Texas is moving deeper into its review of large data center projects connected to the state power system. Utility Dive reported on August 21, 2026 that the Electric Reliability Council of Texas intends to complete a broad audit by December 10. The review follows Governor Greg Abbott’s August 3 call for a moratorium on new data center interconnections until questions about electricity, water and public financial support can be answered.
The headline number is extraordinary. ERCOT’s interconnection queue contains about 474 gigawatts of requests, and Abbott said roughly 90 percent of new power requests are associated with data centers. That does not mean 474 gigawatts of projects will be built. It shows how difficult it has become for utilities to separate serious demand from speculative or duplicated requests. Sensaka’s guide to data center capacity planning describes the facility level version of the same problem: nominal capacity is less useful than capacity that can actually be powered, cooled and operated.
The audit is about project credibility as much as electricity
ERCOT is reviewing hundreds of large proposed facilities and asking developers for more information. Utility Dive reported that around 300 data centers of 75 MW or larger are navigating the Batch Zero process. Community impact reviews will also apply to data centers and crypto facilities of 25 MW and above.
The state wants to know how much electricity projects expect to use, whether they plan to provide their own generation, how much water they need and whether they depend on public incentives. Those questions turn a data center proposal into a broader infrastructure commitment. A campus is no longer evaluated only on land, financing and customer demand. It must also explain its relationship with the grid and the surrounding community.
This makes power modeling central to development. Sensaka’s data center power calculator illustrates the basic relationship between device load, current and continuous operating limits. At utility scale the same discipline applies, although the numbers involve substations, transmission and generation rather than rack circuits.
The queue shows the difference between requested and usable capacity
Large interconnection queues can exaggerate future demand because developers may submit requests for several possible sites or reserve capacity before financing and customers are fully committed. Utilities still have to plan for the possibility that some projects are real, because transmission and generation can take years to build.
Texas is therefore trying to establish which loads are credible enough to shape long term planning. The delay matters because a data center can complete other parts of its development while waiting for certainty about energization. That creates financial risk for developers and their suppliers.
Inside a facility, a similar mismatch produces stranded capacity. A hall may have open rack positions but no remaining power or cooling headroom. Sensaka’s overview of data center management connects power, cooling, assets and operational capacity because each limit changes what the building can actually support.
Water and community impact are moving onto the critical path
Texas is also asking data centers to explain water consumption and whether they will rely on supplies needed by local communities. This matters because cooling design can shift the balance between water and electricity use, especially in large AI deployments.
Community impact is becoming harder to separate from technical design. A project may need new transmission infrastructure, backup generation, water systems and road access. Residents may care about noise, land use and utility prices as much as the developer cares about latency and fiber routes.
For data center companies, those concerns should be treated as project dependencies. Cooling architecture, site design and power strategy can influence approval timelines. Transparent operating data can also become useful evidence when regulators want to understand whether promised efficiency or demand management measures are actually working.
December will not necessarily end the planning problem
ERCOT’s target is to deliver a comprehensive report around December 10, but Utility Dive reported that the original Batch Zero study timeline is already unlikely to hold. Grid officials said they do not yet know how much the audit will shift the next stages of interconnection review.
That uncertainty is the larger lesson. AI investment can move faster than transmission planning, power plant construction and public approval. Developers may have capital and customers while the physical infrastructure underneath the project remains unresolved.
Texas is still likely to remain one of the most important United States data center markets. The current pause shows the condition attached to that growth: projects will increasingly need to prove that their requested capacity is credible, supportable and compatible with regional infrastructure. In the AI era, the ability to demonstrate usable capacity may become as important as the ability to announce it.
Originally published on the Sensaka blog.
Top comments (0)