The Variable Nobody Priced Into the Data Center Model

Beneath the Waterline | Issue 5

Every site selection model has four lines. Power. Land. Cost. A cooperative jurisdiction.

Three of those are numbers. The fourth is a wish wearing a number's clothing. And in Texas right now, that fourth line is on fire.

Texas was the safe bet. Cheap power, open land, and a regulatory posture that rolled out the welcome mat. Roughly 400 data center projects are in development or operating across the state. Every one of those models carried "friendly regulatory environment" as a fixed input, treated like the price of concrete. Settled.

It was never settled.

In May, Hill County commissioners voted to pause data center and power plant construction in unincorporated areas, citing as many as eight projects bearing down on them at once. Within weeks, they were sued in federal court, and they swapped the moratorium for a checklist. Hood County tried twice and failed, after a state lawmaker declared the ban illegal before the vote was even cast. A Houston senator asked the Attorney General to investigate any county that dared to try. And then the governor, the man who courted all of this, told the utility commission in June that data centers will pay for their own power and stop pushing costs onto residents. Five months before he stands for reelection.

That is not a policy debate. That is a fixed input becoming the most volatile number on the page, in real time.

Here is what actually changed, and it is not legal. It is arithmetic that finally became visible to the people paying for it.

For years the benefits of these facilities left the county: the tax abatement, the handful of permanent jobs, the press release. The costs stayed home: the grid strain, the water draw, the rate hike on a retiree's bill. A community will tolerate that trade for a long time, right up until it does the math. Last year, about half of Americans didn't want one near them. A few months later, seventy percent. Across the country, more than 75 projects worth roughly 130 billion dollars were blocked or delayed in a single quarter. That matched the total for the entire prior year.

The consent of the people living next to the machine was always part of the cost structure. The industry just never wrote it down.

So here is what I would tell the executives I work with, the ones building five-year plans that assume the welcome mat stays put.

First, treat the jurisdiction the way you treat the power draw. You would never model a facility without knowing the megawatts. Stop modeling one without knowing the political load: the election calendar, the water table, and the temperature of the last three commissioners' meetings. Consent is a line item now. Price it.

Second, stop calling this a legal problem. Legal handles the permit after the architecture is already set, and by then, the constraint is baked in and the only options left are the expensive ones. The siting decision is an architecture decision. The compliance posture, the water profile, the cost-passing structure: these belong in the design phase, next to the storage and the network, not in a memo from counsel after the neighbors show up.

Third, build for the moratorium you cannot see yet. The agent-ready data estate I help organizations design has always had to account for what sits outside the building, not just inside the rack. A foundation that assumes a fixed regulatory environment is a foundation built on someone's mood. Assume the rules will move. Design for the version where the easy answer disappears, because in Texas, it just did.

None of this is an AI problem. The agents and the models are downstream of a decision made years earlier by someone who modeled compute and capital and never modeled consent. You cannot deploy what the county will not let you build. You cannot govern an estate whose foundation assumed a welcome that was never guaranteed.

The enterprise priced the silicon, the steel, and the electricity. It forgot to price the people. That bill is coming due, and it is local.

If you found this briefing valuable, share it with a colleague who is navigating the shift from AI hype to operational reality.