Yesterday’s piece set out the four levers utilities are using to cover the gap between when 2027 to 2029 capacity is needed and when new gas can actually show up: storage acceleration, coal life-extension, demand-side reliability, and behind-the-meter customer deals. Storage gets the trade-press coverage. Coal life-extension is the one moving most quietly through state regulatory dockets.
The shape of the lever is this. A coal unit that had a retirement date filed in a 2022 or 2023 IRP gets a revised retirement date in a 2025 or 2026 IRP amendment, pushed out anywhere from two to seven years. The utility cites updated load forecasts (almost always data-center driven), capacity market signals, gas turbine delivery constraints, or some combination. The state PUC opens a docket. After a review cycle that usually runs four to nine months, the amended retirement date is approved, sometimes with a tracker mechanism or a contingent gas replacement project attached. The unit keeps running. The capacity market accreditation gets updated. The reliability picture for that planning area gains, on paper, several hundred to a few thousand megawatts of firm dispatchable capacity that the prior IRP had counted as gone.
How many units, where
A working count of units publicly deferred, re-permitted, or moved to reliability-must-run designation since the start of 2025 lands at 18, with about 9.4 GW of nameplate capacity affected. The list is dominated by PJM and MISO units, with a smaller cluster in SPP and the Southeast vertically integrated utilities.
PJM footprint deferrals. Talen Energy’s Brandon Shores complex in Maryland is the most-publicized case, with the FERC-approved cost-of-service agreement keeping the units running past their 2025 announced retirement under a reliability-must-run construct. The total capacity covered by the Brandon Shores arrangement and the related Wagner agreement is approximately 1,800 MW. Other PJM deferrals in 2025 and 2026 include AES Indiana’s Petersburg Units 3 and 4 (no longer retiring at the originally announced 2026 date), Duke Energy Indiana’s Gibson Station (one unit deferral), and AEP’s Mitchell and Amos units in West Virginia, where the state-level political support has been unambiguous.
MISO footprint deferrals. Ameren Missouri’s Rush Island has the most complex history (a federal court-ordered closure tied to New Source Review litigation, intersecting with the utility’s own reliability case). Outside that, the MISO deferral list includes Vistra’s Baldwin Energy Complex in Illinois, where the May 2026 IRP amendment moved the retirement date for the largest unit out to 2030, and Alliant Energy’s Edgewater Station in Wisconsin, which had a 2025 retirement date pushed to 2028 in a docket the Wisconsin PSC approved in March 2026.
Southeast and SPP. Georgia Power’s Plant Bowen has been the focal example. The 2022 IRP retirement schedule had Units 1 and 2 leaving service in 2027. The 2025 IRP update, approved by the Georgia PSC in late 2025, pushed Bowen Units 1 and 2 to a 2032 retirement date and signaled that further extension was possible if the data-center load growth in the Atlanta metro continued. Duke Energy Carolinas filed a similar IRP amendment for the Belews Creek and Cliffside complexes in 2026. In SPP, Evergy’s Jeffrey Energy Center extension and Westar Wolf Creek-adjacent fossil units have moved through Kansas Corporation Commission review with bipartisan political backing.
The cumulative result, across the four ISOs and the non-RTO utilities, is on the order of 9 to 10 GW of coal capacity that had been counted on the retirement side of the IRP ledger in 2022 to 2023 and is now back on the reliable-dispatchable side for some portion of 2027 to 2029. That is enough to materially shift the capacity-margin picture in several planning areas.
Why the PUCs are mostly approving
State public service commissions are not, in general, ideologically aligned. The pattern of approvals across the deferral list is too consistent to attribute to politics alone. Three structural reasons are doing most of the work.
First, the load-growth case is now empirically harder to dispute. Five years ago, a utility load forecast that showed 3 to 5 percent annual growth would have been challenged by intervenors as inflated. In 2026, the hyperscaler interconnection request volumes, the announced data-center campuses, and the actual interconnection queue at PJM and MISO make those forecasts defensible. PUCs are increasingly persuaded that “we need this capacity” is a serious claim, not a rate-base preservation play.
Second, the alternative is increasingly visible and increasingly difficult. The gas turbine delivery story (covered Tuesday) is now common knowledge in the regulatory community. The fact that an IRP cannot produce a 2028 commercial-operation gas plant on demand is something PUCs read in the same trade press utilities do. Saying “retire this coal unit on the original schedule and we will replace it with new gas” is no longer a fully credible IRP narrative for the 2027 to 2029 window.
