Data center site selection is often described as a hunt for cheap power. Power matters enormously, but it is only one variable in a complex optimization. The sites that win combine affordable electricity with available land, water for cooling, fiber connectivity, tax treatment, workforce, and — increasingly decisive — community acceptance. Fail on any one dimension and the project stalls, no matter how attractive the power price.
The site-selection checklist
Start with the physical requirements. A hyperscale data center campus needs a large, flat parcel — 50 to 200 acres for a major campus — with suitable geology and minimal flood, seismic, or wildfire risk. The site needs proximity to high-voltage transmission infrastructure, or at least a feasible path to build it. It needs fiber-optic connectivity with diverse routing — data centers live and die by network latency and redundancy, and fiber follows population and existing network hubs more than it follows cheap power.
Water is the requirement most often underestimated. Data centers use water primarily for cooling — evaporative cooling systems, the most common type, consume water as they reject heat. A large data center can consume millions of gallons per day, comparable to a small town’s water supply. In water-rich regions this is a non-issue; in arid regions it is a binding constraint and a political lightning rod. Alternative cooling technologies — air-cooled systems, closed-loop liquid cooling — reduce or eliminate water consumption but typically cost more or use more electricity, trading one constraint for another.
Then come the policy variables. State and local tax incentives for data centers are widespread and often decisive: sales tax exemptions on equipment, property tax abatements, and similar measures can shift project economics by tens of millions of dollars. Electricity rate structures matter — industrial rates vary enormously across states. Permitting speed and predictability matter enormously to developers working against AI deployment timelines. And workforce availability — electricians, technicians, construction labor — constrains how fast a region can absorb data center construction.
Why development clusters
These variables explain why data center development concentrates in a handful of regions. Northern Virginia — the largest data center market in the world — won on fiber (proximity to internet backbone infrastructure), power (historically adequate transmission and generation), land (available suburban and exurban parcels), and policy (state tax incentives). Similar logic built major markets in Texas, the Pacific Northwest (cheap hydropower), the Southeast, and parts of the Midwest.
Clustering has self-reinforcing effects. Once a region has data centers, it has the fiber routes, the electrical contractors, the permitting precedents, and the utility experience to serve more of them. Developers prefer known quantities. Utilities in established markets have large-load processes and tariff structures already in place. The result is concentration: a small number of counties and utility territories absorbing a large share of national data center growth.
But concentration creates its own problems. The grid in a cluster region faces the lumped impact of many simultaneous large loads. Land prices rise. Water resources come under pressure. And communities that welcomed the first data centers may balk at the tenth — which brings us to the politics.
The community pushback
Data centers are, by design, unobtrusive neighbors: windowless buildings, minimal traffic, no smokestacks. But at hyperscale, their footprint becomes hard to ignore, and community opposition has grown from a rarity to a regular feature of data center development.
The objections cluster around a few themes. Electricity rates top the list: residents fear — often with reason — that the infrastructure costs of serving massive new loads will land on household bills. Water use is second, especially in drought-prone regions, where the image of millions of gallons evaporating daily for server cooling provokes strong reactions. Noise is third: the constant hum of cooling equipment and backup generators can carry to nearby homes, and some facilities have faced sustained complaints. Land use and aesthetics follow — large industrial buildings and their associated transmission lines change the character of suburban and rural areas. And in some communities, there is a simple fairness question: the facility employs relatively few people per megawatt, so the local economic benefit can feel thin relative to the resources consumed.
For illustration, consider the employment math. A 200-megawatt hyperscale facility might employ 100 to 300 permanent staff once operational — well under one job per megawatt. A manufacturing plant of comparable power demand might employ thousands. Communities that grant tax incentives expecting manufacturing-scale job creation can feel shortchanged, even though the construction phase does bring significant temporary employment and the facility contributes to the tax base.
Opposition takes concrete forms: zoning challenges, demands for greater setbacks and noise limits, water-use restrictions, requirements for community benefit agreements, and in some cases outright moratoriums on new data center development. Several localities have paused approvals to study the cumulative impacts of multiple facilities — a rational response to clustered growth, but a source of deep frustration for developers on tight timelines.
The tax incentive debate
Few aspects of data center development are as contested as the tax incentives used to attract it. The standard package — sales tax exemptions on servers and equipment, property tax abatements, sometimes income tax credits — can be worth tens of millions of dollars per project. Proponents argue the incentives are decisive in site selection and that the resulting investment, construction employment, and property tax base justify the cost. Critics argue that data centers, with their low permanent employment and massive resource demands, are poor candidates for public subsidy — and that the incentives primarily shift projects between competing jurisdictions rather than creating new investment.
