(SBA) - Across the United States, communities are being asked to make decisions about developments unlike almost anything local zoning ordinances were written to address. Companies are proposing data centers covering hundreds of acres, requiring hundreds of megawatts of electricity and carrying investment figures measured in billions of dollars. Artificial intelligence has accelerated the demand, adding another layer to an industry already expanding because of cloud computing, streaming, online commerce and the enormous amount of digital information Americans generate every day.
The possibility does not necessarily mean there is an imminent “data-center bubble” that is about to burst. Demand for computing is real, and some of the world's largest and financially strongest companies — including Amazon, Microsoft, Google and Meta — are spending enormous sums expanding computing infrastructure. But history demonstrates that rapidly expanding industries can overbuild, technology can change faster than expected, projected demand can fall short, financing conditions can tighten and individual developers can fail even when the larger industry remains healthy.
That creates an important distinction for communities considering data-center projects. The question is no longer simply whether a company can afford to begin construction. Local officials also need to understand whether the community is financially and environmentally protected if the project is delayed, reduced in size, sold to another company or abandoned after construction has begun.
The phrase “data center” encompasses facilities that differ tremendously in size, purpose and effect on surrounding property. An enterprise data center might operate the computer systems of a hospital, bank, university or government agency. Colocation facilities lease server space to multiple customers. Cloud facilities operate the infrastructure behind services such as Amazon Web Services, Microsoft Azure and Google Cloud. Hyperscale campuses contain enormous concentrations of computing equipment and may consist of numerous buildings constructed over many years.
Cryptocurrency mining is another category sometimes loosely described as data-center development but capable of having a very different effect on its neighbors. Bitcoin miners use thousands of specialized computers to perform calculations continuously in pursuit of cryptocurrency rewards. Many mining operations have traditionally relied heavily on high-speed fans drawing enormous quantities of outside air through rows of machines.
That difference becomes especially important when discussing noise.
A common misconception is that a building containing computers must be relatively quiet. Inside a home or office, computers usually are. Multiply the equipment by tens of thousands, however, and keeping those computers cool becomes an industrial operation.
Many Bitcoin mines have historically operated differently. Thousands of mining computers equipped with high-speed fans may be installed in long buildings or shipping-container-like structures with enormous quantities of air moving continuously through them. Instead of intermittent industrial noise, neighboring residents may encounter a constant mechanical roar, drone or hum that continues throughout the night.
Residents around Granbury, Texas, became one of the nation's best-known examples of this conflict. People living near a large Bitcoin-mining operation complained of constant mechanical noise associated with cooling equipment and said it interfered with sleep and the peaceful use of their property. The dispute produced noise citations, extensive media coverage and litigation alleging the operation constituted a nuisance. Operators pursued mitigation measures, while some health claims made by residents remained disputed and should not be treated as medically established simply because residents associated symptoms with the facility.
Cryptocurrency-mining disputes have also occurred in New York and North Carolina, where residents and local officials confronted around-the-clock fan noise from mining operations. Several communities responded with stricter noise requirements or permitting rules after discovering conventional zoning ordinances had never contemplated thousands of high-speed cooling fans operating continuously.
None of this means an AI or cloud data center will necessarily be quiet. It means communities should know exactly what type of facility is being proposed and what cooling technology it will use. A responsible noise review should include nighttime conditions, low-frequency sound, tonal components, backup-generator testing and measurements at the nearest existing residence — not merely at the developer's property line.
The scale of electricity consumption is one of the biggest differences between today's largest data-center proposals and previous generations of commercial development.
The range is wide because no one knows precisely how quickly AI demand will grow, how efficient future processors will become or how computing workloads will change. But even the uncertainty illustrates the scale of what electric utilities are being asked to plan for.
A facility requiring 10 megawatts is a significant customer. A campus requiring 500 megawatts is something fundamentally different. A project approaching one gigawatt — 1,000 megawatts — can require electrical infrastructure comparable in scale to that serving a substantial population.
Providing that power may involve new substations, transformers, transmission lines and eventually additional generating capacity. Those investments can extend far beyond the borders of the data-center property.
Northern Virginia offers the clearest example of how this can change a region. Loudoun County and surrounding areas became the largest concentration of data centers in the country. The development generated substantial local tax revenue, but the enormous concentration of computing also contributed to the need for additional transmission and utility infrastructure. Proposed transmission corridors have in turn generated opposition from property owners who may receive none of the direct benefits of the data-center development but could live near new high-voltage lines serving it.
For residents, one of the most important questions is therefore not simply, “How much electricity will this facility use?”
It is, “Who will pay for the infrastructure necessary to provide it?”
