What happens when an emerging industry decides to secede from the power grid?
In Pecos County, deep in West Texas’ Permian Basin, a data center developer claims to hold the largest power-generation air permit in the United States. According to Pacifico Energy, the Texas Commission on Environmental Quality has authorized 7.65 gigawatts of gas-fired generation on a single ranch (paired with batteries and solar) to power hyperscale data centers. In February, TCEQ commissioners gave final approval for the nation’s second-largest Clean Air permit to a different developer for Project Matador, authorizing the use of 90 Siemens combustion turbines at a site outside Amarillo. A month later, that same developer sought air permits to add another five gigawatts of generation capacity and expand its planned private grid to nearly 17 gigawatts (including proposed nuclear, solar, and battery capacity).
None of these power generation facilities will ever appear in the list of proposed power projects seeking to connect to the Texas electric grid (the “interconnection queue”) operated by the Electric Reliability Council of Texas (ERCOT). They instead will run around the clock to feed artificial intelligence workloads. And the agency that Texas empowered to permit refineries and petrochemical complexes has found itself, without ever asking for the job, serving as a principal state regulator of AI infrastructure siting.
Why and how did the meter move? As usual, the numbers tell the story. In a recent white paper on the Texas grid, the University of Houston projected that data centers and other large loads could push statewide electricity consumption up 75 to 300 percent by 2035. This sudden increase in demand would leave an annual capacity shortfall of roughly 27 to 40 GW unless the state makes major investments in new power generation and transmission infrastructure. This jump in projected demand led to expected infrastructure constraints, with unsurprising logjams as ERCOT’s large-load interconnection queue swelled above 56 GW, transmission expansion projects facing five to seven year timelines, and lead times for heavy-duty gas turbines extending into the 2030s. As a result, data center developers, whose standard business models have depended on “five nines” (99.999%) of reliable uptime, began to co-locate power generation on site long before the Texas Legislature weighed in.
But when the Legislature finally weighed in, it laid a heavy hand on the industry. Senate Bill 6, signed in June 2025, sought to protect ERCOT ratepayers from stranded infrastructure costs and to give the grid operator emergency leverage over large loads. SB 6 added new Section 37.0561 to the Texas Public Utility Regulatory Act (PURA) and ,authorized ERCOT, during grid emergencies, to direct facilities with large power loads that include substantial non-exporting on-site generation to curtail or to deploy that generation. In essence, ERCOT can reach into the facility and commandeer its power assets in emergencies. Additionally, new §§ 39.169 and 39.170 now empower ERCOT and the Public Utility Commission of Texas (PUCT) to review and approve attempts by developers to co-locate their facilities with existing grid-facing generators. The agencies can also require curtailment-capable equipment for loads interconnecting after 2025.
As a result, the price of grid membership for these data centers has begun to rise sharply: the PUCT’s draft rule in the Texas Administrative Code (16 TAC § 25.194), published in March with adoption expected this summer, would impose six-figure study fees, full contribution-in-aid-of-construction obligations, and financial security postings of $50,000 per megawatt with steep forfeiture provisions. On June 18 the PUCT approved ERCOT’s “Batch Zero” process for triaging the large-load requests already in line. SB 6 may not have spurred a flight from the Texas grid, but it certainly converted a queue-avoidance tactic into a full secession strategy. Under the new Texas regulatory regime, a wholly islanded private grid (like the Pecos County campus) now escapes ERCOT’s curtailment authority, its interconnection costs, and its multiyear studies altogether.
These power archipelagos, however, cannot escape the Clean Air Act. Large power facilities at data centers can still emit enough criteria air pollutants or hazardous air pollutants to constitute major sources under the federal and Texas Clean Air Acts, and even smaller sources may need to comply with Texas rules for standard permits or permits by rule. That trade, however, remains remarkably favorable for developers because Texas air permitting offers a ladder whose rungs can emphasize speed. The smallest sources, for example, might rely on permits by rule. Data centers’ emergency diesel arrays—sometimes dozens of engines totaling more than 150 MW at a single site—could qualify under 30 TAC § 106.511 with no individual public notice and no opportunity for a hearing. Moving to the next rung, standard permits for electric generating units under Texas Health & Safety Code § 382.05195 would offer a streamlined registration process to authorize nominally minor sources within weeks. Because the public participated (if at all) when TCEQ adopted the underlying standard permit in 2007 (with a natural-gas-engine variant added effective January 30, 2025), individual registrations draw neither notice nor contested case rights. As a result, the Environmental Integrity Project now estimates that two dozen of the roughly 130 gas plants now proposed in Texas will take advantage of this expedited route. At the top rung, even full PSD review can move at a pace that would impress practitioners in other delegated states. For example, TCEQ reportedly granted one 519 MW reciprocating-engine plant its permit three weeks after application, aided by an expedited-processing program under Texas Health & Safety Code § 382.05155 that lets applicants pay surcharges to fund overtime and contract reviewers.
