The Federal Energy Regulatory Commission handed the data center industry a split decision this week, and the fallout is already rippling through grid operators well beyond the mid-Atlantic. On October 9, 2026, FERC approved a proposal from PJM Interconnection, the grid operator serving 13 states and Washington, D.C., to build a formal registry tracking large electricity users, including the data centers driving the AI boom. In the same order, the commission rejected the more contentious half of PJM’s plan: a framework that would have let data centers connect to the grid faster in exchange for accepting power cuts ahead of ordinary households during emergencies.
Reuters was first to report the dual ruling, noting that the panel approved PJM’s registry of data centers while rejecting the plan to curtail power to them. The registry provisions take effect October 12, 2026, just three days after the commission’s order landed. For an industry racing to connect gigawatts of new load to an already strained grid, the decision draws a clear line: regulators want visibility into who is drawing power and how much, but they are not yet ready to let a regional transmission organization decide whose lights stay on during a shortage.
What FERC Actually Approved: The PJM Large-Load Registry
The approved piece of PJM’s filing is narrower than the headlines suggest, but it still marks a meaningful shift in how the largest U.S. grid operator by footprint will track its customers. The registry requires large-load facilities, a category that captures data centers alongside other heavy industrial users, to report a set of operational details to PJM: the facility’s location, its ramp schedule (how quickly it can increase or decrease power draw), its capacity supply arrangements, and other technical information the grid operator says it needs for planning purposes.
Until now, PJM has had limited visibility into exactly how many large loads are in various stages of development across its territory, how much power they will eventually draw, or how flexible their consumption patterns actually are. The registry is meant to close that gap. It does not, on its own, change how a data center gets approved to connect to the grid, and it does not give PJM new authority to shut off power to any specific site. What it does is create a standing dataset that PJM, state regulators, and utilities can use to plan transmission upgrades and generation additions years ahead of when a facility actually needs power.
Why FERC Rejected PJM’s Power Curtailment Plan
The part of PJM’s filing that drew the most attention, and the part FERC turned down, was built around a mechanism PJM called the Interim Resource Adequacy Service, or IRAS. Under IRAS, certain large loads would have been allowed to connect to the grid without first securing matching new generation capacity, as long as they agreed in advance to curtail their demand before a grid emergency hit. In practice, that meant some data centers could jump the interconnection queue faster than today’s rules allow, trading speed for a contractual promise to go dark, or at least scale back, when the system got tight.
FERC found that key pieces of the IRAS framework reached too far into territory that belongs to state regulators rather than federal wholesale-market overseers. Utility Dive’s reporting on the order describes the commission’s reasoning as resting on the idea that the rejected provisions effectively regulated retail rates and retail service arrangements, an area where FERC’s jurisdiction is limited and states hold primary authority. The commission also flagged concerns that the curtailment-first approach could discriminate against large loads relative to other types of customers, treating data centers as a separate, lesser-priority class of electricity user without a clear, consistently applied standard for doing so.
Inside PJM’s Original Curtailment Proposal
PJM’s curtailment plan did not appear out of nowhere. It followed months of internal debate within the grid operator over how to handle a surge of interconnection requests from hyperscale data center developers, many of them tied to AI training and inference capacity. Coverage of the original proposal described a plan that would have prioritized cutting or shutting off power to certain data centers and other large loads ahead of residential and commercial customers whenever the grid faced a shortage. The logic was straightforward from PJM’s perspective: if new large loads are going to connect faster than the traditional queue allows, someone has to absorb the risk when supply runs short, and PJM argued that risk should fall on the new, flexible-by-agreement loads rather than on existing ratepayers.
Legal analysis published around the time of the filing, including a review from the law firm Babst Calland, framed the proposal as PJM’s attempt to thread a narrow needle: the firm’s breakdown of the filing noted that PJM was trying to accommodate explosive large-load growth without forcing a full buildout of new generation capacity before each project could connect. FERC’s rejection suggests the commission was not convinced that tradeoff could be structured in a way that stayed within federal jurisdiction and treated all large loads fairly.
The Risk-Shifting Logic Behind IRAS
Strip away the acronyms and PJM’s pitch was a risk trade: developers get speed, PJM gets a contractual escape valve for emergencies. FERC’s rejection means that trade, at least in the form PJM proposed, is off the table for now, and any revised version will need to find a different way to allocate that risk.