Utilities can create specialized tariffs and contracts requiring large customers to pay the costs they create. They can require deposits, minimum-demand commitments or long-term agreements intended to protect ordinary customers if a major industrial load does not materialize as expected.
That risk becomes particularly important when projects are developed in phases.
Data-center water consumption varies dramatically depending upon cooling design. Some facilities rely significantly on evaporative cooling. Others use closed-loop systems, dry cooling, air cooling, reclaimed wastewater or combinations of those technologies.
One of the most widely documented disputes occurred in The Dalles, Oregon, where Google operates data centers. Public-record litigation eventually resulted in disclosure of water-consumption information showing that Google's facilities accounted for a substantial portion of the city's water use. Publicly reported records showed approximately 274 million gallons in 2021, roughly 29 percent of the city's total water consumption that year.
The controversy was about more than the quantity of water. It also became a transparency issue because residents and journalists had sought information about how much of the community's water was being consumed by the data centers.
Google has emphasized water-efficiency efforts and water-restoration programs. Nevertheless, the Oregon case illustrates why communities should establish reporting requirements before a project is approved rather than attempting to determine consumption years afterward.
For agricultural communities, groundwater deserves particular scrutiny. A community depending on an aquifer shared by residences, agriculture and industry faces a fundamentally different calculation than a facility with access to an abundant surface-water source or reclaimed municipal wastewater.
Companies should therefore be asked to disclose average daily water consumption, maximum summertime consumption, the source of that water and how much water is actually consumed rather than merely circulated and returned.
One of the most impressive numbers in almost every data-center announcement is the total capital investment.
A company may announce that it plans to spend $3 billion, $5 billion or even considerably more. Those figures are meaningful because the buildings, electrical equipment, cooling systems and computer hardware truly are expensive.
But investment and taxable investment are not necessarily the same thing.
Servers themselves may account for billions of dollars of an investment and are replaced periodically as technology advances. If that equipment receives favorable tax treatment, the headline investment figure can greatly exceed the value on which normal local property taxes are ultimately collected.
For taxpayers, the more useful question is therefore:
After every exemption, rebate, abatement and development agreement is applied, how much will the company actually pay annually to the city, county, school district and other local taxing entities?
That number should then be placed alongside the cost of any public infrastructure being provided for the project.
A $5 billion investment that provides substantial net tax revenue is one economic proposition. A $5 billion investment receiving extensive exemptions while requiring publicly financed roads, utilities and other infrastructure may be another.
Data centers can generate significant employment during construction. Large campuses can employ hundreds or thousands of electricians, plumbers, concrete workers, equipment operators, engineers and other trades over several years.
That construction activity is genuine economic activity.
Once the buildings are complete, however, employment drops considerably. Data centers are extremely capital intensive rather than labor intensive. A facility containing billions of dollars of servers may employ only a few hundred people — and some facilities considerably fewer.
Permanent employees can include electricians, network technicians, security personnel, HVAC specialists, engineers, facilities technicians and managers. Many of those are skilled, well-paying positions.
But communities need to distinguish between construction jobs, projected permanent jobs and guaranteed permanent jobs.
History shows why.
In 2018, cryptocurrency company Bitmain announced ambitious plans for a massive mining operation at a former industrial property in Rockdale, Texas. Public announcements discussed hundreds of millions of dollars in investment and the potential for approximately 400 jobs.
Then cryptocurrency markets deteriorated.
Bitmain rapidly scaled back the project and reduced its workforce. The original employment expectations did not materialize on the timetable initially envisioned. The property later remained important to the cryptocurrency industry under evolving ownership and operations, but the experience demonstrated how quickly projected employment can change when market conditions change.
It also demonstrates why the wording of economic-development announcements matters.
“There could eventually be 400 jobs” is not the same thing as a binding agreement requiring a company to employ 400 people.
This is increasingly one of the most important questions communities can ask.
That does not mean those forecasts are wrong.
But forecasts remain forecasts.
AI processors may become more efficient. New models may require less computing power. Companies could discover that customers are unwilling to pay enough for certain AI services to justify every proposed facility. Interest rates or construction costs could change. Electricity constraints could make some projects uneconomic. A recession could reduce demand. Companies could consolidate computing operations or shift investment into different regions.
And some developers are far more financially secure than others.
A campus financed and developed directly by Microsoft, Amazon, Google or Meta is financially different from a speculative developer acquiring farmland and electrical capacity with the expectation that a major technology company will eventually lease the buildings.
If demand weakens, several things could happen.
