There's a version of the AI story that gets told at every conference: chips are scarce, power is scarce, and capital is limitless, pouring in by the hundreds of billions. It is a good story in hindsight. It is also, increasingly, a lie of omission. Because while the industry obsessively solves for hardware and energy, it leaves out the single factor that can halt a project overnight: whether the people living next to these campuses will actually let them stay. You can buy the land, secure the chips, and contract the grid capacity, but without local consent, the entire buildout stops.
For twenty years, Northern Virginia was the center of the digital universe, built on cheap land, early fiber, and proximity to federal networks. It quietly became the largest concentration of data center capacity on the planet, and it still is, at nearly 7 GW. But empires outgrow their borders and the grid is maxed out. The queue to plug in new capacity now stretches for years, with no amount of capital moving you up the list. You cannot build what the grid cannot carry, no matter how important you think you are.
So the industry did what empires do when they hit a wall: it expanded outward. Today, 77% of all data center capacity under construction in the US sits in what the industry calls "frontier markets"—places like West Texas, Ohio, Louisiana, and the Carolinas, all territory that had no data center industry to speak of five years ago.
What do the frontier markets offer? Not fiber. Not customers nearby. Just power that’s available right now. Speed to power has become the only site selection criteria that seems to matter anymore and the appetite is staggering. Vacancy across North America has sat at essentially zero for three straight years, and more than 66 GW is under construction right now—more electricity than the entire country of Germany consumes. The industry has moved into towns that have never hosted anything at this scale, largely treating the power connection as the only approval that mattered. But a power connection isn't the same as a green light from the people who live there, and San Marcos, Texas made that difference impossible to ignore.
A developer walked into San Marcos with a $1.5 billion plan and power demand that could hit 2.5 times the entire city's peak electrical load. More than a hundred residents showed up, said no, and the project didn't survive the room. In the first three months of this year alone, local opposition blocked or delayed roughly $130 billion in data center projects across the country. Two years ago, this was framed as a race for three things: chips, power, and capital. Nobody priced in the fourth. Permission—and permission, unlike power, cannot be bought or engineered around.
Can any industry outrun the consent of the people it's building on top of? History has answered this question before, in oil towns, in coal country, and in every boom that assumed the local population was scenery rather than a stakeholder. The answer is always the same: eventually, no. Every frontier market absorbing this pipeline is walking toward its own San Marcos moment. The industry solved for power in five years. Can it solve for permission just as fast?
None of this means the buildout should slow down or that frontier markets should be wary of what's coming. Done well, these projects bring real investment, tax revenue for schools, construction jobs, and grid upgrades that benefit the whole region. Plenty of developers are already getting this right, going in with community benefit agreements, transparent water and energy commitments, and local hiring plans. Those projects tend to move faster, not slower, because they've already done the hard part.
The lesson from San Marcos isn't that communities are against data centers. It's that the industry can't assume consent anymore. It must earn it, the same way it earns everything else it's paid billions for.