The Jurisdiction Problem: Retail Rates vs. Wholesale Markets
The jurisdictional split at the center of this ruling is not a new tension in U.S. electricity regulation, but data centers have pushed it to the surface faster than almost any other category of customer in recent memory. FERC regulates wholesale electricity markets and interstate transmission. States and their public utility commissions regulate retail rates, meaning what an individual customer, residential or commercial, actually pays and under what terms. When PJM’s curtailment framework effectively asked large loads to accept different service terms than other customers in exchange for faster interconnection, FERC read that as edging into retail-rate territory that the commission does not have clear authority to set.
That distinction matters enormously for how the next version of any curtailment plan gets built. A federally approved framework that applies uniformly across PJM’s wholesale market is one thing. A plan that effectively creates a new retail service class, data centers that pay less or connect faster in exchange for accepting outages, is something state regulators across PJM’s 13-state footprint would likely need to sign off on individually, state by state, utility by utility. That is a far slower, far more fragmented path than PJM’s original filing envisioned.
PJM Can’t Unilaterally Cut Power, Even With FERC’s Blessing
One detail getting lost in the headline framing deserves more attention: even if FERC had approved PJM’s curtailment plan outright, PJM itself does not have the operational ability to independently cut power to a specific data center. PJM is a transmission and wholesale-market coordinator, not a utility that owns distribution wires running into any given facility. Actually curtailing a named site would require active cooperation from the local utility delivering that power, and in most cases, sign-off from state governments overseeing that utility’s retail service obligations. The registry FERC just approved does not change that basic structural fact. It simply gives PJM, and by extension the utilities and states it works with, better data to plan around.
The Scale of PJM’s Data Center Pipeline
The urgency behind this fight comes down to raw numbers. PJM has told FERC in its own filings that data centers could eventually account for a substantial share of total demand across its footprint, and the grid operator’s planning documents point to large-load growth measured in tens of gigawatts over the next decade. The table below summarizes the figures that have surfaced through PJM’s regulatory filings and trade-press coverage of the docket.
| Metric | Figure | Source Context |
|---|---|---|
| PJM footprint | 13 states plus Washington, D.C. | PJM’s wholesale market and transmission planning territory |
| Large-load definition threshold | 50 MW+ peak demand | Definition approved in the October 9 FERC order |
| Projected data center share of peak demand | Up to roughly 30% by 2036 | Figure PJM has cited in its own FERC filings on long-term load growth |
| Projected data center load by 2036 | Roughly 67 GW | PJM long-term planning projection cited in filings tied to the large-load docket |
| Typical hyperscale single-site request | 500 MW to 1 GW | Range reported by trade press covering interconnection requests in PJM’s queue |
| Registry effective date | October 12, 2026 | Effective date set in the FERC order approving the registry |
Those figures explain why PJM felt pressure to move fast on some kind of framework for new large loads, and why FERC’s rejection of the curtailment half of the plan leaves a real gap. The grid operator’s own numbers suggest data centers are not a marginal category of customer anymore; in PJM’s telling, they are on track to become one of the largest single drivers of peak demand growth across the entire 13-state region.
A Pattern of FERC Rulings on Data Centers Through 2025 and 2026
This week’s order is not an isolated event. It is the latest step in a string of FERC actions over roughly the past year addressing how large loads, and data centers specifically, connect to the PJM grid. Late in 2025, FERC issued guidance addressing co-located generation, essentially power plants built alongside data centers to supply them directly, within PJM’s territory. Through the first half of 2026, PJM filed additional proposals and compliance documents addressing how those co-location arrangements and large-load interconnections should work, and FERC issued partial approvals and rejections along the way before this month’s ruling on the registry and the IRAS curtailment framework.
| Timeframe | Action | Outcome |
|---|---|---|
| Late 2025 | FERC addresses co-located generation rules within PJM | Guidance issued for pairing on-site generation with large loads |
| Early-to-mid 2026 | PJM files large-load and co-location compliance proposals | Partial FERC approvals and rejections on specific provisions |
| October 9, 2026 | FERC rules on PJM’s registry and IRAS curtailment framework | Registry approved; curtailment plan rejected |
| October 12, 2026 | Registry provisions take effect | Large loads must begin reporting required operational data |
Read together, the pattern shows a commission willing to give PJM tools for visibility and planning, the registry, co-location guidance, but far more cautious about handing PJM or any RTO the authority to reorder who gets power first during a shortage. That caution has been consistent across multiple orders, not a one-off reaction to this specific filing.