A partially completed project might simply be delayed. A developer might finish one building rather than six. An unfinished site could be sold to another operator. A completed facility could be converted to another computing use. A financially distressed company could enter bankruptcy and its lenders could take control of the property.
Or a site could sit unfinished.
That last possibility is where local governments need to pay particularly close attention.
If hundreds of acres have already been graded, roads constructed, drainage patterns changed and utility infrastructure installed, walking away is no longer equivalent to abandoning an empty farm field.
There may be unfinished concrete foundations, industrial buildings, electrical equipment, fuel storage, construction debris, stormwater systems and disturbed land requiring stabilization or removal.
And the company that originally made the promises may no longer exist in its original form.
Communities do not have to imagine what a sudden reversal in a computing boom looks like.
The cryptocurrency industry experienced one.
During periods of rapidly rising Bitcoin prices, mining companies rushed to construct facilities and obtain inexpensive electricity. Some companies borrowed heavily on the assumption that high cryptocurrency prices would continue.
When cryptocurrency prices fell and electricity costs rose, the economics changed rapidly.
Major industry participants experienced serious financial distress. Compute North filed for Chapter 11 bankruptcy protection in 2022, while Core Scientific, then one of the largest publicly traded Bitcoin-mining companies, also entered Chapter 11 later that year before eventually restructuring.
Those bankruptcies did not mean every facility was abandoned. Valuable assets were sold, transferred or continued operating under reorganized ownership.
That is actually an important lesson.
The physical assets of a data center — particularly a site with large quantities of available electricity and fiber-optic connectivity — may remain valuable even if the original operator fails.
But communities cannot assume another buyer will always appear.
A development agreement should be designed for the possibility that one does not.
Another question emerging as local governments debate moratoriums and tighter regulation is whether banks and other lenders are becoming hesitant to finance data centers.
Many of the largest data centers are backed by some of the world's strongest corporations, which can finance construction through existing cash, corporate borrowing and long-term capital markets. Other developers rely on conventional construction loans, private-credit firms, institutional investors, real-estate financing, securitizations and joint ventures. Some projects obtain financing only after securing a long-term lease from a large technology company with a strong credit rating.
For lenders and investors, the biggest risks can include the availability of electricity, construction costs, interest rates, whether a major tenant has actually committed to the facility, whether the developer can obtain permits and whether the equipment and buildings will remain economically useful over the life of the loan.
Regulatory uncertainty also matters.
If a local government adopts a temporary moratorium, considers substantially larger setbacks, requires new environmental studies or changes the permitting process, a lender may have to reassess when — or whether — a project can proceed. An uncertain permit can affect construction schedules, interest expenses and ultimately the project's value.
That would require evidence from lending institutions, credit markets or financing data demonstrating such a connection.
Political developments can certainly become part of a lender's risk analysis. A project proposed in a jurisdiction where officials are considering a moratorium, changing zoning standards or imposing new infrastructure requirements may represent greater regulatory risk than an otherwise identical project with all permits in hand.
But the larger financing questions are more fundamental: Who is borrowing the money? Who has committed to use the data center? Is sufficient electricity available? What happens if demand falls? And who is responsible for infrastructure costs if the project never reaches its promised size?
Those issues would matter regardless of which candidates won an election.
There is another distinction communities should understand.
A data center already leased for 15 or 20 years to one of the world's largest technology companies presents lenders with relatively predictable revenue.
A developer constructing millions of square feet of data-center space because it expects someone will eventually lease it presents a different risk.
The same issue applies to electricity.
Available power has become one of the scarcest resources in many major data-center markets. Developers may therefore have an incentive to reserve very large quantities of electrical capacity before every customer or building has been secured.
But utilities have to distinguish genuine future demand from speculative requests.
If every project requesting 500 megawatts never actually needs 500 megawatts, utilities could overestimate future demand and construct infrastructure for loads that never appear.
Utilities and regulators can reduce that risk through deposits, milestone payments, minimum-demand charges, long-term service agreements and other financial commitments. Those protections make it harder for a developer to reserve enormous quantities of scarce electrical capacity, trigger costly infrastructure investment and then walk away without financial consequences.
Local governments can apply the same principle to development agreements.
If a community wants protection against an unfinished or abandoned development, relying exclusively upon a company's promise to clean it up presents an obvious weakness.
The name appearing on the local deed may therefore not be the name of the multibillion-dollar parent company residents recognize.
A contractual cleanup obligation belonging only to a thinly capitalized subsidiary may have limited value if that entity becomes insolvent.
Financial assurance can take several forms.
A local government could require a performance bond, irrevocable letter of credit, escrowed funds, parent-company guarantee or another form of dedicated financial security sufficient to complete required public improvements or address an abandoned project.
