VALLEY FORGE, Pa. — PJM has updated its thinking on the design of its reliability backstop procurement to meet rising data center load, gravitating toward a model in which the RTO would determine the amount of capacity to be purchased and act as the administrator and counterparty to the resulting agreements.
Rebecca Carroll, executive director of market design and economics, repeatedly stressed during a workshop March 4 that PJM does not have a proposal yet and will be working on its final design through at least April 10. The RTO aims to file a proposal with FERC by late May.
PJM is considering allowing data centers that procure capacity through the procurement to avoid being enrolled in its proposed Connect and Manage system, which would require them to curtail ahead of demand response resources during strained system conditions. While the amount of capacity purchased in the backstop would be determined by PJM, Carroll said the buyers may be able to submit their own preferred amount to purchase.
The procurement is intended to be a one-time measure that awards 15-year capacity commitments, possibly starting in 2030.
Core questions Carroll said PJM’s package must answer include how the RTO should balance reliability and over-procurement risks; whether changes to credit and collateral rules would be required to account for the greater risks associated with 15-year commitments; and when resources would need to be capable of coming online to participate in the backstop — with 2029 or 2030 being possible requirements.
A model in which PJM is the counterparty could pose substantial risks if either the data center or generation default, which could force the RTO to pick up the remainder of the multiyear commitment or to suddenly procure capacity for the data center. PJM presented examples of how securitization could be used to shift the risk to investors in a model similar to the bonds issued in the wake of February 2021’s Winter Storm Uri.
PJM Chief Risk Officer Carl Coscia said such a high collateral requirement would likely make any project unviable; however, the threshold should be high enough to prevent backstop participants from walking away from their commitments. The RTO’s risk provisions were designed around a three-year advance capacity auction awarding one-year commitments and are not well positioned to account for the uncertainty with 15-year obligations, he said.
Gwen Kelly, PJM senior director of credit risk and collateral management, said if the current credit policy was applied to a 15-year, 1-GW unforced capacity commitment at $400/MW-day, there would be a $662 million pre-auction credit requirement, $224 million of which would be returnable. Coscia said this accounts for deficiency charges over the 15 years.
PJM CFO Lisa Drauschak repeated that staff still do not have a proposal, and the presentation only illustrates possible pathways and outcomes.
Several stakeholders have presented their own perspectives and proposals during several workshop meetings held over the past month. (See PJM Stakeholders Begin Discussions on Reliability Backstop Design.) The workshops will be on hiatus over the next month until PJM has a complete package to present.
Many of the same sticking points dominated the discussions: how to define the amount of capacity to be procured; whether the procurement should be one-time; which resources are eligible to offer; and whether PJM, utilities or the data centers should be the counterparties to backstop commitments.
Independent Market Monitor Proposal
The Independent Market Monitor proposed a separate backstop procurement auction awarding 15-year commitments to new resources that reach agreement to serve data centers in the same locational deliverability area (LDA).
Bowring stated the key goal of any acceptable backstop auction design is to ensure that the data centers pay for the costs they impose on the system. Bowring stated that the Monitor’s proposal is the only that does not attempt to shift costs and risks to other customers and therefore the only proposal consistent with the statement of principles from the White House Office of Energy Dominance and the PJM governors.
The backstop auction would use the basic Base Residual Auction (BRA) design, modeling capacity transfer capability and limits between LDAs and providing a single clearing price up to a maximum based on the net cost of new entry (CONE) for the reference resource. Unlike the BRA, the backstop maximum price would be based on an assumed 15-year lifespan for the reference resource to match the commitment term. The contracts would cover the full cost of energy, ancillary services and capacity.
The Monitor’s backstop would not be a one-time measure and would be run after each BRA to procure capacity for data center load, which would be excluded from the standard capacity auctions. Bowring said limiting the number of backstop auctions is desirable, artificially limiting the number to one auction would counter the goal of the White House and governors to ensure that data centers pay for the costs they impose rather than shifting those costs to others. In the absence of future backstop auctions the increased demand by additional data centers would be borne by other customers, but just delayed by a year.
