The DOE’s Generator Retirement Policy Needs Guardrails

The U.S. Department of Energy (DOE) has been employing emergency authority to block power plant retirements. Section 202(c) of the Federal Power Act gives the Secretary of Energy emergency authority to direct power generation and transmission during grid emergencies. Traditionally, the DOE has used this authority at the request of grid operators responding to immediate threats (e.g., hurricanes, winter storms, wildfires, summer heat) and has terminated the orders within days or weeks. When Winter Storm Fern hit in late January 2026, the agency issued 20 emergency orders to squeeze extra output from generators during the crisis, none of them lasting more than 20 days. That is the traditional use.

What is not traditional is using Section 202(c) to block power plant retirements. Starting in May 2025, the DOE began issuing 90-day orders to prevent scheduled closings—first a coal plant in Michigan and gas-oil units in Pennsylvania, then additional coal plants in Indiana, Washington, and Colorado. Neither plant owners nor the grid operators responsible for reliability in these regions requested the orders. When the 90-day orders expire, the DOE simply issues new ones.

Short-term emergencies and long-term resource planning call for different tools. Using Section 202(c) as a long-term resource-planning tool requires the planning discipline that long-run policy decisions demand. Durable policy requires analysis of costs and benefits, consideration of alternative actions, transparent criteria, and protection against market distortions. The current orders lack all four. If the DOE continues using emergency authority to defer unit retirements, then the policy needs guardrails to ensure the orders promote grid reliability rather than undermining it.

How Retirements Normally Work

Grid operators and utilities already have processes for managing power plant retirements. In regions outside of regional transmission organizations (RTOs), utilities manage retirements internally under the oversight of state regulators or cooperative boards. In regions with an RTO, when a plant owner notifies the RTO of a planned closing, the RTO conducts a detailed grid-reliability review. If the review identifies a problem, the RTO can offer the plant owner a reliability-must-run (RMR) agreement to keep the unit available until the issue is resolved. Depending on the nature of the concern, the issue may be resolved by adding new generation, transmission upgrades, or demand-side resources. These RMR arrangements are governed by rules filed with the Federal Energy Regulatory Commission (FERC). RMR rules include vetted cost-allocation provisions and incorporate market protections to prevent plants given cost-recovery guarantees from undermining market-based incentives for investment and operations.

Maryland’s Brandon Shores and Wagner plants illustrate the process working as designed. In October 2023, Talen Energy announced intentions to retire both plants. PJM-conducted reliability reviews found the retirements would create grid stability issues in the Baltimore area until planned transmission upgrades could be completed. PJM’s planned offer of an RMR agreement sparked controversy, leading to an agreement negotiated among Talen, PJM, the Maryland Public Service Commission, Maryland utilities, consumer representatives, and environmental groups that was filed with FERC and approved in May 2025. The RMR specified compensation, provided performance incentives, addressed how retention of the plants would interact with PJM’s capacity markets, and established an end date tied to completion of transmission upgrades. While the process was contentious and expensive, it was also transparent and governed by existing rules, producing an agreement that allocated costs to the customers who benefited from the reliability services.

Not every retirement raises reliability issues. The owner of the Eddystone power plant in Pennsylvania proposed retiring two gas-oil units at the plant in December 2023. The units had run less than 1 percent of the time over the four years before their planned retirement date. PJM’s reliability review found no reason to block their planned May 31, 2025 retirement; however, the day before the units were to be retired, the DOE ordered them to stay online.

RTO retirement studies have limits worth acknowledging. Retirement reviews typically assess whether a plant closure would cause transmission reliability violations—usually thermal or voltage problems on specific lines or at substations. They do not always evaluate whether the region has sufficient generation to meet peak demand. Even when an RTO review identifies a reliability concern, generators are usually not obligated to accept an RMR contract. And the decision to retire plants owned by regulated or cooperative utilities often rests with state regulators or utility boards, where regional reliability concerns may be given less weight than political or financial considerations. These limits mean that, while well developed and integrated into RTO rules, RMRs do not address every potential reliability concern.

Tail-Risk Insurance, Bought Badly

Why would the DOE override a retirement process that by all appearances (and despite the limits mentioned) appeared to be functioning well? The charitable interpretation is that the DOE sees a risk that existing regional processes miss. Each RTO plans for its own worst-case conditions, asking whether it has enough resources to ride out the stress. What this approach may not fully capture is the scenario in which multiple regions face extreme conditions simultaneously. This happened during Winter Storm Fern, when grids from the plains states to the mid-Atlantic were stressed for three days. MISO entered Energy Emergency 2 status for roughly 11 hours, and the three MISO plants subject to retirement-deferral orders ran as the winter storm blew across the country.

