The Data Center Next Door
- 2 days ago
- 8 min read
How the AI Boom is Reshaping Local Communities
On a massive, windswept swath of private farmland in Box Elder County in northern Utah, local residents are worried about what “the cloud” will feel like when it finally lands.
The Stratos Project Area artificial intelligence (AI) and defense mega data center that they are debating—and which the Box Elder County Commission approved in May—would not be a vague digital abstraction. It would be a concrete complex of server halls, substations, cooling towers, and backup generators on three sites over 40,000 acres—large enough to draw power as if it were a small city.
What truly grabbed headlines about the new data center was its projected heat generation. A Utah State University physics professor estimated that, under certain assumptions, the facility could release waste heat into the air and water equivalent to the energy of 23 atomic bombs per day.
The analogy was intentionally dramatic, but it captured a real concern: In a warming West already facing shrinking snowpack and a stressed Great Salt Lake, many fear that a vast new heat source and water user could exacerbate local conditions in the wrong direction.
Supporters of the new “hyperscale” data center, which is backed by Utah’s Military Installation Development Authority, counter that the campus “will fund modern buildings at Hill Air Force Base,” the Salt Lake Tribune reported. It will also “leave ratepayers unaffected by generating its own power with natural gas, reuse its water, bring in new jobs and keep the United States competitive against China in the artificial intelligence race,” the newspaper said.
Box Elder County’s project—and its public outcry—are not an isolated case. Across the United States and beyond, communities from Arizona to Virginia and even faraway India are wrestling with the local consequences of an AI‑driven data center boom.
Once tucked quietly into industrial parks, data centers are now among the most visible—and contested—pieces of infrastructure being built.
Once tucked quietly into industrial parks, data centers are now among the most visible—and contested—pieces of infrastructure being built.
Behind our phones’ gentle notifications lies a hard question: Can we design the infrastructure powering AI so that it supports, rather than strains, the places where it is built?
From Invisible Servers to Electric Cities
The first shock for many communities is the sheer scale of electricity that modern data centers demand.
Hyperscale facilities serving AI workloads often require 100 megawatts of power or more—roughly equivalent to the load of a small city—concentrated on a single campus or cluster.
For local utilities, that can mean massive new substations, high‑voltage transmission lines, and grid upgrades compressed into a few years rather than decades. When several projects arrive at once, planners worry that they will have to delay connecting other industrial customers, lean more heavily on gas‑fired peaker plants, or even consider new fossil‑fuel capacity to keep up.
Communities understandably ask whether they will end up paying higher rates so that faraway users can chat with AI chatbots or train the next generation of models.
Communities understandably ask whether they will end up paying higher rates so that faraway users can chat with AI chatbots or train the next generation of models.
Yet energy experts also see an opportunity. Some types of computing, especially AI model training and batch analytics, do not have to run at a precise minute or in a specific place. In principle, these “flexible computing” workloads could be shifted to times of day when solar and wind output are strongest, or to regions with abundant clean power, turning data centers into adjustable loads that smooth rather than stress the grid.
Water: The Peak-Day Problem
If electricity tends to dominate public debate, water may be the subtler constraint.
Prof. Shaolei Ren, associate professor at the University of California, Riverside campus, who studies the environmental footprint of digital infrastructure, suggests that communities must look beyond simple annual totals.
Annual water use is useful, he tells The Earth & I, but it can hide the real infrastructure issue: Water systems are stressed by peak daily demand, not just the sum over a year. A project with modest annual use can still cause large spikes on hot, dry days—exactly when local systems are already under pressure. Communities should therefore insist on both annual estimates of water usage and peak daily demand, including under drought conditions, before they sign off on a project.
Ren also emphasizes that water and energy need to be “planned together,” not treated as separate checkboxes. The best strategies are not simply “use less water at any cost,” because some alternatives—such as abrupt shifts to dry cooling—can drive up electricity consumption and emissions.
More promising approaches include using nonpotable or reclaimed water where feasible: With on‑site storage, the data center can draw less from the system during peak hours, improve cooling efficiency, and transparently evaluate the trade‑off between water use and electricity demand.
Diesel Backups and Air Quality
Most people never see the diesel generators that sit behind modern data centers, but local residents may certainly hear and smell them. These rows of engines are there to satisfy strict reliability expectations: If the grid fails, the servers must stay on.
Ren points out that even though backup generators are intended for rare emergency use, they can still matter locally. Diesel units emit high levels of nitrogen oxides and fine particulate pollution when they run; the risks of air pollution are especially acute in communities already overburdened by highways, ports, warehouses, or power plants.
Diesel units emit high levels of nitrogen oxides and fine particulate pollution when they run; the risks of air pollution are especially acute in communities already overburdened by highways, ports, warehouses, or power plants.
Regulators therefore need to ask where the generators are located, how often they may run (including tests), and how their emissions combine with existing local burdens. That cumulative perspective is particularly important in places like Northern Virginia’s “data‑center alley” and emerging clusters such as Box Elder County, where concerns about air quality and heat are now intertwined.