Those mechanisms are not interchangeable.
A parent-company guarantee is only as strong as the corporation actually guaranteeing the obligation. A performance bond covers only the work identified in the bond. A normal construction bond covering a public street may not provide one dollar toward demolishing an abandoned data-center structure unless the agreement explicitly includes that obligation.
The details of the agreement matter considerably more than a general promise that the company “will take care of it.”
Local governments have increasingly learned this lesson from other forms of large-scale development.
Wind and solar developments in many jurisdictions are now subject to decommissioning requirements intended to establish who removes equipment and restores land when a project reaches the end of its useful life.
The same principle can be applied to large data centers.
A data-center decommissioning plan could define when a project is considered abandoned, how long construction may remain inactive, what equipment and structures must be removed, how disturbed land must be stabilized and who is responsible for the expense.
The estimated cost should also be periodically updated.
A cleanup estimate calculated before construction may bear little resemblance to demolition and restoration costs 10, 20 or 30 years later.
Communities should also be cautious about relying heavily on predicted salvage value. Electrical equipment, copper and structural material may have significant value someday, but future markets cannot be guaranteed.
Perhaps most importantly, decommissioning requirements should follow the property.
If the project is sold three times, the third owner should remain bound by the same environmental, financial and cleanup obligations imposed upon the first.
Communities also have options for reducing the financial risk associated with economic-development incentives.
If a project is expected to involve billions of dollars of investment over several construction phases, incentives can likewise be structured around investments that actually occur rather than every phase envisioned in an initial master plan.
Development agreements can contain clawback provisions allowing governments to recover some benefits when legally specified investment, employment or construction commitments are not met.
The wording remains critically important.
A clawback cannot recover an economic-development “promise” that was never made legally binding.
Whenever an announcement says a project could, may, is expected to, has the potential to, or could employ up to a certain number of workers, residents should understand those phrases generally describe projections rather than guarantees.
Local governments should also consider what happens to public infrastructure if development stops before completion.
Suppose a city extends a large water main, reconstructs a road or expands wastewater capacity based primarily on anticipated revenue from a proposed data center.
If construction stops halfway through, the bonds used to finance that infrastructure do not disappear.
Someone still has to repay them.
Similarly, an electric utility may construct a new substation or transmission infrastructure anticipating hundreds of megawatts of demand. If the customer never reaches that level of consumption, regulators need a mechanism preventing unrecovered costs from simply becoming somebody else's burden.
Development agreements can therefore address not only construction but also failure to construct.
Projects can be required to meet development milestones by specified dates. Zoning or permits can expire if construction does not proceed. Developers can be made responsible for defined infrastructure costs even if future phases are canceled.
Those protections shift some of the risk away from existing taxpayers and utility customers.
As proposals have grown larger, some communities have considered temporary moratoriums on data-center development while elected officials review existing ordinances.
They are not necessarily the same thing.
A temporary moratorium generally pauses certain applications, permits or approvals for a defined period while a government studies an issue and considers new regulations. Depending upon state law and the language of the ordinance, a moratorium might last several months or a year and could allow projects already possessing certain approvals to continue.
An outright ban, by contrast, prohibits a particular land use rather than temporarily delaying new approvals while rules are developed.
The distinction matters for both residents and developers.
A community confronted with a new type of 500-megawatt industrial use may discover that its zoning code was written when nobody contemplated such a facility. Existing ordinances may contain no meaningful requirements governing low-frequency noise, backup generators, groundwater consumption, decommissioning, transmission infrastructure or financial guarantees.
A temporary moratorium can give local officials time to answer those questions before approving projects under outdated rules.
There are also costs associated with delay.
Landowners may have pending transactions. Developers may face financing deadlines. Construction costs can change. A company may decide to take its investment elsewhere. Governments therefore have reason to define what they intend to study, establish a clear timetable and avoid leaving temporary restrictions in place indefinitely without action.
The relevant question is not simply whether a community is “for” or “against” data centers.
It is whether the community has adequate rules for a development of the size and type being proposed.
Debate over data centers is often framed around whether cities and counties should enact stricter ordinances.
A more precise question is whether their existing ordinances are adequate and specific enough.
Likewise, a setback measured from the wall of a data-center building may provide little protection if the loudest cooling equipment or substation is considerably closer to a neighboring home.
Communities can evaluate standards addressing residential setbacks, nighttime noise, low-frequency and tonal noise, generator testing, exterior lighting, landscaping, stormwater management, groundwater use, water reporting, transmission infrastructure and emergency response.