Seller eligibility would be limited to new generation, with no allowance for uprates, DR or resources which canceled deactivations or did not clear in the capacity market. Bowring said this is the only way to counter the strong incentives to game the process and attempt to be designated as new. Consumers could offer varying bids into the auction, with the highest winning if there is insufficient supply offered.
Monitor Joe Bowring said the core goal is to avoid shifting risk and costs to general load by using the auction structure to directly pair resources and data centers. Any model in which PJM or utilities would be the counterparty would risk requiring other customers to pick up the commitment if the data center defaults. He referenced the ratepayer protection pledge large tech companies signed on March 4, which states they will avoid shifting the costs of their interconnection and service onto other customers. (See Trump Gets Tech Execs to Sign ‘Ratepayer Protection Pledge’.) Bowring stated that the pledge is a good model for all data centers.
Data centers larger than 5 MW would be required to participate in the backstop or be subject to curtailment similarly to PJM’s Connect and Manage proposal. PJM would work with electric distribution companies (EDCs) to identify the data center customers behind large load adjustments (LLA) the utilities submit for inclusion in PJM’s load forecast. The purpose of the comparatively cutoff for data centers is to prevent gaming by adding incremental load to a data center in order to avoid the rules.
GQS New Energy Strategies Principal Pamela Quinlan, representing the Data Center Coalition, said it would be a difficult task to tie LLAs to specific customers and seeking to allocate costs to a class of consumers based on how the electricity is used would be undue discrimination.
Bowring responded by saying the capacity shortfall PJM is facing is due to data center load growth, so the costs to mitigate those reliability risks should be assigned to those customers. Bowring said that costs must be directly assigned to data centers in order to prevent costs from being shifted to other customers.
“This is not an allocation to a class of customers. It is a direct assignment via a bilateral contract between willing participants that covers all the mutual risks in the contract. The concept of allocation means that the costs could be reallocated to other customers if a data center fails,” he said in an email to RTO Insider. “The idea that it is difficult for large corporations to reach a bilateral deal is far fetched.”
She argued the Monitor’s analysis assumes available capacity would remain the same in the absence of data center load growth, ignoring the likelihood of resources deactivating without that growth.
Quinlan said using a 15-year net CONE to set the maximum price, on the grounds that the commitment term would be 15 years, misses that resources could participate in PJM’s other markets after the commitment has expired.
Bowring responded that the basic BRA would have higher prices and would continue to attract the capacity needed to meet the needs of other customers. “As our analysis has shown, the current crisis in the capacity market is a result of data centers and not the organic growth of other customers.” It is fine that data centers participate in PJM markets at the end of the 15 year contracts.
“The proposal from the Data Center Coalition would have PJM serve as the counterparty. This means that the data center risk would be imposed on other PJM members. That is inappropriate and inconsistent with the basic goal of a backstop auction to address the costs imposed by data centers. It is not a general concern to be put off to the uncertain future but must be a core element of any backstop proposal,”he said in an email.
Data Center Coalition
Quinlan presented a set of priorities the Data Center Coalition believes should be incorporated into PJM’s design, centering around the position that the RTO should not make substantial changes to the capacity market while designing a one-time procurement structure.
The coalition recommended a backstop design in which PJM would be the counterparty and limited to participants which could be in service for the 2028/29 delivery year, with some allowance for the following year. The RTO’s design should not seek to determine resource adequacy for specific load-serving entities or use “uncertain” long-term forecasts to determine the need for capacity.
Concurrent with the procurement, the RTO should initiate a comprehensive review of the capacity market design, including improvements to load forecasting and consideration of “LSE-based frameworks,” Quinlan said.
Responding to questions on how the risk of a data center default could be managed, Quinlan said risk allocation is an important question to consider, but one that should be part of a long-term discussion. The ideal way to manage the risk associated with multiyear commitments is to ensure that the backstop is a short-term measure that buys time for more substantial market changes, she said.
Quinlan said the coalition considered ways of allocating costs that did not fall to LSEs, but there are practical questions on implementation and whether that can be accomplished in time for a May filing.
