On a quiet, dusty stretch of rural land, the sale goes through with little fanfare. There’s no factory announcement. No promise of hundreds of jobs. No glossy renderings of a future downtown center. In many cases, local officials and residents don’t even know who the buyer is until the deed is recorded. What matters isn’t who will live there. What matters is how much power the land can deliver.

Across the U.S., a new land rush is underway, driven by the explosive growth of data centers that support artificial intelligence. Developers are scouring the map for places where electricity is available now — or can be made available faster than somewhere else — where water rights are secure, and where fiber can be brought online without years of delay. The result is a scramble for land that looks less like traditional economic development and more like a competition for raw inputs.

The shift is fundamentally about power, according to Jerry Poon, a principal engineer who works on large commercial and industrial power infrastructure. Sites are ruled out not because of zoning or cost, he said, but because the grid simply can’t support the load without years of upgrades. “The driver is electricity availability,” he explained, as well as water access and fiber reach. “Everything else is secondary.”

In other words, land that would once have been considered remote or undesirable is now suddenly valuable — and it’s not because people want to live there, but because machines do.

Megawatts > municipalities

For decades, states and cities competed to attract employers by offering tax breaks, infrastructure upgrades, and workforce pipelines. Data centers invert that logic. Forget labor availability or proximity to customers; today, the most important factor in site selection is whether or not the grid can deliver massive, continuous electricity demand sans interruption.

From an electrical standpoint, data centers behave very differently from traditional industrial users. Their loads are concentrated, steady, and intolerant of outages, and can require new substations, redundant feeds, transmission upgrades, and years of coordination with utilities before construction even begins. In rural or semi-rural areas, that infrastructure often doesn’t exist yet, which can create what engineers describe as hidden constraints on land that looks empty on a map.

One of the most underestimated factors is time. Power upgrades move on multi-year timelines, shaped by permitting, equipment availability, and rights-of-way, among other issues. In contrast, demand from hyperscale customers can materialize quickly once a site is selected. The mismatch sometimes produces a leapfrogging pattern — some projects wait, some downsize, and others hop to regions where capacity already exists or can simply be delivered faster.

As Poon pointed out, water access follows close behind, particularly in the western U.S., where cooling requirements increasingly shape site viability. In some locales, energy and water planning have necessarily become inseparable conversations. Jobs, however, are often an afterthought. A hyperscale data center may require a large construction crew and cost billions to build, but may employ only around 30 workers once it’s up and operating.

The result is a geography shaped by electrons rather than people, and land value pegged to megawatts rather than population growth.

Grids under pressure?

Many of these projects are landing in areas served by rural electric cooperatives, not large investor-owned utilities. But for those co-ops, data centers represent both opportunity and risk. A single facility can dramatically increase electricity demand, potentially justifying new transmission investment. Still, the upfront costs of modernizing grids—whether borne by developers themselves or co-op members—can be enormous, while the benefits may be unevenly distributed.

From a systems perspective, the strain goes beyond capacity. Woody Zhu, an assistant professor at Carnegie Mellon University who studies power systems and data centers, said large facilities introduce highly concentrated, continuous demand at single locations, changing how grids fail during extreme events.

“Rather than stress propagating gradually through diversified demand,” Zhu said, “stress can localize sharply around specific nodes.” In other words, problems pile up in one spot instead of spreading out. This increases the risk of “cascading outages or forced load shedding,” or deliberate power cuts to prevent grid collapse, in nearby communities.

In regions not originally designed for this level of sustained load, rapid data center growth can reduce operational flexibility and lengthen recovery times after disruptions. Without changes, Zhu warned, “an infrastructure-first approach” risks creating grids that operate in a “near-permanently stressed regime.” These systems may work, but offer significantly less margin for error.

“Long-term sustainability will require more proactive planning,” Zhu said—from coordinating siting decisions with transmission expansion to integrating uncertainty into load forecasts, and “pairing growth with strengthening measures like local generation, storage, or adaptive demand management.” Otherwise, the risk is more outages, and “slower recovery when failures do occur.”

