Texas Demanded Data Center Receipts. New York Hit Pause. The AI Buildout Just Got a New Gate.

On Aug. 3, Texas Gov. Greg Abbott directed the state's grid operator and utility regulator to audit every large data center project waiting to plug into the Texas grid. Within a week, ERCOT, the Electric Reliability Council of Texas, which manages about 90% of the state's load, postponed its "Batch Zero" transmission planning study. Three weeks earlier, New York had imposed the first statewide pause on permits for new hyperscale data centers, the largest class of facility, typically drawing hundreds of megawatts.

The bottleneck on the AI buildout is no longer power, water, or chips. It is disclosure.

ERCOT is currently tracking more than 1,800 large-load projects, new industrial electricity customers above a defined size, representing more than 474 gigawatts of proposed demand. That figure is more than five times the grid's record peak load. Abbott attributed roughly 90% of that demand to data centers, according to EE Times.

The scale of the queue is what turned voluntary disclosure into mandatory disclosure. Texas officials had asked 377 companies with active data center proposals to fill out a state survey on water use, power needs, and cooling. Twenty-eight responded. That 7.4% response rate is the diagnostic. Voluntary compliance failed; the August directive is the lever being pulled in response.

Texas's directive requires developers to disclose ownership, incentives, power and water needs, cooling systems, self-generation plans, and local-impact measures before any project moves forward. New York's July 14 order, by contrast, is a temporary moratorium on certain discretionary environmental permits while rules are drafted on electricity costs, water, emissions, noise, and community benefits. The two mechanisms differ, but they share an underlying premise: the public now gets to see the deal before the deal gets built.

Public skepticism is the political backdrop. A June Reuters/Ipsos poll found 57% of Americans would oppose a data center in their community, 14% would be comfortable with one nearby, and 77% were concerned the AI data center boom would raise local costs, as EE Times reported. The two state actions land on top of that polling, not separate from it.

The trade and analyst press has been describing this same pattern as a "social license to build" problem. Teneo, a global advisory firm, has put the price of closing that gap at roughly $1 trillion, not the cost of building, but the cost of building under terms the public will accept. Utility Dive frames the constraint as community opposition, transparency, and rate-base allocation, meaning who pays for transmission upgrades and who gets cheaper power.

The disclosure test, not the megawatt count, is now the binding constraint on the AI buildout. A project that cannot answer who owns it, where its water comes from, how much grid capacity it will consume, and what local ratepayers will be on the hook for is a project that does not break ground. That is a new condition. Until this year, the gating factors on hyperscale construction were transformer lead times, GPU allocations, and interconnect study backlogs. To those, add a fourth: a state-level information test.

The next test case is the Texas survey itself. ERCOT and the Public Utility Commission of Texas now have the authority to require answers; the question is whether developers fill in the form or contest the directive. The most likely response is selective compliance, where large hyperscalers with existing utility relationships submit, while smaller developers stall, which would push Texas toward a formal rule rather than an executive ask. New York's path is slower: a draft framework is due after the moratorium expires, and the substantive rules will be hashed out in rulemaking.

The next data center fight in any state will turn on the same question Texas just asked and 349 of 377 developers declined to answer: what is this project, who owns it, and what does it cost the public?