Cleaner options exist. Large‑scale battery storage paired with renewables can provide fast, clean backup for at least short outages, while cleaner fuels or stringent emissions controls can reduce pollution from any remaining generators. But, as with water and energy, these alternatives must be designed into the project from the start; communities rarely get a chance to renegotiate once the concrete is poured.
Land, Jobs, and the Local Bargain
A large AI campus does not just draw electricity and water; it also remakes land and local economies.
A large AI campus does not just draw electricity and water; it also remakes land and local economies.
A single facility can convert hundreds of acres of farmland or open space into fenced compounds, access roads, and electrical yards. For some towns, this is a welcome sign of investment; for others, it feels like an industrial takeover.
Construction brings a wave of temporary jobs, but long‑term employment at a highly automated facility may be modest relative to its footprint, especially once maintenance and security personnel are established.
Generous tax incentives can further dilute the fiscal benefits that were used to justify the project in the first place.
Urban‑planning scholar and University of Michigan Assistant Professor Xiaofan Liang tells The Earth & I that some of the most important local impacts—shifts in housing markets, transportation patterns, land‑use change, or social cohesion—may only emerge over longer time horizons and are still poorly documented. Because these changes are hard to quantify and fall outside the immediate permitting checklist, they are often overlooked during approvals.
Many jurisdictions currently permit data centers under broad industrial or commercial labels. Liang argues that the underlying concerns—energy demand, water consumption, noise, setbacks, environmental impacts—are not entirely new to zoning, but the scale and intensity of those impacts can be far greater than for traditional industrial uses.
Cities may be able to gain finer-grained control over siting and design by asking for explicit zoning language for AI data centers while following new practice reports from planning associations. For communities trying to negotiate better outcomes, Liang suggests starting with a big-picture understanding of how a data center actually works—asking which regional power lines and water systems it will depend on, what the development life cycle will look like, and which agencies or companies will control each decision point.
Practical tools such as public hearings, community benefit agreements, and clear permit conditions can then turn that knowledge into leverage.
What a Community-Integrated Data Center Looks Like
Ren argues that a truly community-integrated data center would be designed around several elements:
Local infrastructure limits, not just the company’s computing roadmap
Use of water sources that do not diminish community water security
Reporting both peak and annual water and energy demand
Choosing cooling technologies based on local grid and climate conditions
Investing in shared infrastructure whenever the facility creates new burdens.
Moreover, for backup power, a community-integrated data center would minimize diesel reliance where feasible and carefully evaluate local air quality impacts.
A truly community-integrated data center would be designed around local infrastructure limits, not just the company’s computing roadmap.
Liang also sees two complementary ideas emerging under the same banner.
The first is the concept of community-integrated data centers as large commercial facilities that deliberately build community goals into their projects. That can mean outreach or sustainability teams, formal community benefit agreements, funding for nonprofit initiatives such as digital access programs, and architecture and streetscapes that genuinely fit the surrounding neighborhood.
The second idea is more experimental but goes further and imagines community‑run data centers. These remain mostly conceptual, but researchers are exploring the technical feasibility of smaller‑scale, locally controlled facilities that process local data under local governance. In these visions, a data center might reuse an abandoned factory, share a building with a workforce‑training center, bundle rooftop solar and broadband upgrades, and be governed through community institutions rather than only corporate boards.
Around the world, existing projects hint at what this future could look like in practice.
In Odense, Denmark, for instance, Meta’s data center was designed with heat recovery in mind. Instead of simply releasing warm exhaust air, the facility captures low‑temperature waste heat from its servers and feeds it into the city’s district heating network via large heat pumps, donating up to 100,000 megawatt‑hours of heat each year—enough to warm thousands of local homes.
In Hong Kong and several US metros, cloud providers now cool data centers with recycled or nonpotable water, developed in partnership with local utilities, to reduce pressure on drinking water supplies. These projects are far from perfect, but they show that with planning and partnership, resource footprints can be dramatically reduced.
Choosing the Future of the Cloud
In the debate over the unknowns of AI data centers, certain questions arise: What if projects started with a mindset to size the facility to fit within local grid and water constraints? What if the center was designed to use reclaimed water and on-site storage to shave peak demand? Other concerns are to equip them with batteries and minimal diesel, and connect to nearby homes, farms, or greenhouses via a system that turns waste heat into a community asset. Finally, can the centers be paired with investments in broadband, workforce training, and local public spaces?
In these scenarios, a data center next door might still be controversial—but it would look less like a looming industrial threat and more like an unusual but honest neighbor, one that does its share to keep the community livable in a warming world.
The AI boom guarantees that more data centers are coming. The question is whether they arrive as opaque, resource‑hungry bunkers, or as carefully planned pieces of shared infrastructure.
Treating data centers as grid citizens and potential community partners—rather than invisible clouds floating above the physical world—opens up room for negotiation and creativity. The examples already emerging in places like Odense, Quincy, and Hong Kong show that when communities, regulators, and companies make different choices early enough, the cloud can support the ground beneath it instead of overwhelming it.
*Dhanada K. Mishra is a PhD in Civil Engineering from the University of Michigan and is currently working as the Managing Director of a Hong Kong-based AI startup for building technology for the sustainability of built infrastructure (www.raspect.ai). He writes on environmental issues, sustainability, the climate crisis, and built infrastructure.