DOE Pushes PJM to Make Data Centers Pay Their Own Power Costs

Editorial aerial image of a large data center, substation, transmission grid, power plants and nearby neighborhoods representing PJM data center electricity cost allocation.

The U.S. Department of Energy is backing Federal Energy Regulatory Commission guidance that would push PJM Interconnection to assign more of the cost of new generation and grid infrastructure to the large electricity users driving the need for it, including data centers. DOE filed its position with FERC on October 7 and publicly announced it on October 9, 2026.

Key Takeaways

  • DOE filed a formal Notice of Intervention and Statement of Position in FERC dockets ER26-3380-000 and EL26-108-000.
  • The department supports revising PJM’s Reliability Backstop Procurement so new data centers and other large loads do not shift their capacity and infrastructure costs onto existing customers.
  • DOE wants project-level tracking so PJM can see whether planned large loads actually enter service, shrink, slip or disappear.
  • The filing supports a “build, bring, or buy” model in which large loads can supply, import or contract for qualifying new capacity.
  • This is not yet a final PJM tariff. DOE is supporting FERC’s recommended reforms, and PJM has until October 29 to submit revised tariff provisions under the accelerated path described in the filing.

The primary source is DOE’s nine-page filing with FERC. The document says PJM should change its cost-allocation methodology so the costs of serving new data centers and other large electric loads are not unfairly shifted to existing ratepayers. DOE’s October 9 public announcement summarizes the same position.

Energy-sector discussion around FERC’s September 29 action shows how closely PJM’s delayed backstop procurement and large-load rules are being watched.

This Is a Cost-Allocation Fight

The policy question is not simply whether data centers should be allowed to connect. It is who pays when their electricity demand requires new power plants, capacity commitments or grid infrastructure.

DOE says FERC was right to question whether PJM’s proposed Reliability Backstop Procurement would allocate costs consistently with the traditional principle that customers should pay costs roughly connected to the burdens they create or benefits they receive. The department warns that PJM’s current method could miss some large loads already embedded in baseline demand forecasts and create what DOE calls free ridership.

BitcoinVersus.Tech previously covered FERC delaying PJM’s 6.8 GW backstop over data-center cost rules. DOE’s new filing is the next step in that same dispute: the department is now explicitly supporting FERC’s direction and asking PJM to revise its tariff quickly.

BloombergNEF’s 2026 PJM discussion explains why rapid data-center load growth is colliding with tight power supply and difficult capacity-market choices.

DOE Wants PJM to Track Each Large Project

One of the most consequential parts of the filing is DOE’s support for mandatory large-load identification and transparency. The department says PJM needs enough project-level information to know whether a proposed data center or other large customer actually enters service, is delayed, is reduced in size or is cancelled.

That matters because forecasts can drive procurement decisions years before a project begins consuming electricity. If PJM buys capacity for a projected large load that later slips or disappears, other customers can be left paying for resources procured around a demand forecast that never fully materialized. DOE argues that continuously updated project data would help PJM revise its targets, avoid double counting and attribute demand growth to the correct zones and load-serving entities.

PJM Is Already Short on Capacity

DOE’s filing says electricity use inside PJM rose more than 9% from 2020 through 2025, increasing by about 67.9 terawatt-hours. At the same time, PJM’s capacity market has struggled to secure enough supply for projected demand growth.

The document points to a 6,623 MW shortfall in the 2027/2028 delivery-year auction and a 6,831 MW shortfall in the 2028/2029 auction. DOE cites those deficits as evidence that PJM’s existing procurement framework is not keeping pace with large-load growth.

The broader demand trend is not limited to PJM. BitcoinVersus.Tech recently covered U.S. electricity demand heading toward record highs as AI and crypto data centers expand. PJM is simply one of the first regions where the financial and reliability consequences are becoming difficult to separate.

“Build, Bring, or Buy” Is the Preferred Path

DOE supports allowing large loads to meet their reliability needs by building qualifying new power supply, bringing qualifying capacity into the system from elsewhere, or contracting for qualifying new capacity. The department says PJM’s current backstop process does not yet provide a sufficiently clear path for that flexibility.

That model is already appearing in major corporate power deals. BitcoinVersus.Tech recently reported on Google and Constellation’s 3.59 GW PJM power agreement, which includes 890 MW of new nuclear capacity and a larger power-supply arrangement. Deals like that illustrate what “bring or buy” can look like in practice, although each contract has its own regulatory and market details.

What This Does Not Do Yet

DOE’s filing does not itself rewrite PJM’s tariff, set a new universal data-center rate or determine the final bill for any specific project. FERC has recommended reforms, DOE has now endorsed them, and PJM still has to submit revised tariff provisions for regulatory review.

It also does not mean every future transmission or generation cost can automatically be traced to one data center. Power systems are shared networks, and cost allocation remains a regulatory exercise that has to account for who causes a need, who benefits from an upgrade and how forecasts change over time.

What Happens Next

DOE urges PJM to use FERC’s accelerated option and submit a revised tariff by October 29, 2026. The filing specifically asks for stronger cost allocation, better large-load identification and a clearer path for large customers to build, bring or buy the capacity they need.

The next document to watch is therefore PJM’s revised tariff filing. That filing should show whether the principles in DOE’s position become concrete billing, forecasting and procurement rules—or remain broader policy goals.

Bottom Line

DOE’s October filing moves the data-center power debate from general promises toward specific market mechanics. The department is asking PJM to identify large projects more precisely, track whether they actually materialize and allocate the costs of new capacity more directly to the customers responsible for the demand.

For AI data centers, the message is increasingly clear: access to grid power may depend not only on securing a connection, but also on proving where the new electricity will come from and who will pay for the infrastructure needed to deliver it.

Editor’s Note: DOE filed its Notice of Intervention and Statement of Position on October 7, 2026, and publicly announced its support on October 9. The filing represents DOE’s position in an active FERC proceeding. It does not by itself establish a final PJM tariff or final cost allocation.

Support independent reporting: BitcoinVersus.Tech volunteers daily to help keep its research verifiably true. Bitcoin donation address: 3C9o19EH5HSiWEPyCTmEKzxhNCbo2X6TTb.

BitcoinVersus.Tech is not a financial advisor. This media platform reports on technology and financial subjects for informational purposes.

2 responses to “DOE Pushes PJM to Make Data Centers Pay Their Own Power Costs”

  1. […] also recently examined how regulators are pushing large data centers to shoulder more of the infrastructure costs their dem…. The TCDC filing shows the private-market version of that same reality: before a server rack goes […]

    Like

  2. […] regulatory and infrastructure risks facing large-load projects. BitcoinVersus.Tech recently covered federal pressure to make large data centers shoulder more of the generation and grid costs their dem…. Different grid regions use different rules, but the economic question is similar: enormous AI […]

    Like

Leave a Reply