The usual worry about Texas’ data center boom is that the state is flying partly blind. It does not yet know how much electricity these projects will actually need, how much water they will take, or whether the tax breaks offered in the name of growth will return enough to justify their cost.
That sounds, at first, like a paperwork problem: collect better numbers, build a better spreadsheet, close the gap, and the problem is solved. But the longer you sit with the missing information, the less accidental it looks. The uncertainty does not fall evenly. Again and again, it gives developers room to maneuver while leaving the public to plan around promises it cannot fully test.
Start with demand. ERCOT’s queue of large-load requests, meaning any single site asking to draw more than 75 megawatts, now totals roughly 439 gigawatts. Data centers account for almost 90 percent, with crypto mining operations and industrial plants making up most of the rest. For scale, the most electricity Texas has ever drawn at one time is 87,403 megawatts, set on July 21 of this year.. The queue is more than five times that. Separate estimates put the capacity that data centers have actually connected and are drawing today at closer to 7.5 gigawatts.
For years the development queue cost little to join and bound no one to build, but that inflated total did real work by manufacturing a sense of inevitability that pressured the grid to build ahead of it. Texas has only just now begun closing that door, requiring large-load applicants to post real financial security and prove land control before a request counts. It’s a genuine fix, but it only touches one ledger.
Water is the project resource that Texashas always asked about but never tied to a real requirement, and it works the same way demand once did—only more brazenly. Data centers heavily consume water in order to cool servers and to run the plants that power them. When the state asked the industry to report how much they are using, fewer than a third of the companies queried responded. The Texas Water Development Board runs a separate, mandatory survey, yet only 17 percent of data centers answered in 2025 (despite the penalty for ignoring it being a Class C misdemeanor). Required or not, the number never arrives, and a figure the state never receives is one it cannot price, cap, or otherwise meaningfully regulate.That silence is less a gap in the record than a form of leverage. Vagueness here is not a hole in the industry’s case; It is the entire case.
The gas plants are hardest to hide, and yet developers still have a sanctioned way around scrutiny. Rather than wait in line, they increasingly build their own power on site—withsome 38 gigawatts of it in development—and the same reform that now demands firm commitments from the grid queue leaves this exact path open. Those plants still need air permits, so the pollution does get counted. But a permit only measures what comes out of the stack. It never asks whether the data center it serves is real, or who covers the cost of the infrastructure if it never arrives. Those questions belong to the grid connection process, and a company that builds its own power never has to enter it. The smoke gets measured. The gamble does not.
The media coverage of the boom keeps arriving at this pattern and stopping short of naming it. Whatever any one company intends, the pattern is too consistent to pass for accident. The lever is not missing information. Sometimes the industry files far too much, like that padded queue of connection requests, and sometimes it files almost nothing, like the water usage survey. What runs through both is asymmetry: the developer usually knows more than the state and reveals only what suits the project. Call it what it is: a system that rewards whoever controls the information, and never the public. That same imbalance lets Texas grant a sales tax exemption now on pace to become the nation’s costliest, without a clear account of what it was buying. “Gather more data” assumes the data is lying around waiting to be picked up, when really it sits with whoever profits from Texas never getting its hands on it.
Texas has courted this industry, and that exemption is one measure of the welcome. So it was notable when Governor Greg Abbott wrote to state regulators in June about data center costs. Look at what carries a deadline and what does not. The PUC must make data centers fund their own infrastructure; the PUC and ERCOT must file a joint memo; the commission must act on transmission costs by July 31. All electricity, all with dates attached. Then come the water items, and the register shifts. Water reporting, water-efficient cooling, an end to outdated tax incentives, these are things the governor pledged to take up with the Legislature next session. That session convenes in January 2027. The mandatory water survey Texas already has, the one 17 percent of data centers bothered to answer, is already law. (Abbott also recently touted that a controversial East Texas data center project was pulling its plans because it didn’t meet the governor’s requirements, providing him a useful pelt to hold up to anti-data center voters as proof that his recent actions are having an effect. )
None of this means the data demand surge is imaginary. AI computing is growing fast, and some of those filings will become real buildings pulling real power off the grid. The industry brings real investment, and Texas has reason to want it. That was never the question. The question is whether Texas should commit decades of generation and water to requests never guaranteed by the companies making them. Growth does not become a public obligation just because it sounds inevitable. In an interconnection queue, a firm plan and a hedged bet look exactly alike. Only one of them should be enough to break ground.
The useful move is not to plead for numbers. It is to make the absence of them expensive for whoever withholds them. And on energy demand, Texas has finally started doing that by doing whattktk Ohio ran the same test and the result was striking. Data center developers had filed roughly 30 gigawatts of interconnection requests with AEP Ohio. Once regulators required binding contracts backed by real collateral, just 5.6 gigawatts signed. The rest evaporated, which was the point.
Water deserves the same treatment, and so does the on-site generation path left open alongside it. Tie either hookup to real disclosure so that a company earns it by opening its books. And when a speculative plant winds up half-used, let the company that built it swallow the loss, not the ratepayer who never asked for it. None of that requires predicting the future. None of this is theoretical either. Texas has run into it before. Oil and gas has powered the state for a century, and most of its wells are plugged and retired properly. The trouble was always narrower than the industry: bonds have long been set below what plugging a well actually costs, so when a company fails, the well becomes the state’s. More than 11,000 now sit orphaned on the state’s rolls, and recent watchdog estimates put the state-funded plugging cost above $55,000 per well, with taxpayer dollars covering the difference. The point is not that a data center is an oil well, or that the industry did anything wrong. It is that Texas already knows what happens when a cost is not secured up front, and the current boom is a chance to secure it early instead of late.
Texas does not have to know the future to guard against it. It has to ask for firmer commitments before public systems plan around a company’s word, and make the people selling the boom carry the risk if it fails to arrive. The computing is real and it is not going anywhere. The only open question is who pays when the promise outgrows the thing beneath it.
The post With Texas’ Data Center Boom, Uncertainty Is the Business Plan appeared first on The Texas Observer.
Leave a Reply