Environmental and energy practitioners are typically accustomed to working with siting boards, energy facility councils, and environmental impact reviews. Texas, however, uses a different and tailored process. With no state siting statute, no state environmental policy act, and no TCEQ jurisdiction over noise or, in most respects, groundwater pumping, TCEQ’s preconstruction air permit process as a practical matter offers the primary state approval that can affect these multi-gigawatt private power campuses. Most of the siting conflicts that other states distribute across several proceedings therefore, in Texas, funnel into an air permitting process that the state never designed to carry such weight. Nor are the stakes merely local: modeling published this year in Environmental Research Letters suggests that data center and cryptocurrency growth could raise U.S. power sector CO₂ emissions 13 to 28 percent by 2030. Nearly half the gas-plant capacity now proposed in Texas would power data centers directly.
This array of legal forces creates two interesting, and important, focal points. The first fight centers on how to calculate the centers’ potential to emit. In an April 2025 letter to EPA Region 6 and TCEQ—written before SB 6 passed, and aimed at grid-connected Texas Energy Fund peakers rather than data center plants—the Environmental Integrity Project and Sierra Club protested that TCEQ had approved standard permit registrations for gas plants that actually constituted major sources under the state’s implementation plan. Their argument relied on straightforward logic: 30 TAC § 116.610(b), a SIP-approved provision, bars standard permits for any new major stationary source, and the SIP’s potential-to-emit definition counts only federally enforceable limits. Where a registration caps annual emissions just below the 250 ton-per-year thresholds of 40 C.F.R. § 51.166(b)(1) but imposes no rolling limits, no operating-hour restrictions, and no monitoring adequate to enforce them, the source’s potential to emit arguably can exceed major-source levels under the synthetic minor doctrine of EPA’s Hu Honua Title V order. The letter added that the facilities also likely constituted hazardous air pollutant major sources using EPA’s all-loads AP-42 emission factors (rather than high-load factors). With this revised approach, the same plants’ formaldehyde potential to emit would cross the 10 ton-per-year major source threshold of CAA § 112 and trigger 40 C.F.R. Part 63, Subpart YYYY. That theory maps directly onto behind-the-meter data center plants permitted through the same standard permit. As a result, the reclassification of these facilities as major sources would push them into noncompliance for construction without a PSD permit. Of course, citizen suits under 42 U.S.C. § 7604(a)(3) and re-permitting with full notice and comment could follow.
The second focal point is the contested case hearing process. Residents near the 1.2 GW Sandow Lakes plant in Lee County requested a hearing before the State Office of Administrative Hearings to challenge its air permit application. Last October, the commissioners disposed of the request - in what opponents clocked at forty-five seconds - and issued the permit. Ranchers and the Sierra Club similarly sought a hearing on the Amarillo project, and in February 2026 the commissioners denied those requests as well. They denied the request over the contrary recommendation of the agency’s own Office of Public Interest Counsel.
This abbreviated process reflects the commissioners’ increased discretion to narrow the contested case hearing process pursuant to steady intervention by the Texas Legislature over a quarter century. House Bill 801 in 1999 restructured public participation and conditioned hearings on a demanding “affected person” showing under Texas Water Code § 5.115. Senate Bill 709 in 2015 went further and designated the draft permit itself as prima facie evidence that the applicant had met all legal requirements. This small step effectively shifted the burden of producing rebuttal evidence to protestants and confined the issues referred for hearing to disputed fact questions that the affected person actually raised. It also capped hearings at 180 days. And, by design, the standard permit and permit-by-rule tiers carry no contested case right at all. The denials that drew hundreds of comments and national coverage simply reflect the system functioning as the Legislature built it.
That design does not make the denials unreviewable. The Sandow Lakes opponents moved for rehearing in November and have launched a legal challenge to the permit. The route has recent precedent: in 2023, a Travis County district court reversed TCEQ’s denial of a contested case hearing and its issuance of an NSR permit in the Max Midstream matter (a judgment since appealed), and such appeals now flow to Texas’s new Fifteenth Court of Appeals. The Texas Legislature created this new court in 2023, effective September 2024, to hear appeals involving state agencies and from the new Texas business courts (which means that an appellate bench Texas built with commercial litigation in mind will now decide most of the state’s important environmental appeals – the subject of a future blog post). Beyond state court, the avenues narrow to Title V petitions to EPA under 42 U.S.C. § 7661d(b)(2) after issuance of the operating permits, EPA oversight of the SIP-approved PSD program, and - if the potential-to-emit theory holds - citizen suits for construction without a required permit. This last route seems especially troubled after the U.S. Department of Justice’s attack on the constitutionality of citizen suit enforcement of the Clean Air Act (ironically, in a challenge to generator emissions from xAI’s massive data center in Southaven, Mississippi). Each of these alternative pathways, however, offers slower, narrower and more treacherous footing than the straightforward contested case hearing originally sought by the protestants.
This farrago has led to an ironic result. Texas has spent three decades deregulating its electricity market precisely so that generation decisions would answer to markets rather than regulators. But in 2026, efforts to protect ratepayers and reassert control over large loads have spurred developers to migrate to off-grid power models – i.e., facilities whose primary regulatory handholds are environmental permits designed for entirely different purposes. TCEQ’s permitting velocity may not survive sustained contact with EPA oversight under a future administration, potential-to-emit litigation, a new Texas appellate court, and revived assertion of hearing rights by rural communities. The hyperscalers may have opted out from one regulator’s queues and curtailments, but their shift behind the meter has left environmental lawyers at the forefront of the next great siting fight in American energy law.
