The single most important grid-policy document of 2026 will not be a piece of legislation, a White House executive order, or a court ruling. It will be a PJM tariff filing due at the Federal Energy Regulatory Commission on February 16, 2026, drafted in response to a December 18, 2025 FERC order, that defines for the first time how a large customer load can sit physically beside a generator and contract directly for its output. That document will determine which of the roughly 50 gigawatts of hyperscaler load currently in the PJM development queue gets built on the announced timelines, which gets restructured, and which moves to other interconnections entirely.
The order matters because it ends the ad-hoc phase of co-location policy. For the last two years FERC has handled co-location case by case, with each Amended Interconnection Service Agreement, each Wholesale Distribution Service tariff filing, and each PJM Operating Agreement waiver evaluated on its own facts. That worked when the deals were exotic. It stopped working once the deals became the dominant new-build pattern.
What the December 2025 order actually did
FERC’s December 18, 2025 order directed PJM to file revised tariff language defining two new transmission services available to a co-located large load. Both services were left as proposals in the order, with PJM tasked to develop the detailed terms.
Firm Contract Demand transmission service. A guaranteed firm transmission product, sized to a contract demand level negotiated between the co-located load and the generator, that permits the load to receive a defined quantity of grid-supplied energy when the host generator is unavailable. The host generator continues to participate in PJM markets and reliability obligations. The load is charged transmission and ancillary services at the firm rate for the contracted quantity. This is the product that allows a behind-the-meter style arrangement to maintain genuine grid-backstop optionality, at a defined and ratepayer-protective price.
Non-Firm Contract Demand transmission service. An interruptible product that gives the co-located load access to grid energy only when system conditions permit, at a lower transmission and ancillary cost. The load is exposed to interruption during reliability events, but pays less for transmission service when it does draw from the broader grid. This is the product that allows a hyperscaler to take a “cheaper but exposed” position when its workload genuinely tolerates interruption, which for AI training workloads is more often than the public framing has assumed.
In addition to the two transmission products, FERC required PJM to file a separate informational report by January 19, 2026 addressing reliability concerns associated with co-location arrangements, including the impact on PJM’s capacity, ancillary services, and reserve markets when a generator’s output is partially or fully committed to a co-located load. PJM has confirmed publicly that the informational report was filed on schedule, with the substantive tariff brief still being prepared for the February 16 deadline.
Why this exists at all
The order is the policy resolution to a question that the Susquehanna case forced on FERC in 2024. In November 2024, FERC voted 2 to 1 to reject the amended ISA between Talen Energy’s Susquehanna Nuclear LLC, PPL Electric Utilities, and PJM that would have raised the co-located Amazon Web Services data center load from 300 megawatts to 480 megawatts. The majority, Commissioners Christie and See, concluded that PJM had not justified the nonstandard provisions and that the arrangement risked cost-shift to other PJM ratepayers and reliability degradation. Chairman Phillips dissented. Talen filed a rehearing request in December 2024.
The rehearing did not produce a reversal. It produced something more durable. FERC opened a Section 206 show-cause proceeding in February 2025 directing PJM and the PJM Transmission Owners to demonstrate that the existing PJM tariff was still just and reasonable in light of co-location requests, and held a technical conference in late 2024 that surfaced the structural questions the case-by-case approach could not resolve. The December 2025 order is the outcome of that proceeding.
The substantive question the order had to answer was specific. When a hyperscaler signs a contract that takes a large block of a nuclear plant’s output directly, what transmission service does the load take, and at what price? If the load takes full Network Integration Transmission Service (NITS) at the standard rate, the economics of the arrangement break, because the entire point of the co-located structure is to avoid the cost stack and the queue delay of a full network customer. If the load takes no transmission service at all, every other ratepayer in the PJM footprint is implicitly subsidizing its access to grid backstop during generator outages. Neither extreme is acceptable to FERC. The two new products are the middle.
What this changes for the deal stack
The order is not a green light for behind-the-meter co-location. It is a framework that makes properly-priced co-location possible. The actual permissibility of any specific arrangement still depends on the PJM tariff language that comes out of the February 16 filing and the subsequent FERC review.
That said, the directional implications are clear.
The nuclear PPA structure now has a regulated template. Through 2024 and into early 2025, the major nuclear restart and uprate deals (Constellation’s Three Mile Island reactivation contract with Microsoft, Talen’s Susquehanna arrangement with Amazon, Vistra’s Comanche Peak negotiations) were being structured against a moving target. Each deal had to manage the risk that its specific contract structure would be rejected at FERC on cost-shift or reliability grounds. The December 2025 order removes that overhang for any deal that fits within the Firm or Non-Firm Contract Demand template. Deals structured outside the template will face higher regulatory friction, deals inside will move faster. This matters for nuclear IPP valuations more than for hyperscaler economics, because the nuclear plant operator is the party that needs the regulatory certainty to commit to capital improvements and license renewals.