Market Impact: What This Means for Data Center Developers and Utilities
For data center developers with projects already in PJM’s interconnection queue, the immediate practical effect is paperwork, not a change in connection timelines. Large loads now need to supply PJM with the location, ramp schedule, and capacity-supply details the registry requires. That is a compliance burden, but not a blocking one. The bigger consequence is strategic uncertainty. Developers who were counting on IRAS-style fast-track connections in exchange for accepting curtailment risk now have to plan around the traditional interconnection process, which generally requires new generation capacity to be secured before a large load can connect, a slower and more capital-intensive path.
Utilities and state regulators across PJM’s footprint are arguably the bigger winners from this order in the near term. By keeping curtailment authority out of PJM’s hands and within state jurisdiction, FERC preserved the ability of state utility commissions, the bodies that answer most directly to residential ratepayers, to decide how aggressively, if at all, data centers should be asked to shoulder outage risk in their territory. That sets up a patchwork likely to emerge over the next year: some states may move quickly to approve their own large-load curtailment tariffs, while others may resist anything that looks like preferential treatment for data centers, creating uneven rules for developers operating across PJM’s 13-state region.
How PJM’s Approach Compares With Other Grid Operators
PJM is far from the only regional grid operator wrestling with how to absorb a wave of data center interconnection requests. Other RTOs and independent system operators across the country, including those covering Texas, the Midwest, and California, have been working through their own versions of large-load interconnection reform, generally weighing the same tradeoff PJM tried to solve: how to let big, flexible loads connect faster without shifting reliability risk onto everyone else on the grid. None of those other operators has yet secured FERC approval for a curtailment-first framework as sweeping as what PJM proposed, which makes this week’s rejection a useful data point for the rest of the industry. It signals that federal regulators are, for now, more comfortable approving data-visibility tools like registries than they are approving mechanisms that let a grid operator unilaterally decide which customers lose power first.
That makes PJM something of a bellwether. Because PJM’s data center growth has outpaced most other U.S. grid territories, driven partly by its proximity to major fiber routes and existing data center clusters in Virginia and the surrounding region, whatever framework eventually clears FERC review in PJM’s territory is likely to become a template other RTOs study closely when they draft their own large-load rules.
The Data Behind the Debate
Industry data helps explain why this fight is happening now rather than five years ago. Lawrence Berkeley National Laboratory’s data center energy research projects that data centers could account for roughly 11.8% of total U.S. electricity consumption by 2030, with a plausible range between 9.5% and 15.3% depending on how fast AI-driven demand grows. Deloitte’s technology, media, and telecommunications forecasting puts global data center electricity consumption at around 536 terawatt-hours in 2025, close to 2% of global electricity use. PJM Interconnection’s own planning documents go further for its specific territory, stating that data centers already represent more than 7% of energy use across the region PJM serves, and that more than 90% of data center load in the PJM region runs continuously, around the clock, rather than fluctuating the way most other large industrial loads do.
That last figure, the near-constant nature of data center demand, is central to why curtailment is such a contentious idea. A factory that can ramp production up and down has flexibility to offer a grid operator. A data center running AI training workloads around the clock has much less room to simply switch off without disrupting the service it supports, which is exactly the tension FERC’s ruling leaves unresolved.
Why Round-the-Clock Load Changes the Calculus
That near-constant draw is exactly what made PJM’s curtailment pitch so attractive on paper and so disruptive in practice. Asking a factory to idle for a few hours is routine. Asking an AI training cluster to go dark mid-run carries a much higher operational cost, which is part of why the proposal drew pushback from data center operators even before FERC weighed in.
What Happens Next: PJM’s Likely Compliance Path
PJM is not left without options after this ruling. The grid operator can refile a narrower curtailment framework that avoids the retail-rate and discrimination issues FERC flagged, likely by working more closely with state regulators from the outset rather than proposing a PJM-administered mechanism that touches retail service terms. PJM could also pursue curtailment arrangements on a bilateral basis, negotiated directly between individual data center developers and their local utilities, with state commission approval, rather than as a blanket PJM-wide tariff. That approach would be slower and more fragmented, but it would sidestep the jurisdictional objections baked into this month’s rejection.