Those are land-use questions.
Separate development agreements can deal with the financial questions: performance bonds, letters of credit, infrastructure reimbursement, tax-incentive clawbacks, construction milestones, parent-company guarantees and decommissioning security.
Neither mechanism requires officials to predict whether AI will continue growing at its current pace.
They simply establish responsibilities before something goes wrong.
There remains considerably less rigorous research on residential property values near data centers than many people assume.
It is not currently defensible to state that a home located a particular distance from a data center will automatically lose a specific percentage of its value.
The variables are too great.
A landscaped facility inside an established industrial area a mile from residential property is fundamentally different from a massive campus constructed several hundred feet from rural homes that previously overlooked farmland.
At the same time, the value of farmland or industrial property suitable for data-center development can increase dramatically. Sites possessing access to large quantities of electricity, fiber infrastructure and adequate water can become exceptionally valuable.
It is therefore entirely possible for the same development to dramatically increase the value of land purchased for a data center while creating uncertainty for nearby residential property owners.
An independent baseline property-value study conducted before construction can provide useful information if disputes arise years later.
Before a major data-center project receives final approval, local officials and residents should know the maximum electrical demand of the entire proposed campus, not simply its first building. They should know what additional generation, transmission and substations will be necessary and precisely who will pay for those improvements if the project succeeds — or if it does not.
They should know predicted nighttime noise levels at the closest existing residence, including low-frequency and tonal noise. They should know how many permanent jobs are legally required and how many are merely projected.
They should know the value of every tax exemption and how much the company is actually expected to pay annually after those incentives are applied.
But communities should also ask the questions that seldom appear in a groundbreaking announcement.
What happens if only one building out of six is completed?
What happens if construction stops for two years?
How long before an unfinished project is legally considered abandoned?
Who secures the property?
Who maintains unfinished stormwater systems?
Who restores roads damaged by construction?
Who removes abandoned foundations, generators, fuel systems and electrical equipment?
Who pays if the limited-liability company that owns the property declares bankruptcy?
Is the cleanup obligation guaranteed by the parent corporation?
Is there a performance bond, letter of credit or other financial security available without relying on future tax revenue?
Are incentives reduced if later construction phases never occur?
Do all environmental, financial and decommissioning obligations automatically transfer to future owners?
And if public infrastructure is being constructed specifically for the project, who pays for that infrastructure if the expected customer never appears?
Those questions may sound pessimistic when billions of dollars of investment are being announced.
They are actually the same kinds of questions lenders, investors and insurance companies routinely ask before committing their own money.
The current data-center expansion may continue for many years. Artificial intelligence could ultimately require every bit of computing capacity now being contemplated and considerably more.
Neither local officials nor residents need to correctly predict which future will occur in order to protect their community.
They can evaluate a development according to what it consumes, what it produces, what it pays, what effect it may have on surrounding property and what legally enforceable responsibilities the developer accepts.
They can distinguish between a projected benefit and a guaranteed one.
They can distinguish between a temporary moratorium intended to develop appropriate regulations and a permanent prohibition on development.
And they can require agreements that remain workable if everything goes according to plan — and if it does not.
For residents, the important numbers remain megawatts, gallons, acres, decibels, setbacks, permanent jobs, infrastructure costs, tax exemptions and actual annual tax payments.
But several more should now be added:
The amount of financial security guaranteeing unfinished work.
The cost of decommissioning the site.
The amount of infrastructure investment for which the developer remains responsible if future phases are canceled.
Because a company's announcement describes what it intends to build.
A zoning ordinance determines what it is allowed to build.
And a carefully written development agreement determines who is responsible if what was announced is never completed.
Editor's Note: City records from The Dalles, Oregon, show that Google’s data centers used about 355 million gallons of municipal water in 2021, roughly 29 percent of the city’s total water use. Google separately reported 274.5 million gallons consumed that year, a figure it calculated by subtracting the water it discharged from the water it withdrew.
Lawrence Berkeley National Laboratory’s December 2024 report projected that data centers could account for 6.7 to 12 percent of U.S. electricity use in 2028. Its June 2026 update projected a share of 9.5 to 15.3 percent in 2030.
After the Nov. 3 midterms, Reuters reported that analysts saw a greater risk of project delays, and the Financial Times reported that some banks were using credit insurance to manage their exposure to data-center loans. As of Nov. 11, a nationwide lending pullback tied to the election results remained unverified.
Sources
Data-center technology, electricity and infrastructure
Water, noise and nearby property values
Jobs, tax incentives and projects that scaled back
Financing, the midterms and bankruptcy
Local ordinances and financial safeguards
