Texas regulators freeze new data center grid connections pending a comprehensive reliability audit.
Texas has halted processing new data center connections to its electric grid. In an August 3 letter to state regulators and the grid operator, Governor Greg Abbott ordered a comprehensive audit. The directive was triggered by a figure that stopped resembling a forecast and started looking like a filing cabinet: approximately 474 gigawatts of pending interconnection requests.
Texas recorded an all-time peak demand of 91,089 megawatts on July 22. The current interconnection queue exceeds that figure by more than five times, encompassing over 1,800 projects. Abbott noted in his letter that “Approximately 90 percent of the new power requests are data centers.”
This surge was not gradual. Interconnection requests hovered near 48 gigawatts in 2023 but surpassed 474 GW by this summer. Abbott warned that load growth at this magnitude threatens grid reliability and stability, mandating that any project failing the verification and audit process be denied grid access.
ERCOT responded within hours. Through Market Notice M-A080326-01, the grid operator suspended the Batch Zero interconnection study, effectively canceling the classification notifications developers had anticipated receiving by August 7.
Utility planners refer to the discrepancy between requested and actual capacity as “ghost demand.” Submitting an interconnection request carries minimal cost, prompting developers to file simultaneously across multiple territories to preserve options before constructing facilities in whichever jurisdiction clears first. Until recently, these filings went entirely unscreened before being incorporated into demand forecasts.
Evidence that much of the queue is speculative is already emerging. After tightening its qualification criteria, Exelon saw its high-probability data center load estimate fall by approximately 40%, settling at roughly 11 GW. Similarly, when Ohio raised interconnection study fees to $100,000, reported data center demand at AEP Ohio dropped by more than half. Imposing even a modest upfront cost successfully filtered out the majority of speculative requests.
A review of utility filings across the Midwest, Mid-Atlantic, and South indicates that national large-load requests now exceed 700 gigawatts. To contextualize this scale, the entire operational U.S. data center fleet currently consumes an estimated 60 to 70 GW at any given time.
In June, PJM stated that data centers are driving the bulk of its demand growth, cautioning that new facilities can be constructed two to three times faster than the generation capacity required to support them. Meanwhile, New York paused new data center approvals in July for up to one year while drafting updated development regulations.
Neither ERCOT nor the Public Utility Commission of Texas has released a timeline for completing the audit, and Abbott’s directive established no deadline. Consequently, developers currently positioned in the queue lack a definitive date to plan around—a delay that carries its own financial burden.
BloombergNEF quantified the financial impact of the delay, estimating losses of up to $15 billion across affected projects, with 49.8 GW at risk. The firm also highlighted the political context, noting that Governor Abbott faces re-election in November.
The root issue is a systemic design flaw that distributed systems engineers resolved decades ago, yet utilities are now forced to address from scratch. When an interconnection queue accepts unlimited, cost-free requests, it ceases to function as a reliable indicator of actual demand. Grid operators spent two years treating these filings as legitimate forecasts and basing generation plans around them—resulting in a state preparing for five times its historical peak electricity usage. Global analysis of developer behavior reveals that some entities submit speculative requests for five to ten times their intended capacity solely to secure a position in line.
The implications extend far beyond Texas. Approximately one-fifth of the national data center pipeline is now stalled pending an open-ended audit in the state that has already absorbed more AI infrastructure than any other. Utilities nationwide are closely monitoring which screening mechanisms prove effective—whether deposits, milestone requirements, credit checks, or study fees. The successful model will likely establish the admission standards for the broader industry buildout. Anyone managing a project in any interconnection queue should review the full details of the Texas executive order and ERCOT’s subsequent response.