Google
Google recommended PJM adopt a backstop in which it procures capacity on behalf of load and allocates the costs across the region, leaving it to states to develop end-user rates. While the company shared several design components it prefers, it stated it does not have a complete proposal.
The company expressed support for one-time solution targeting a specific delivery year with well defined needs, leaving long-term capacity commitments as a separate issue. The backstop should focus on a fuel-neutral framework for incentivizing high-accreditation resources, with the capacity to be purchased defined by the deficiency in a particular auction — rather than targeting individual customers or a class of end-users.
Joint Stakeholders
A cohort of generation owners presented a backstop focused on meeting the shortfall expected for the 2028/29 BRA, scheduled to open in June 2026. The proposal was signed onto by Constellation Energy, Vistra, AlphaGen and Earthrise.
The one-time auction would be conducted in September and mirror the 2028/29 BRA clearing price for commitments up to 15 years. Resource offers would clear first based on the delivery year in which the project can come online, then by the length of the commitment term the offer requested. Procurement would be capped at the reliability requirement for the 2028/29 delivery year.
Seller eligibility includes new resources, uprates, DR, reactivated resources and existing resources that cleared above the maximum price in the 2028/29 auction.
Constellation’s Erik Heinle said the proposal is agnostic on how costs would be allocated, though it specifies that it would respect bilateral contracts. The risk of over-procurement and large loads not coming online would be managed by restricting procurement to load that is already accounted for in the capacity auction through capping the amount purchased at the reliability requirement for the 2028/29 delivery year.
Voltus
Voltus advocated for PJM allowing behind-the-meter capacity to participate in the backstop, arguing behind-the-meter resources have some of the quickest development times — making them well suited to a process intended to rapidly bring on new capacity.
Senior Manager of Regulatory Affairs Kimaya Abreu said PJM should be focused on procuring new capacity from resources not receiving a sufficient price signal from BRAs. That effort would be best served by taking a tech-agnostic approach which allows behind-the-meter capacity to participate. Not allowing DR, behind-the-meter storage, and other DERs’ participation would run afoul of requirements that BTM resources be treated comparably to generation, outlined by FERC in orders 719, 745, 841, and 2222.
Voltus argued including behind-the-meter resources in the backstop is consistent with proposals stakeholders made throughout the 2025 Critical Issue Fast Path process focused on meeting large load growth, as well as the statement PJM’s Board of Managers released at the conclusion of the process. (See PJM Board of Managers Selects CIFP Proposal to Address Large Load Growth.)
The company also endorsed a proposal by the Natural Resources Defense Council to define new capacity, which would allow resources that have completed the third phase of the interconnection process, or are in the surplus interconnection service process, to qualify so long as they are not already subject to the capacity must-offer requirement. For DR, resources that did not offer into the capacity market between the 2025/26 and 2027/28 auctions would be permitted, as well as those seeking to increase the amount of capacity offered.
NRDC
The NRDC’s proposal included an auction design in which capacity would be procured for a pool of buyers that would share the costs and risks, while sellers would receive 10- to 15-year commitments. If participating consumers default or do not come into service, either the capacity payments would be reduced, or the remaining load would pay more. The auction would be a permanent addition to the capacity market, conducted during each queue cycle’s final agreement phase. For Transition Cycle 2, this would be December 2026 or the following month.
Participating resources would be required to offer into BRAs during their commitment terms, with the revenues flowing to load with long-term commitments, which would also be responsible for capacity deficiency penalties. The auction would be open to large loads as well as LSEs seeking to offer long-term firm service to new customers.
The maximum procurement would be set at the amount bid into the auction, and any load that does not receive a commitment would be required to go through PJM’s proposed Connect and Manage system.
Eligibility would be limited to projects that have already cleared the interconnection queue but not yet entered service, as well as DR. The NRDC said the backstop should not be allowed to become another expedited interconnection queue following the example of the Reliability Resource Initiative, which allowed 51 projects to be inserted into TC2. Several of those projects have dropped out of the queue after running into high network upgrade costs.
