after a developer has spent hundreds of millions of dollars. Continuous oversight consumes staff time and technical expertise that many counties do not have. Applicants resist open-ended obligations that can create future costs or enforcement exposure. It is easier for both sides to settle on a setback that can be drawn on a site plan.

But a setback does not report a megawatt.

Green Oak Charter Township, Michigan, is one of the places trying to close the gap. Its ordinance requires annual reporting on energy, water and noise. Little Rock, Arkansas, requires larger facilities to perform pre- and post-construction acoustic work, use permanent monitoring and file annual reports. 

Emporia, Kansas, uses a recorded development agreement to carry utility commitments, studies, mitigation, monitoring, reporting, decommissioning and phasing obligations forward with the property and its successors.

The same weakness appears at the end of a project’s life. Only 34 percent of the documents addressed decommissioning, and about 25 percent required financial assurance. That leaves many communities with no clearly funded plan for removing equipment, restoring land or dealing with an abandoned shell. A promise to decommission is useful. A bond, letter of credit or other security sized to the actual cost—and periodically updated—is enforceable.

The strongest ordinances, therefore, treat a data center less like a warehouse and more like a long-lived industrial system. Approval begins the relationship. It does not end it.

Square feet is losing the argument

Local zoning codes are comfortable with square footage. It is printed on building plans, familiar to assessors and easy to compare with a warehouse, shopping center or factory.

For data centers, it doesn’t work. 

A compact artificial-intelligence or high-performance-computing hall can consume more electricity and reject more heat than a much larger warehouse-style facility. Server racks are getting denser. Cooling systems and on-site generation change the site’s impact without changing the building envelope. A square-foot trigger can miss the smaller but more intense project while capturing a large building with relatively modest load.

It also invites relabeling. A developer can divide a campus into phases, split contiguous property among related entities or describe the use as cloud infrastructure, a computational facility or an advanced technology campus. If the rule regulates the name on the application instead of the facility’s function, the rule will age quickly.

Longmont addresses part of the problem by aggregating phased and related campuses when it applies its power cap. Green Oak goes further. It regulates a “High Resource Demand Facility,” a function-based category triggered by operating characteristics and thresholds such as more than 10 megawatts of demand or more than 100,000 gallons of water per day. The developer’s preferred label does not decide whether the rule applies. The load does.

That approach also survives the next technological turn. A cryptocurrency mine, AI training cluster or future high-density computing use can produce similar local effects even when none is called a data center. Regulation based on megawatts, peak demand, water draw, heat rejection or generator capacity follows the impact.

The harder question is which unit should control. Only about 13 percent of the documents examined used meaningful scale or type tiers, and the corpus showed no common threshold. Some use megawatts. Others use water, square footage, server counts or combinations. The field has not agreed on the unit of regulation, which means many bright lines remain arbitrary.

Potentially, the sharper model uses more than one measure, aggregates commonly owned or contiguous parcels and counts planned phases toward the threshold. It asks not how large the first building is, but what the full campus is designed to do.

Noise shows how far the rules still have to go

Noise is the complaint local officials hear and the technical issue many ordinances still handle poorly.

Almost three-quarters of the documents addressed noise. Only 31 percent supplied a numeric standard. Just 14 percent addressed C-weighting, infrasound or low-frequency sound.

It’s the difference between continued resident complaints and a solved problem. 

Data center noise is often continuous, tonal and concentrated at low frequencies. Chillers, transformers, cooling fans and generators can create a steady hum that carries differently from traffic or occasional industrial activity. Yet many ordinances rely on a single A-weighted decibel limit, or dBA. A-weighting discounts low-frequency energy to approximate aspects of human hearing. A facility can comply with a 55-dBA property-line limit and still produce the kind of persistent drone residents describe inside homes or at night.

C-weighting, or dBC, is flatter across low frequencies. It is not a complete solution by itself, but it can expose energy that an A-weighted reading minimizes. A technically serious rule also identifies where sound is measured, the averaging period, day and night limits, how tonal components are penalized, how baseline ambient sound is established and when post-construction verification occurs. It tells an operator what to do if the completed facility fails.

Most ordinances in the review do not reach that level. They name noise, add one number and rely on distance.

Large setbacks can reduce exposure, and for small governments they are administratively attractive. But distance is not an acoustic outcome. Topography, equipment orientation, weather and multiple sound sources can make a modeled buffer perform differently in practice. Bradley County’s 3,000-foot rule may prevent most conflicts by preventing most projects. It does not prove that a compliant facility will meet a measured low-frequency standard at a home.

Little Rock’s combination of preconstruction analysis, post-construction testing, permanent monitors and recurring reports is more demanding because it checks the result. That is precisely why such provisions remain rare. Performance standards are expensive, technically adversarial and capable of producing an uncomfortable answer after a facility is already built.

They are also where regulation becomes real.

The next ordinance

The first generation of local data center rules was written to answer a zoning question: Where can this use go?

The next generation has to answer an operating question: What must this facility continue to prove?

That does not mean abandoning setbacks, screening, acreage minimums or utility letters. Bright lines are useful, especially in places with small staffs. It means pairing those tools with performance standards and an auditable life cycle: load and water thresholds that capture the actual operation; aggregation rules that prevent a campus from being split below the limit; baseline and post-construction acoustic testing; annual energy, water, generator and noise reporting; correction plans when performance drifts; and financial assurance that follows the project through closure.

The most sophisticated ordinances in the collection are not simply the longest or the harshest. They are the ones that preserve leverage after opening day.

For now, America’s local data center code is mostly still built around the permit. The facilities it governs will run for decades. The law is beginning to catch up—but only at the edges, and one annual report at a time, if that.