On August 3, 2026, Texas Governor Greg Abbott ordered the Public Utility Commission of Texas and ERCOT to audit every data center in the state's interconnection queue. The directive paused all new grid connections until the audit was complete — with no published deadline for when that would happen.
The numbers behind the decision explain why Abbott acted. ERCOT's interconnection queue had swelled to roughly 474 gigawatts of requests. That is more than five times the grid's record peak demand of 91,089 megawatts, set on July 22, 2026. Data centers accounted for approximately 90 percent of that total.
BloombergNEF estimated that the freeze could delay 49.8 gigawatts of data center load — nearly 20 percent of the entire U.S. development pipeline — and cost projects up to $15 billion. By the first quarter of 2027, the financial impact could reach $8 billion.
The queue's scale suggests a significant portion of requests may be speculative.
What the Queue Actually Contains
The audit is not about slowing Texas down. It is about distinguishing real projects from ghosts.
ERCOT officials have acknowledged that some interconnection requests may be speculative — developers holding a place in line without financing, site control, or a credible path to construction. "A place in the queue is not proof the grid should plan around it," said Neil Osnato, founder of Persistence Analytics Group.
Under the new audit, developers will need to show financing, site control, permitting, water and power availability, and a credible construction timeline — rather than simply holding a position in line. Abbott's directive explicitly requires disclosure of whether data centers provide their own power or rely on the grid, their water use, and whether they are using state or federal incentives.
Carbon Direct found that of ERCOT projects that began screening by 2020, only 40 percent reached interconnection agreement status or became operational. The other 60 percent have not — and likely never will.
The national picture is worse. Lawrence Berkeley National Laboratory puts the active U.S. interconnection queue above 2,600 gigawatts, with median wait times above 4 years and rising.
The Waiting Time Problem
Even for projects that do get built, the timeline is punishing.
FERC's federal target for an interconnection agreement is 8 to 11 months. The national average is 25 months. In PJM, the grid serving the mid-Atlantic and Midwest, it is 40 months. Active projects in data center load-growth zones are waiting 36 to 48 months in both PJM and ERCOT.
"The interconnection queue is one of the most consequential and least understood bottlenecks in the energy evolution," said Derya Eryilmaz, PhD, Vice President of Power Commercialization at Carbon Direct.
The bottleneck is not generation. It is the transmission infrastructure required to deliver power to where it is needed. Utility-scale solar projects can be developed in four to five years. Major new transmission projects take seven to ten-plus years to deploy. The two timelines do not match.
The Comparison That Matters
China faced the same physics. It did not face the same queue.
China's "East Data, West Computing" strategy, launched in 2022, designated eight national computing hubs and ten cluster nodes, concentrated in western provinces with abundant land, cheap renewable energy, and cool climates. The strategy explicitly calls for coordinated planning of computing and power infrastructure — a principle that Chinese regulators have operationalized through mechanisms like the "2+3+N" coordination framework, which replaces isolated approvals for data centers and grid upgrades with network-level integration.
The results are visible in the grid connection process. In Ningxia, State Grid has embedded industrial electricity demand into its mid- and long-term grid development plans, with rolling supply-demand assessments and advance construction of main grid frameworks and supporting transmission projects. The goal, as the company put it, is to achieve "immediate signing and connection, full connection where possible" for computing enterprises.
In Guizhou, grid companies have aligned data center layouts with grid planning to ensure computing clusters achieve "immediate arrival and connection, immediate connection and use."
In Wuxi, Jiangsu province, the "stepped commissioning" model allows computing clusters to connect to the grid in phases, with pre-laid external network infrastructure enabling each phase to "plug and play" without the enterprise building its own external transmission lines.
Inner Mongolia — the region with the largest concentration of China's western computing hubs — has made "computing-power coordination" an explicit policy priority. The region's total computing power reached 345,000 petaflops as of July 2026, with more than 85 percent of data center electricity coming from green power. The region generated more than 270 billion kilowatt-hours of renewable power in 2025.
The distinction lies in whether grid connection is structured as a queue or as a coordinated plan. In the U.S., developers submit requests and wait. In China, grid operators plan capacity around designated computing hubs, build the transmission infrastructure in advance, and connect projects as they arrive.
The Industry's Response
What has struck observers is how little resistance the Texas audit has drawn from the companies it affects most directly.
The Data Center Coalition, whose members include most of the world's largest cloud and AI infrastructure operators, framed the audit as an opportunity rather than a threat. "This review can showcase the good actors," said Dan Diorio, the coalition's executive vice president.
The cooperative tone reflects a recognition that the queue had become a liability for serious developers. If speculative projects are holding positions in line, they are delaying real projects and inflating the grid planning problem. An audit that clears the queue of ghosts benefits the companies that actually intend to build.
QTS, which operates three Texas facilities and has two more in development, welcomed the directive on similar terms, calling for clear guardrails around transparency and accountability.
What Comes Next
The Texas freeze is not a rejection of data centers. It is a rejection of the queue as a planning mechanism.
ERCOT's new Batch Zero large-load process — approved in June 2026 — was designed to organize demand into batches, study eligible projects, allocate transmission capacity based on what the grid can support, and publish a transmission plan identifying necessary upgrades. The audit paused the first batch before it could begin.
When the Public Utility Commission met on August 14, it clarified that the audit would cover roughly 300 projects under Batch Zero rather than the entire 474-gigawatt queue. But the practical effect for developers is the same: nothing new connects until the audit clears it.
The broader implication is that the U.S. grid connection process — designed for a world of individual generation projects competing for transmission capacity — is not suited to the demands of AI infrastructure. A 474-gigawatt queue that contains 90 percent data center requests, where 60 percent of projects never materialize, is not a functioning market. It is a speculative holding pattern.
Whether other states and FERC will follow Texas's approach remains to be seen.
Sources:Capacity (August 25, 2026); Institute for Energy Research (August 19, 2026); Data Center Dynamics (August 12, 2026); Carbon Direct (May 14, 2026); FERC (September 25, 2024); POWER Magazine (August 6, 2026); Utility Dive (August 5, 2026); EIA (July 2026); Lawrence Berkeley National Laboratory (December 2025); International Energy Agency (2026); China Smart Grid (September 11, 2026); National Energy Administration of Ningxia (August 31, 2026); China Energy News (August 25, 2026); Sina Finance (August 4, 2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.
Forecasts from third-party analysts can change with market conditions.
Cost and pricing examples are point-in-time estimates; actual rates vary.
Country and company comparisons rely on public reporting, not operational data.
This sector moves fast; timelines and deal terms may be updated later.
Company deals and regulatory rulings may evolve; verify current status.
AI infrastructure is changing quickly; claims can become outdated soon.
Sources
- Capacity (August 25, 2026)
- Institute for Energy Research (August 19, 2026)
- Data Center Dynamics (August 12, 2026)
- Carbon Direct (May 14, 2026)
- FERC (September 25, 2024)
- POWER Magazine (August 6, 2026)
- Utility Dive (August 5, 2026)
- EIA (July 2026)
- Lawrence Berkeley National Laboratory (December 2025)
- International Energy Agency (2026)
- China Smart Grid (September 11, 2026)
- National Energy Administration of Ningxia (August 31, 2026)
- China Energy News (August 25, 2026)
- Sina Finance (August 4, 2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.; Forecasts from third-party analysts can change with market conditions.; Cost and pricing examples are point-in-time estimates; actual rates vary.; Country and company comparisons rely on public reporting, not operational data.; This sector moves fast; timelines and deal terms may be updated later.; Company deals and regulatory rulings may evolve; verify current status.; AI infrastructure is changing quickly; claims can become outdated soon.