Jim Matheson, the CEO of the NRECA, a national association representing roughly 900 not-for-profit electric co-ops and rural utilities, noted the chance and challenge for these rural providers. Co-ops remain “laser-focused on providing reliable and affordable power to their members,” Matheson said. “Data center development can bring tangible benefits to certain locations, including increased local tax revenues, both direct and indirect job creation, and broader infrastructure improvements. Many co-ops are still in the early stages of evaluating an influx of requests, including the potential system impact.”

“To date, we have not seen an increased strain on rural grids,” he said.

When physics turns political

The issue of who pays for what is no longer theoretical. As data-center growth begins to haunt household electricity bills—whether as reality, in some cases, or as more a potential specter amid a larger affordability crisis—the question of who pays for AI’s power appetite has become a live political issue.

On Tuesday, Microsoft $MSFT pledged to pay higher electricity rates for its data centers and to subsidize grid expansion, promising that its AI buildout would not raise costs for residential customers. The move follows growing opposition to new data-center projects in states including Virginia and Pennsylvania, where local governments have already moved to block or slow construction amid concerns about rising utility bills.

President Donald Trump seized on the issue this week, describing data centers as a cost-of-living threat and warning that voters would not tolerate subsidizing AI’s power binge. “I never want Americans to pay higher electricity bills because of data centers,” Trump said, claiming Microsoft was the first company to agree to “major changes.” Or at least to pledge them.

The timing is convenient. And the math is not. Electricity prices rose more than 6% nationally over the past year, according to federal data, and new estimates suggest surging demand from data centers is already pushing emissions higher by keeping fossil-fuel generation online longer than planned. It’s one part technical infrastructure challenge, one part rolling referendum on affordability.

Capital moves first, questions come later

From an investment standpoint, the data center boom looks anything but speculative. Occupancy rates appear high, much of the construction pipeline is pre-leased, and billions of dollars are flowing into land acquisition, construction, and specialized equipment.

At the local level, however, the process often feels opaque. Land purchases are frequently conducted through shell companies and non-disclosure agreements that function to obscure the ultimate buyer. In some towns, residents have only learned a massive facility is coming once zoning hearings are already underway, or even after construction has already begun.

The speed of these deals and their secretive nature can leave communities scrambling to understand just what they’ve agreed to. By the time public debate catches up, the fundamental decisions have already been made—about land use, about infrastructure, about long-term resource demand. That is, virtually the entire scope of projects.

This is not a break from American history

It’s tempting to frame the data-center land rush as something entirely new, and much media coverage presents the development in just that way, as a bleeding-edge, digital-age anomaly where infrastructure for machines takes precedence over human settlement.

In historical context, however, the pattern is much more familiar and of a piece with the past than you might guess.

In fact, American land booms have rarely been driven primarily by towns or jobs. From 19th-century railroads to gold rushes across California, Nevada, and the Black Hills of South Dakota and Wyoming, large-scale development prioritized control of land, energy, and capital-intensive infrastructure—and were frequently justified as inevitable progress.

Speaking to Quartz, the historian and author John Reeves pointed to the language of inevitability that often accompanied westward expansion. Even when settlers understood and acknowledged that land seizures violated treaties or laws, many believed economic development could not—and what’s more, should not—be stopped. Capital moved first; governance followed later, if at all.

That logic echoes today. Earlier booms wrapped infrastructure expansion in the language of settlement and opportunity. Data-center projects make similar claims about transformation, often touting community benefits and job creation even as their most significant requirements are resources and time.

“An often-overlooked principle of American capitalism is its tendency toward monopolistic forms of production,” Reeves, whose upcoming book concerns the theft of the Black Hills, said. That was “certainly the case” in the 19th century with fur-trading empires and mining companies dominating their respective industries. “And we see unregulated (more or less) monopolies like Amazon $AMZN, Meta $META, and Google $GOOGL dominating everywhere today. In both eras, there was a lot of rhetoric about free markets and the superiority of small-time producers, but it was the monopolies that reaped the lion’s share of the wealth creation.”

In this sense, then, the AI land boom is a distilled version of a much older American story—one in which decisions about land use and infrastructure are made before the public has a chance to weigh in, or even truly understands the issues at stake. The question now is not whether data-center expansion will reshape the map, but who will have a say in how.

📬 Sign up for the Daily Brief