This interpretation frames retirement-deferral orders as tail-risk insurance. Good insurance practice requires a clear understanding of the risk being covered, a fair premium, and confidence that no cheaper alternative is available. The DOE’s orders meet none of these conditions. Instead, plant owners are obligated to comply with DOE orders, and ratepayers are stuck with the bill. For example, Consumers Energy spent $254 million operating the J.H. Campbell plant in Michigan following the DOE’s retirement deferral order, collected $119 million in market revenue, and will seek recovery of $135 million in excess costs from ratepayers in MISO’s northern and central regions.

The fact that these units ran during the late-January storm does not establish that they were necessary nor that the cost of the “insurance” was fair. Evaluating the orders’ contribution to reliability during Fern would require detailed counterfactual analysis. The director of reliability assessments at the North American Electric Reliability Corporation (NERC) called the orders “a blunt instrument,” observing that “some are helpful, some maybe not.” Note that NERC focuses primarily on grid reliability, not costs. An assessment including consideration of ratepayer costs would be less flattering.

While the DOE’s July 2025 resource adequacy report helps clarify the department’s lens on reliability, it does not make the case for any specific order—nor does it diagnose current planning failures. Without a clearer view of what’s missing, NERC, RTOs, and utilities cannot update their own reliability rules to fill whatever gap the DOE believes it has found.

The DOE has treated plant retirements during a period of rising demand as evidence that market signals are failing. That view misses what markets have done. When markets signal scarcity, investors respond. PJM’s more recent capacity auctions have motivated retention of older plants once set for retirement. Energy Information Administration (EIA) data shows that overall power plant retirements dropped in 2025 to the lowest level in almost 20 years.

Why These Plants Were Retiring

Three of the five states with power plants subject to retirement-deferral orders—Washington, Colorado, and Michigan—have ambitious carbon-free energy goals and related policies that have pushed plant retirements faster than market conditions alone would have dictated. Consumers Energy announced its retirement plan in 2021 as part of a settlement agreement with regulatory staff, the state attorney general’s office, and environmental groups. MISO approved the retirement in March 2022. The utility then spent years working with state regulators and MISO to plan the transition, acquiring a 1,200-megawatt combined-cycle gas plant in Michigan in 2023, building wind facilities, and planning a solar farm. In early May 2025, just days before the plant’s planned retirement, MISO concluded the region would have sufficient generation to meet peak load that summer. The DOE declared an emergency and blocked retirement anyway.

The Centralia coal plant in Washington presents a similar pattern. The state passed a law in 2011 to phase out coal-fired electric generation by 2025, with the date worked out in discussions between the plant’s owners, state officials, and environmental groups. The plant’s owners planned to convert the unit to natural gas, a plan currently blocked by the DOE’s order. Colorado’s Craig coal plant has a simpler story: Its owners agreed in 2016 to retire the unit by the end of 2025, with the company citing both state and federal policies and economic reasons for the closure.

State policy is a reasonable target for federal attention if it contributes to reliability problems that RTOs cannot address. RTO reliability policies (e.g., RMR) cannot override state law, and market prices simply reflect efforts to comply with it. Addressing this gap would require coordinated work between FERC, states, and NERC to clarify how federal reliability authority interacts with state retirement mandates.

But state laws did not drive all these retirements. The Pennsylvania and Indiana plants subject to DOE orders were cases in which market conditions were no longer sufficient to keep the generators in service. One unit at the Schahfer plant in Indiana had been out of service for months when the DOE ordered it back into service. As previously noted, Pennsylvania’s Eddystone units were rarely used in the years before their planned retirement date.

These orders did not fill a vacuum; rather, they overrode processes that appeared to be working. To the extent that the DOE has identified a real gap in retirement practices, the right response is to spell it out clearly and work with FERC, the states, and RTOs to close the gaps in existing reliability processes. The DOE should not seek to run a permanent workaround from Washington, D.C.

Guardrails for a Blunt Instrument

If the retirement-deferral orders address a real reliability gap, good policy practice requires they have the same guardrails as the better RTO-managed RMR agreements. Over two decades, RTOs have developed protections to address the tension between keeping uneconomic plants online for reliability purposes and preserving market incentives for investment and efficient operation. The DOE’s retirement-deferral orders lack these protections, and the consequences are predictable from the RMR experience.

