Cloud & InfraArticle Landowner fights over what counts as 'public use' now shape where, and how fast, new cloud capacity gets built.
Emeka Okafor The AI buildout has a land problem, and it's about to become a legal one. Across Georgia, Maryland, Virginia, and Pennsylvania, utilities are invoking eminent domain — or threatening to — to route high-voltage transmission lines to data center campuses. Landowners are suing, and the question the courts have to answer is older than the republic's grid: when a power line exists mostly to feed a handful of private hyperscale customers, is taking someone's farm for it still a "public use"?
That question sounds like it belongs in a law review, not a dev publication. It doesn't. Power delivery is now the single hardest constraint on where new cloud and GPU capacity gets built, and eminent domain is the mechanism that determines how fast transmission gets sited. If the courts tighten the doctrine — and there's a real chance they will — the timeline for every "coming soon" region and GPU cluster stretches with it.
The extension cord problem #
The scale is worth stating plainly. Data centers consumed more than 4% of US electricity in 2024, and the Lawrence Berkeley National Laboratory projects that share could roughly double or triple by 2028. Over 3,000 data centers operate in the US today with another 1,500 in development. Meeting that load means new generation, but mostly it means new wires — and wires cross other people's land.
The standard playbook: a utility offers to buy an easement. If the landowner refuses, state law typically lets the utility condemn the property, because courts have long treated electric transmission as the canonical public use under the Fifth Amendment's takings clause. That assumption held for a century because the grid genuinely served everyone downstream.
The data center era breaks the assumption. Georgia Power is acquiring more than 300 parcels — including occupied homes — for a 35-mile, 500 kV line out of Plant Wansley, and the company itself says 70–80% of the power on that line will serve data centers. In Maryland, PSEG's roughly 70-mile Piedmont Reliability Project cuts through farmland in three counties largely to relieve congestion driven by Northern Virginia's data center cluster. Locals call it "the extension cord," and the label is doing real legal work: Maryland's ratepayer advocate has filed a complaint at FERC arguing that around $2 billion in data-center-driven transmission costs is being unfairly assigned to Maryland households.
"Public use" was never built for this #
The constitutional floor is low. In Kelo v. City of New London (2005), the Supreme Court blessed takings for private economic development, so a federal challenge to a data center line is a long shot. But Kelo triggered a backlash — 45 states passed eminent domain reforms afterward — and that's where the fight actually lives.
State law is a patchwork, and law professor Aaron Walayat's recent analysis maps the fault lines well. Michigan, Ohio, and Oklahoma courts bar takings whose benefit is purely private. Texas requires that a project "serve the public" rather than exist for the builder's exclusive use. South Dakota and Vermont have upheld utility condemnations — but on the theory that the lines served in-state customers. Mississippi's supreme court rejected a taking in 1984 precisely because the line ran to Louisiana and delivered nothing to Mississippians.
Notice what all of those tests hinge on: who actually benefits. A transmission line that's 75% dedicated to four AI campuses, or that wheels power across a state to another state's server farms, is a genuinely harder case than anything these doctrines were built for. Utilities will argue — correctly — that electrons are fungible and any new capacity improves regional reliability. Landowners will argue — also correctly — that "reliability" here means reliability for the data centers whose load created the problem. Both framings are defensible, which is exactly why this is headed for state supreme courts.
My read: the utilities win most of these cases on current law, but the litigation itself is the cost. The Piedmont project shows the shape of it — PJM, the mid-Atlantic grid operator, warned of blackout risk and set a June 2027 in-service deadline, and the project is instead in the Fourth Circuit arguing about whether PSEG can even survey land before it has a construction certificate. Every month of that is a month of capacity that doesn't energize.
What this means if you build on the cloud #
Three concrete consequences for people who ship software rather than power lines.
Capacity geography gets stickier. Transmission fights favor incumbency: it's far easier to upgrade an existing easement than to condemn a new corridor through angry counties. That's a structural force pushing new capacity into places that already have it — one more reason us-east-1 and the Northern Virginia colo market keep sprawling instead of demand dispersing to cheaper regions. If your DR plan assumes a wide menu of well-stocked regions materializing on schedule, discount it.
Power politics shows up in your invoice. The Maryland FERC complaint is the leading edge of a cost-allocation war: who pays for data-center-driven grid expansion, ratepayers or the customers driving the load? The likely settlement is special large-load tariffs that push transmission costs onto data center operators — several states are already drafting them. Hyperscalers won't eat that; it flows into colo lease rates and, eventually, cloud and GPU pricing in affected regions.
Behind-the-meter generation stops being exotic. If wires take a decade and a lawsuit, generation you build inside the fence line doesn't. This is the same pressure behind hyperscaler nuclear PPAs — Microsoft's deal to restart Three Mile Island's Unit 1 being the marquee example — and the on-site gas turbines showing up in new campus designs. Expect "grid-independent" to become a genuine selling point in colo negotiations, and expect the sustainability numbers attached to it to deserve scrutiny.
Where this lands #
Seven in ten Americans say they don't want a data center near them, and the industry's social license is eroding faster than its megawatt pipeline is growing. Eminent domain is where that abstract resentment gets concrete standing: a farmer with a condemnation notice is a much more sympathetic plaintiff than a county board worried about "character."
The AI industry has treated land and wires as a procurement detail. The next few years of state court decisions will decide whether that stays true. If even a couple of state supreme courts hold that a line built 75% for private compute isn't a public use, utilities will have to restructure how these projects are justified and financed — and every capacity roadmap downstream of them will slip. Watch the Georgia PSC docket on Project Wansley and the Piedmont litigation in Maryland. They're infrastructure cases, but the thing being decided is how fast the cloud gets bigger.
Sources & further reading #
When can a power company take your land for data center infrastructure?— theconversation.com - Power companies are using eminent domain to seize land for data centers as 70% of Americans say not in my backyard— fortune.com -
[Georgia family says they're forced to sell home to help power AI data centers](https://www.cbsnews.com/news/georgia-power-ai-data-centers-eminent-domain/)— cbsnews.com -
[Blackout fears debated at 4th Circuit hearing on proposed Maryland power line](https://marylandmatters.org/2026/05/05/piedmont-power-line-4th-circuit/)— marylandmatters.org -
Maryland lawmakers back data center transmission cost complaint at FERC— utilitydive.com -
[Georgia Power has a massive plan for new power lines](https://www.gpb.org/news/2026/06/23/georgia-power-has-massive-plan-for-new-power-lines-what-does-mean-for-the-homes)— gpb.org
[Emeka Okafor](https://sourcefeed.dev/u/emeka_okafor)· Security Editor
Emeka has spent over a decade tracking threat actors, vulnerability disclosures, and the evolving landscape of application security, bringing a sharp continent-spanning perspective to his reporting. He's known for translating dense CVE advisories into clear, actionable context that developers and security teams alike actually read.
Discussion 0 #
No comments yet
Be the first to weigh in.