Third, the FERC and NERC posture has shifted in a way that gives state regulators top cover. NERC’s 2024 and 2025 reliability assessments specifically flagged the elevated risk in the load-growth scenarios. FERC Order 1920’s emphasis on long-term transmission planning, while not directly about coal retirements, has added to a regulatory tone that prioritizes reliable capacity over schedule discipline on retirements. The reliability-must-run construct, used heavily at Brandon Shores, gives FERC a federal mechanism to bless what a state PUC might be reluctant to authorize on its own.
The risks that are not being priced into the IRPs
The lever has limits, and several of them are not being fully reflected in the amended IRPs that are getting approved.
Fuel and reagent supply is the first. US coal production has been declining for a decade. The mines and rail logistics that supplied a 350 GW coal fleet are not the same as what is left to supply a 180 GW fleet. Several of the deferred units rely on Powder River Basin coal that comes from a specific set of mines with a specific set of rail commitments. Extending operations another five years for a fleet that the supply chain had begun to wind down for produces a price-volatility tail that the IRP amendments have generally not modeled aggressively. The 2022 winter storm episodes, where some plants ran low on coal stockpiles before they could be restocked, are a precedent that has not been forgotten by reliability planners.
Environmental compliance is the second. The 2024 EPA rule package targeting greenhouse gas emissions, mercury and air toxics, and effluent limitations applied to coal units with retirement dates after specific cutoff years. Several deferred units that originally claimed exemptions on the basis of near-term retirement now fall on the compliance side of the rule. The capital expenditure required to meet those obligations, in some cases hundreds of millions of dollars per unit, is now in the rate base. The litigation status of the rules remains active, but utilities are increasingly making the capex commitments on the assumption that some version of the rule survives.
Asset condition is the third. Coal units operating past their originally planned retirement dates are running with deferred capital maintenance schedules. The boiler, the turbine, the cooling system, and the environmental controls are all aging assets. Reliability availability factors for late-life coal units have trended downward, and the failure-mode tail (unplanned outages during peak periods) is precisely the risk profile that capacity markets are supposed to price in. Whether the accreditation methodologies are capturing that risk accurately is contested.
Rate impact is the fourth, and least technical, but politically most consequential. The capital expenditure to extend a coal unit, the higher fuel costs of running an aging fleet at higher capacity factors, and the eventual decommissioning costs all flow through to customer rates. The data center load that is driving the load growth does not necessarily pay rates in proportion to the cost it imposes. State PUCs that are approving the extensions are increasingly fielding questions from residential ratepayer advocates about cost allocation. The colocation order docket at PJM (covered last week) is one piece of that larger argument.
What to watch through the rest of 2026
Several state-level decisions in the next six months will calibrate how far the lever can be pulled.
The Indiana Utility Regulatory Commission has a pending docket on Duke Indiana’s amended Gibson Station retirement schedule, with a decision expected in the third quarter. The order will be a useful read on how a relatively load-growth-pressured Midwestern state is weighing the trade-offs.
The North Carolina Utilities Commission is reviewing Duke Energy Carolinas’ amended IRP, which includes Belews Creek and Cliffside extensions. Stakeholder testimony has been more contested than in Georgia, and the order will indicate whether the Southeastern utility playbook generalizes.
The Maryland Public Service Commission and FERC will both have to weigh in on what happens after the current Brandon Shores reliability-must-run construct expires. The transmission upgrades the agreement was meant to bridge to are progressing but not on the original schedule.
The Federal Energy Regulatory Commission’s posture on additional reliability-must-run constructs is the largest single variable. The Brandon Shores precedent created a template. Whether FERC chooses to apply the template broadly, narrowly, or modify it under a new administration’s direction will shape how far the second lever can extend.
How this connects to the other three levers
Coal life-extension does not stand alone. Each of the four levers from yesterday’s piece interacts with the others.
Storage acceleration is the most direct substitute. A coal unit deferred from 2027 to 2030 buys the planning area three years to build out storage capacity that can carry more of the reliability burden. The IRP amendments are increasingly explicit about this sequencing.
Demand-side reliability (the third lever, covered next) buys margin on the load side. A planning area that can shed 800 MW of flexible load during a scarcity event can defer the retirement decision less aggressively.
Behind-the-meter customer deals (the fourth lever, covered after) take the load off the bulk grid entirely. The data center campuses negotiating direct interconnection and gas generation arrangements with utilities are precisely the load that would otherwise drive coal life-extension at the system level.
The four levers are being pulled in concert. Which one carries the most weight depends on the planning area, the utility, the regulatory environment, and the specific year. For 2027 and 2028 in PJM and the Southeast, coal life-extension is doing more work than the public-facing reliability narrative acknowledges. Whether that remains true through 2029 depends largely on what storage and demand-response can deliver in the same window.
The next piece in this series will take up the third lever, demand-side reliability, and the surprisingly large numbers showing up in the latest demand-response auction results.