• Transparent criteria. The DOE’s current orders cite a mix of general concerns about regional resource adequacy in coming years and more localized concerns. Broad statements of concern do not allow RTOs, utilities, or state regulators to act on the underlying risk. A detailed statement of the reliability concern—what type of reliability event, in which location, under what conditions, with what probability—would help others take steps to address the identified issue and let the DOE itself measure whether the gap is closing.

• Cost-benefit review. Good RMR agreements require a cost-benefit test before approval; however, the retirement-deferral orders include no such discipline. Campbell ratepayers across MISO’s Northern and Central regions had no say in the decision, and they have not seen an analysis of benefits. The same is true of every other retirement-deferral order. NERC’s “blunt instrument” characterization captures the point. Dispatch is not proof of need, and even when a plant runs, the question of whether the cost of keeping it online exceeds the reliability benefit remains unanswered.

• Consideration of alternatives. The Electric Reliability Council of Texas (ERCOT)’s RMR process requires an open and competitive solicitation of alternatives before approving a cost-of-service contract to retain a plant in service. Such alternatives include demand management, accelerated completion of generator projects under construction, use of mobile generation, and transmission upgrades. ERCOT also commits to a plan identifying how the need will be eliminated (usually by completion of a transmission project). The retirement-deferral orders include no parallel discipline. For example, the DOE has not considered whether mothballing plants now and recalling them later would more cheaply meet resource adequacy gaps projected to arise in several years, nor whether plans pursued by plant owners to replace units expected to retire would adequately meet perceived reliability needs.

• Protection against market distortions. The orders create three related distortions:

• Reduced incentive to operate efficiently. A merchant generator earning market revenues bears the cost of inefficient operation; a plant guaranteed cost recovery does not. Cost-of-service arrangements always carry this risk, which is why state regulators review monopoly utility rates and why FERC-approved RMR agreements include performance incentives and oversight provisions. (The retirement-deferral orders do not.)

• Cost shifting. Had Consumers Energy decided on its own to keep the Campbell coal plant online, the utility would have borne the full cost. The DOE’s 202(c) order spreads those costs across MISO ratepayers, with Consumers’ own customers bearing only a fraction. ISO New England’s Internal Market Monitor noted the potential for “RMR fishing,” filing a speculative retirement notice to test whether a reliability-based cost-of-service guarantee would be forthcoming and withdrawing the notice if it does not. The 202(c) orders extend this incentive across the country.

• Deferred investments in alternatives. If plant owners expect the DOE to socialize the costs of uneconomic plants, then they have less reason to develop replacement resources. Barriers to exit can become barriers to entry. Retaining an existing plant can delay redevelopment of the site itself—as with the Centralia plant, which the owner intends to replace with a natural gas plant—and the uncertain status of the retained plant discourages new development to serve the same area. Intended to protect reliability, the emergency orders risk degrading the market designs and grid rules that provide it.

Building on What Works

The administration argues that Section 202(c) retirement deferrals respond to a real reliability need. Rising demand, new generation backlogs, and the potential for correlated extreme weather across regions are genuine challenges; however, even granting a generous reading of the DOE’s rationale, the current approach fails on its own terms. It is expensive, analytically opaque, and structurally prone to the moral hazard and market distortions that FERC-approved reliability processes are designed to prevent.

The administration has identified data center growth as a primary driver of rising electricity demand—the same demand growth it cites to justify keeping aging plants online. Suppose the administration asked data centers driving demand growth within a region to cover the costs of the orders—or, better yet, allowed those customers to provide equivalent reliability services by other means.

Given the choice, data center operators would likely choose something other than paying to keep a decades-old coal plant running. They would invest in new generation, battery storage, demand response, or transmission upgrades—resources that are cheaper, more flexible, and better integrated into the grid. And if the customers whose demand growth is motivating these orders would choose differently, the current approach is probably not the most cost-effective.

If there are gaps in how the electricity industry manages reliability risk, the answer is not to bypass existing institutions but to strengthen them. Publish the analysis, set the guardrails, and let the entities closest to the action—RTOs, utilities, state regulators, and the customers driving demand growth—find the most durable solutions.

Emergency authority is a necessary tool, and using it well means knowing when not to use it. Days-long emergencies can run on adrenaline; years-long ones need guardrails.