Texas Orders Audit of Data Centers Before They Can Connect to Its Grid

Texas Governor Greg Abbott on August 3 ordered state regulators to audit every data center seeking a connection to the state's main power grid, and said no project can move forward until its audit is done. The directive, announced by the governor's office, effectively puts new data center hookups in one of the country's biggest AI infrastructure markets on hold. For enterprises counting on new cloud and colocation capacity in Texas, it adds a fresh source of delay and uncertainty to AI infrastructure plans.

What did Governor Abbott order?

The directive tells the Public Utility Commission of Texas (PUCT) and the Electric Reliability Council of Texas (ERCOT), which runs most of the state's grid, to verify and audit all data centers moving through ERCOT's interconnection process. Projects that do not meet PUCT, ERCOT and state requirements will be denied a grid connection, according to the governor's office.

Each project will have to provide detailed information, building on a June 10 directive. The list includes whether the project relies on state or local tax incentives, grants or abatements; how much power it expects to use at peak and over a year, and how much it will generate on site; where its water comes from and what cooling technology it uses; how it limits noise, lighting, traffic and other effects on neighbors; and who owns and controls it.

"Our top priority is to protect Texans' safety and quality of life," Abbott said in the announcement. The release did not say how long the audits will take.

Why is Texas auditing data centers now?

The short answer is the size of the queue. ERCOT is considering more than 474 gigawatts of requests to connect new large loads, more than five times the state's record peak electricity demand, and about 90% of those requests come from data centers, the governor's office said.

That number has grown quickly. ERCOT's queue held 233 gigawatts in January and more than doubled in less than six months, TechCrunch reported on August 4. Many of those requests are speculative, and the publication noted that many projects are expected to drop out. That is part of the problem the audit is meant to solve: grid planners cannot tell which projects are real.

TechCrunch also reported that the governor had earlier asked data center developers to answer a voluntary survey and that most did not respond. Rising electricity prices are another factor. The publication cited federal data attributing part of recent increases to data centers and crypto-mining sites, while noting that Texas prices remain relatively low compared with other states.

What does the Texas audit mean for enterprise AI capacity?

Texas has been one of the most attractive places to build AI data centers because of its land, its fast-growing power supply and its relatively light permitting. Google and Microsoft have both been drawn by the state's natural gas supply, according to TechCrunch, which concluded that the state's run as a data center hub could depend on how the audits play out.

For enterprise buyers, three effects are likely. First, timelines for new capacity in Texas may slip while projects are reviewed, which can ripple into the delivery dates cloud and colocation providers promise their customers. Second, projects that bring their own power are likely to look better under the new criteria, since the audit explicitly asks about self-supplied versus grid-supplied electricity. Third, water use and community impact are now part of the approval file, so providers with weak answers may face longer delays.

The physics behind the queue is simple. "A thousand watt GPU needs a thousand watts of cooling," Steve Heibein, Federal AI Chief Technologist at Hewlett Packard Enterprise, said during a AI Connect expert webinar on why AI has become a supercomputing problem. Dense AI racks draw far more power per square foot than the enterprise data centers they replace, and states are starting to push back.

What should CIOs and infrastructure leaders do?

Ask providers where planned capacity is located and whether any of it is waiting on an ERCOT connection. Contracts that promise capacity in a specific region should spell out what happens if a regulatory pause delays delivery.

It is also a good moment to revisit how much compute AI workloads actually need. AI Connect speakers have repeatedly argued that infrastructure and inference cost decisions should follow the workload, from splitting CPU and GPU tasks to right-sizing models. Panelists in a session on AI factories made a similar case for on-premises options when regulated data or capacity limits are in play.

Next to watch: how quickly PUCT and ERCOT publish the audit process and timelines, how developers respond, and whether other states with crowded interconnection queues adopt similar data requests.

Frequently asked questions

Did Texas ban new data centers?

Not formally. The governor ordered an audit that must be completed before any data center project in ERCOT's interconnection process can advance, which effectively pauses new connections.

Why is Texas auditing data centers?

ERCOT's queue of new large-load requests has grown to more than 474 gigawatts, over five times record peak demand, and many requests are speculative.

What information must data centers provide?

Tax incentives, self-supplied versus grid power and peak demand, water sourcing and cooling, community impact measures, and ownership.

Filed under: Data & AI Infrastructure, AI Governance & Trust