ERCOT이 역사상 가장 관대한 미국 전력망 운영자임에도 불구하고 474GW에 달하는 연계 대기열을 갑자기 동결한 이유를 분석했다.

In the spring of 2026, the Public Utility Commission of Texas (PUCT) posed a straightforward question to its data centers: how much water and power do you actually consume? Only twenty-eight companies responded, representing 92 of the state’s 582 operational and planned facilities—less than 16%.¹, ² The survey was voluntary. In July, acting on direction from the House Committee on Natural Resources, the PUCT reopened it for a second round.³ Just three weeks after that deadline passed, the governor halted the queue entirely. A voluntary disclosure request that Texas issued twice without securing answers ultimately triggered the most consequential single action any U.S. grid operator has taken against the artificial intelligence buildout to date.

On August 3, 2026, Governor Greg Abbott directed PUCT Chairman Thomas Gleeson and Electric Reliability Council of Texas (ERCOT) President and CEO Pablo Vegas via letter to conduct a "comprehensive verification and audit of all data centers advancing through ERCOT's interconnection process."⁴ The audit applies to every project in the queue "regardless of specific type or load size," extending well beyond the 75-megawatt (MW) threshold established by Senate Bill 6.⁵ Abbott explicitly warned that projects failing to comply "will be denied connection to the Texas grid."⁴

Under the directive, every developer in the queue must now provide project-by-project disclosures.⁴ This represents a materially different requirement than the PUCT’s spring survey. While the earlier questionnaire was voluntary and effectively anonymous, the new audit is mandatory and directly tied to grid access. Developers may decline to answer a voluntary survey; they cannot refuse Governor Abbott’s directive and still secure grid connection.

The directive carries no deadline. That omission is the single most critical detail in this development: the freeze is open-ended rather than date-certain. Neither the governor’s letter nor the subsequent market notice commits Texas to reopening the interconnection pipeline on any specific day.

Within hours, ERCOT formalized the pause. Issued the same day, Market Notice M-A080326-01 confirms that ERCOT will miss its August 7, 2026 deadline to classify large-load applicants into its queue triage system, known as Batch Zero.⁶ Instead, ERCOT is filing for a good-cause exception to the timelines outlined in Planning Guide Sections 5 and 9, with hearings scheduled for the PUCT’s August 20, 2026 open meeting. Per E&E News, the utility is also postponing the entire Batch Zero study process, including megawatt allocation work previously targeted for completion by April 9, 2027.⁷

Fifty-four days and eight consequential actions compressed a routine cost-allocation memo into a full freeze on 474 gigawatts (GW) of demand. Sources: Office of the Texas Governor, ERCOT, and PUCT public records; E&E News, POWER Magazine, and Texas Tribune reporting.

The queue frozen by Abbott’s directive is far from a rounding error. As of June 2026, ERCOT tracked 474.7 gigawatts (GW) of large-load interconnection requests—90.2%, or 420.8 GW, attributed to data centers—distributed across more than 1,800 projects.⁸, ⁹ This volume exceeds ERCOT’s all-time peak demand of 91,089 megawatts (MW), recorded on July 22, 2026, just twelve days before the governor’s letter.⁴ The pipeline had stood at approximately 233 GW in January 2026, doubling in just six months.⁸

The timing carries its own irony. On June 18, 2026—the same day the PUCT approved Batch Zero—the Federal Energy Regulatory Commission (FERC) issued Section 206 show-cause orders in Docket RM26-4 to all six FERC-jurisdictional grid operators. The orders directed PJM, MISO, CAISO, ISO-NE, NYISO, and SPP to justify or overhaul their large-load interconnection tariffs. ERCOT was explicitly excluded, operating entirely outside FERC’s jurisdiction.¹⁰ Those six regional transmission organizations now face contested dockets, a 60-day comment window, and years of potential appeals before any regulatory shifts materialize.¹⁰ Texas required only six weeks and a governor’s signature.

That dynamic lies beneath the surface of this development. ERCOT’s independence from federal oversight enabled Texas to assemble the world’s largest data center pipeline faster than any FERC-jurisdictional market. That same autonomy is precisely what allowed a single executive letter to halt it. Independence cuts both ways: the U.S. grid operator with the fewest procedural hurdles just demonstrated that minimal process can accelerate or freeze progress instantaneously.