For illustration, consider a hypothetical $1 billion data center campus receiving a 50 percent property tax abatement over 15 years. The foregone revenue can exceed $50 million — real money for a county budget — in exchange for a facility employing a few hundred people permanently. Supporters point to the construction phase, which employs thousands over two to three years, and to the signal the investment sends to other employers. Skeptics note that the developer would likely have built somewhere, and the abatement merely determined where.
The debate is sharpening as the scale grows. Some jurisdictions are attaching conditions to incentives: local hiring requirements, water-use limits, community benefit contributions, or clawback provisions if the facility doesn’t meet its commitments. Others are letting incentive programs expire, betting that power availability and interconnection speed matter more than tax treatment at this point in the cycle. A few are going the other direction, expanding incentives to compete for AI infrastructure specifically. There is no consensus — but the era of unquestioned, unconditional data center tax breaks is ending.
Water: the emerging flashpoint
Water deserves special attention because it is becoming the binding constraint in more regions. The core issue is evaporative cooling: the cheapest, most established way to reject heat from a data center consumes water by design. As facilities scale up and cluster in arid and semi-arid regions — drawn by solar resources, land availability, and policy — water consumption becomes both a physical limit and a political one.
The industry’s responses are real but partial. Air-cooled chillers eliminate water use at the cost of higher electricity consumption and larger equipment footprints. Closed-loop liquid cooling — circulating coolant through the servers and rejecting heat without evaporation — dramatically cuts water use but requires redesigned facilities and higher capital expenditure. Some operators commit to “water positive” goals, funding watershed restoration exceeding their consumption. And siting decisions increasingly treat water availability as a first-order screen, not an afterthought.
Transparency is improving but incomplete. Water Usage Effectiveness (WUE) — liters of water per kilowatt-hour of IT energy — gives the industry a common metric, but disclosure remains voluntary and inconsistent. Communities negotiating with developers increasingly demand water-use commitments as a condition of approval, and some jurisdictions are writing them into permits.
What good siting looks like
The developers navigating this landscape successfully share a playbook. Engage the community early — before permits are filed, not after opposition organizes. Commit to ratepayer protections in writing, so residents don’t fear subsidizing the facility. Address water explicitly, with measured commitments and efficient cooling technology. Design for the neighborhood: setbacks, noise mitigation, architectural treatment that acknowledges the building’s presence. And negotiate community benefit agreements — infrastructure improvements, workforce programs, tax revenue sharing — that make the facility’s value tangible to the people living next to it.
Utilities and regulators have a role too. Hosting-capacity maps that steer developers toward grid-ready sites reduce the land-use conflicts that come from speculative site selection. Coordinated planning across jurisdictions can distribute growth rather than concentrating it in the path of least resistance. And rate structures that guarantee large loads pay their full costs remove the most potent source of community resentment.
Transparency is the thread connecting these practices. Developers that publish water usage, noise studies, traffic plans, and decommissioning commitments before being asked tend to face less opposition than those that disclose the minimum required. Some operators now release community impact assessments modeled on environmental impact statements — voluntary, detailed, and publicly available. In a political environment where trust is scarce, proactive disclosure is becoming a competitive advantage in siting, not just a compliance exercise.
Workforce development is an underused tool. Data center operations jobs — technicians, electricians, facilities engineers — pay well and don’t require four-year degrees, which makes them attractive to communities weighing the employment question. Operators that partner with local community colleges on training pipelines convert an abstract job count into a tangible local benefit: a path from the neighborhood into the building. A few regions have made such partnerships a formal part of the incentive or permitting package, aligning the facility’s hiring with the community’s workforce needs.
The siting outlook
The fundamental tension isn’t going away: the industry needs to build at unprecedented speed, and communities have legitimate interests in how their land, water, and grid are used. The regions that thread this needle — offering developers speed and certainty while protecting residents on rates, water, and quality of life — will capture the investment. Those that can’t will watch it go elsewhere.
Siting is where the data center boom meets democracy, one county zoning hearing at a time. Power may determine which regions are candidates, but land, water, and local consent determine which projects actually get built.
That consent is not a fixed resource — it responds to how the industry behaves. Each facility that protects ratepayers, conserves water, mitigates noise, and invests in its community makes the next project easier to site. Each one that extracts tax breaks while straining local resources makes the next project harder. The industry’s siting challenge is therefore partly of its own making, and its resolution lies partly in its own hands: build facilities that earn their place, and communities will keep making room for them.