In the meantime, the registry itself starts generating data immediately. Over the next several quarters, PJM, state regulators, and the utilities operating within PJM’s footprint will have, for the first time, a consolidated picture of exactly how much large-load demand is coming, where, and how flexible that demand actually is. That dataset will almost certainly shape whatever curtailment framework PJM eventually brings back to FERC.
Predictions: Where This Regulatory Fight Goes From Here
Based on the pattern of FERC’s rulings over the past year and the structural issues raised in this order, several outcomes look likely over the next 12 to 18 months. First, PJM will almost certainly refile some version of a large-load flexibility framework, but routed through state regulatory processes rather than as a PJM-administered curtailment tariff, to address FERC’s jurisdictional concerns directly. Second, expect other RTOs, particularly those managing fast-growing data center clusters outside PJM’s territory, to watch this ruling closely and favor registry-style data collection over curtailment-first proposals in their own near-term filings. Third, state utility commissions across PJM’s 13-state footprint will likely diverge sharply in how they handle large-load service terms, with some states moving toward negotiated curtailment agreements with hyperscalers and others resisting anything that looks like special treatment for data centers.
Fourth, the registry data itself is likely to become a flashpoint in future rate cases and transmission planning disputes, since it will give consumer advocates, utilities, and data center developers alike a common, PJM-verified dataset to argue over. Fifth, and perhaps most significantly for the AI industry, expect more hyperscalers to accelerate behind-the-meter and co-located generation projects, building their own power supply alongside data centers rather than waiting on the traditional interconnection queue, since FERC’s prior guidance on co-location remains a viable path that does not run into the same jurisdictional objections that sank PJM’s curtailment plan.
Related
- US Grids Get 60 Days to Fix AI Data Center Power [2026]
- Senate GRID Act Targets Data Center Energy Crisis
- The $1.4 Trillion Grid Overhaul: How Utilities Are Reshaping America’s Power System
- The AI Data Center Power Crisis: How Big Tech’s GW Appetite Is Reshaping the Grid
- Half of U.S. AI Data Centers Delayed or Canceled: Inside the 7 GW Capacity Crisis
Frequently Asked Questions
What did FERC actually approve on October 9, 2026?
FERC approved PJM Interconnection’s proposal to create a registry tracking large electricity users, including data centers, that requires those facilities to report their location, ramp schedule, capacity supply arrangements, and other operational details. The registry provisions took effect October 12, 2026.
What did FERC reject?
FERC rejected PJM’s proposed Interim Resource Adequacy Service, or IRAS, framework, which would have let certain large loads connect to the grid without first matching new generation capacity, in exchange for agreeing to curtail demand before grid emergencies.
Why did FERC reject PJM’s curtailment plan?
FERC found that key provisions appeared to intrude on state jurisdiction over retail rates and could discriminate against large loads, both issues that sit outside or in tension with the commission’s wholesale-market authority.
Can PJM now cut power to data centers during an emergency?
No. PJM cannot independently curtail power to any individual site. Doing so would require cooperation from the local utility delivering power to that facility and, in most cases, approval from the relevant state government or utility commission.
Which states does PJM’s registry cover?
PJM coordinates the wholesale electricity market and transmission planning across a 13-state footprint plus Washington, D.C., making it the largest grid operator territory affected by this ruling.
What counts as a “large load” under the new registry?
FERC approved PJM’s definition of a large load as an end-use customer with peak demand of at least 50 megawatts, a threshold that captures most hyperscale data center campuses.
Will this ruling slow down data center construction in PJM territory?
Not directly. The registry adds a reporting requirement but does not change interconnection timelines on its own. The bigger effect is removing the faster, curtailment-based connection path some developers were hoping PJM would offer, pushing most projects back toward the traditional, capacity-matched interconnection process.
What happens next in this regulatory process?
PJM is expected to refile some version of a large-load flexibility framework that addresses FERC’s jurisdictional concerns, likely by working more directly with state regulators rather than proposing a PJM-administered curtailment tariff.