Before the pause, ERCOT’s internal triage narrowed the state’s 474.7 GW data center queue to a single 65 GW tier that had already cleared completed interconnection studies—representing just 13.7% of the total. Base Load classification confers firm interconnection rights, not a commitment to construct. Sources: POWER Magazine and E&E News, citing ERCOT (August 2026); RTO Insider, citing ERCOT’s Jeff Billo at the August 3, 2026 ERCOT TAC meeting.

Strip away the headline figure, and a significantly smaller—and far more revealing—number emerges. Prior to the pause, ERCOT had already conducted internal triage on the 474.7 GW queue, sorting applicants by their proximity to Batch Zero eligibility. Reported by POWER Magazine and E&E News based on ERCOT briefings, the results illustrate a rapidly narrowing funnel:

Revisiting the Base Load tier: of the 474.7 GW requested, approximately 65 GW—13.7%—occupies the sole category that has cleared completed interconnection studies and holds firm rights without additional reliability evaluation.⁸ All remaining capacity within the 205 GW intermediate tier is either pending capacity allocation or awaiting categorical classification.

Two caveats apply to this figure, given its likelihood of being cited out of context. First, the breakdown originates from a preliminary ERCOT overview presented during CEO Pablo Vegas’s July 29 testimony before the Texas Senate. It was disseminated via trade press rather than issued as a standalone ERCOT publication.⁸ Second, E&E News described the 65 GW as the segment "most likely to be built between now and 2032." While a reasonable inference regarding the most mature tier, that characterization reflects E&E’s interpretation of a classification category, not an official ERCOT construction forecast.⁷ Base Load status confirms that a project has cleared studies and holds firm rights; it does not signify a binding commitment to construct.

Even when read conservatively, the funnel withstands independent verification. In June 2026, Enverus Intelligence Research applied a proprietary dataset of 197 projects and a twelve-signal readiness matrix, identifying only 55 projects—approximately 21.7 GW—as genuinely positioned to clear Batch Zero. An additional 62 projects, totaling roughly 37 GW, were likely slated for deferral to a future batch.¹¹ The discrepancy between 65 GW and 21.7 GW is not trivial. It encapsulates the broader genuine-versus-speculative debate, distilled into two independently derived estimates that align in order of magnitude and fall drastically short of the 474.7 GW headline.

ERCOT’s Independent Market Monitor had already articulated this reality in blunt terms. Jeff McDonald characterized the agency’s load-growth projections as "an upper bound," cautioning that actual deployments would arrive "considerably lower" amid AI business-model uncertainty, escalating equipment costs, and growing local resistance.⁷ The narrowing funnel was never hidden. It simply required a governor’s letter to compel state action.

ERCOT already possessed a mechanism engineered to address this exact issue, and it failed to deliver. Signed by Abbott on June 20, 2025, Senate Bill 6 (SB 6) mandates that every new or expanded load exceeding 75 MW demonstrate site control and post a uniform per-MW financial security deposit before ERCOT initiates interconnection studies.¹² The premise was straightforward: increase the cost of speculation so that only financially committed developers participate. ERCOT established the Batch Zero security deposit at $50,000 per MW, with a posting deadline of July 10, 2026. This operates separately from SB 6’s non-refundable flat screening fee, which ranges from $100,000 to $300,000 based on project size and applies to all developers regardless of whether they later post the larger security.⁸, ¹² These represent two distinct gates at different stages. Conflating them remains the most frequent error in coverage of Texas’s large-load framework: the screening fee purchases process entry; the $50,000/MW security secures a position in the batch study.

The mechanism did not filter sufficiently. Approximately 205 GW of the queue cleared Batch Zero’s eligibility screens—criteria theoretically calibrated to exclude undercapitalized speculative ventures through requirements for posted security, site control, and completed studies—while the mature Base Load tier contained merely 65 GW. ERCOT projected roughly 100 GW would qualify; it received more than double that amount.¹³ NRG Energy’s Bill Barnes summarized the shortfall in six words that should give pause to any market designer treating capital deposits as a proxy for project maturity: "money did not have the culling effect that we thought it was going to."⁸ Access to capital, it appears, is not synonymous with a viable project.

That dynamic explains the direct link between the failed voluntary survey and the August 3 audit. When a price-based filter underperforms, the state’s fallback is mandatory disclosure—a mechanism requiring cooperation that Texas could not secure voluntarily. Representative Brad Buckley of the House Natural Resources Committee dismissed the response rate as "pretty pathetic."