AI Data Centers vs. Homeowners: How a $16B Grid Expansion Is Rewriting Property Rights in Georgia

When Infrastructure Meets Eminent Domain

Eminent domain is old law doing new work.

Originally designed to serve genuine public necessity — highways, hospitals, public utilities — it’s now being deployed to power private AI infrastructure at scale. Georgia Power, a subsidiary of Southern Company, is legally authorized to exercise it. The company insists it’s a last resort, used in less than 1% of annual land transactions.

That statistic offers cold comfort when you’re one of the 1%.

Between 20 and 30 homes face outright demolition. Hundreds more are being asked to accept permanent easements that plant high-voltage towers feet from bedroom windows. Some research suggests safe residential distance from 500-kilovolt lines starts at 300 feet. One homeowner was offered an easement that would place the line 87 feet from her house, with 12 feet between the structure and the boundary.

The Math Nobody Wants to Do

Cynthia Van Epps has done the math. It’s brutal.

The Coweta County homeowner of 17 years calculated that accepting the proposed easement — which would consume more than half her land — would leave her paying an estimated $300,000 in property taxes over the next 30 years on land she cannot use, build on, or sell. Meanwhile, Georgia Power runs revenue-generating transmission lines across that same land for decades.

“You’re paying property taxes for the very power lines that they’re making billions of dollars off of,” she said.

That’s not a hypothetical grievance. It’s a structural absurdity baked into how easement law interacts with property tax assessment. The land stays on your books. The utility gets the use. You get the bill.

A Viral Moment With a Very Real Backstory

Ansley Brown didn’t set out to become the face of this fight.

She posted a video about her family’s situation — her mother’s childhood home, a USDA loan for single mothers, an appraisal that came in $70,000 to $100,000 short of market value, and a lawyer’s letter threatening condemnation if they didn’t settle. The video crossed 6 million views.

The story resonated because it compressed something abstract — grid infrastructure, AI demand, utility law — into something viscerally human. A federal program helped a single mother buy a rural home. Decades later, a private utility backed by a $26 billion government loan wants to take it to power data centers.

The irony isn’t subtle. It’s structural.

Georgia Power’s Position

To be fair, Georgia Power’s account differs on the specifics.

The company says it negotiated with Brown’s mother for over four months and ultimately offered $205,000 above her own appraisal — more than $275,000 above Zillow’s estimate. It says she accepted the financial terms in May, then declined to proceed. On underground burial, the company says it evaluated the option and found it neither practical nor feasible at this voltage and scale, citing construction complexity, longer repair times, and significantly higher costs.

Georgia Power also pushes back on the “data center project” framing, describing the Ashley-Park Wansley corridor as regional infrastructure serving manufacturing, residential demand, and grid reliability broadly — not built for any single customer.

That’s technically accurate. It’s also a bit like saying a highway isn’t built for trucks because cars use it too.

The Bigger Pattern: AI Infrastructure Has a Footprint Problem

This isn’t a Georgia story. It’s a preview.

As AI model training and inference demand explodes, data centers are multiplying across the U.S. — and each one needs power, cooling, and transmission capacity that didn’t exist five years ago. Utilities are racing to build the grid to match. That means new lines, new substations, and new corridors cut through land that people actually live on.

The tension is real and growing: AI infrastructure investment is accelerating faster than the legal, regulatory, and community frameworks designed to manage its physical footprint. Eminent domain law was written for a different era of public necessity. It hasn’t caught up to a world where “public benefit” increasingly means “private compute capacity.”

State legislators in Georgia are starting to ask questions. Rep. Brian Jack’s office reached out to Brown. State Sen. Greg Dolezal walked her mother’s property. The Public Service Commission is being pressed on whether it granted blanket eminent domain authority over all 330 properties without individual review.

Those are the right questions. Whether they arrive in time is another matter.

What This Means for the AI Tools Ecosystem

Most readers here aren’t Georgia homeowners. But they are AI adopters — people building with, investing in, or evaluating tools that run on this infrastructure.

The physical cost of AI is becoming visible in ways it wasn’t before. Every model inference, every API call, every cloud-hosted workflow has a power address. And that address is increasingly located in someone’s backyard, whether they agreed to it or not.

For founders and operators, this is worth tracking — not just ethically, but practically. Infrastructure bottlenecks, regulatory pushback, and community opposition are already slowing data center permitting in multiple states. That affects capacity. Capacity affects pricing. Pricing affects your stack.

The AI boom is real. So is the grid it requires. And so are the people standing in its path.

The most expensive infrastructure is rarely the kind you can see on a balance sheet. Sometimes it’s a childhood home in Coweta County that a 27-year-old is fighting to keep — while a 6-million-view video quietly rewrites the public conversation about who the AI economy is actually built for.