The “speed to power” advantage of co-location narrows but does not disappear. A behind-the-meter co-located data center can begin drawing load on the timeline of the generator’s commercial operation, bypassing the multi-year network queue. That advantage persists under the new framework, because the Firm and Non-Firm Contract Demand products do not put the load through a standard network interconnection process. What changes is the cost stack. Co-located load now pays for the transmission and ancillary backstop it actually uses, which will be material though smaller than full NITS. The economic case for co-location remains positive against full network connection for queue-bound projects, but the margin tightens.
Hyperscaler capital allocation shifts toward interruption-tolerant workloads. The Non-Firm Contract Demand product creates a real arbitrage for hyperscalers willing to design workloads around interruptibility. AI inference serving needs to be firm, but AI model training can often tolerate hour-scale or even day-scale interruption with minimal economic loss. Hyperscalers that build their internal demand-shifting infrastructure (workload migration, training checkpoint and restart, geographic redundancy across non-co-located sites) can capture the non-firm transmission discount. Hyperscalers that do not, cannot. This is a real software and operations capability difference between the major buyers, and the order makes it visible.
Greenfield gas in PJM gets a partial reprieve. A surprising secondary effect of the order is that it makes new gas peaker buildout slightly more attractive at the margin. The framework explicitly permits a generator-load pairing that is not nuclear, which means a combined-cycle or simple-cycle gas plant developer can now pursue the same structural arrangement that has been dominated by nuclear announcements. Gas plant development cycles are shorter than nuclear license renewals or new builds. For hyperscalers with 2027 to 2029 power requirements that the nuclear stack cannot meet on time, gas co-location becomes a viable bridge. The climate implications of that pattern are a separate question.
Non-PJM ISOs will move next. PJM is the only RTO with the immediate regulatory pressure to act, because it has the bulk of the hyperscaler load. ERCOT operates outside FERC jurisdiction for most matters, but the Public Utility Commission of Texas is already developing parallel rules for large-load interconnection that draw on similar concepts. MISO, SPP, and the New York and New England ISOs have all signaled they expect to develop co-location frameworks in 2026 and 2027, with the PJM tariff language likely serving as the reference model. Regional variation will persist, but the structural categories of Firm and Non-Firm Contract Demand transmission service are likely to converge across ISOs.
What to watch in the February filing
The PJM brief due February 16, 2026 will be evaluated on three operational questions that the order left open.
First, the cost allocation methodology for the Firm Contract Demand product. The order requires that the co-located load pay a “just and reasonable” share of transmission and ancillary costs. The specific allocation formula will determine how attractive the product is. If the formula treats the co-located load as substantially equivalent to a network customer for cost-causation purposes, the economic margin narrows considerably. If the formula recognizes the load’s reduced use of the bulk transmission system, the margin holds.
Second, the curtailment priority and interruption rules for the Non-Firm product. The order did not specify where in PJM’s reserve and curtailment hierarchy non-firm co-located load sits. If it is curtailed first, alongside other interruptible service, the product is genuinely useful to hyperscalers willing to design around it. If it is curtailed after firm load but before generator-side reserves, the product is less differentiated and less valuable.
Third, the reliability-event treatment when the host generator trips. The informational report due January 19 will set the analytical foundation, but the tariff language must specify the operational rules. Under a Firm Contract Demand arrangement, the co-located load expects to receive grid energy when the host generator is unavailable. Under a Non-Firm arrangement, the load expects to be cut. The transition between those states during a generator trip, particularly for a multi-unit nuclear plant where one reactor remains online while another goes offline, is operationally tricky and not yet well-defined in the public record.
Risks to the read
The framework could be rendered moot by a state-level intervention. State public utility commissions retain jurisdiction over retail rates and over generation siting decisions, and several PUCs (Virginia, Pennsylvania, Ohio) have signaled discomfort with the cost-shift and reliability risks of large co-located arrangements. A determined state PUC can effectively block a co-located deal in its jurisdiction even when the FERC-jurisdictional tariff would permit it.
The 2026 capacity auction results in PJM could reorder the economics. The May 2025 PJM Base Residual Auction set a record clearing price near $269 per megawatt-day, reflecting tight reserve margins driven in part by data center load. The 2026 auction has not yet cleared. A sufficiently high or sufficiently low capacity price will shift the relative economics of behind-the-meter versus network-supplied data center power and change which deal structures pencil.
Congressional action on AI infrastructure could overtake the regulatory framework. Bills circulating in 2025 and 2026 propose federal preemption of certain large-load interconnection decisions in the name of national security and AI competitiveness. None has advanced to floor consideration. If one does, the careful FERC-PJM tariff framework would be partially superseded.
The frame
FERC has done what it can do. It has created a structured tariff framework that lets co-located data center load exist legally, at a regulated price, with defined reliability obligations on both sides. That is a substantial accomplishment compared to the case-by-case chaos of 2024. It is not a finished product. The February 16 PJM filing, the subsequent FERC review, and the state-level responses across 2026 will determine whether the framework actually unlocks the hyperscaler buildout or merely formalizes the friction that has already shifted some of that buildout out of PJM.
For investors, developers, and utility planners, the calendar is now legible. Watch the January 19 reliability report, watch the February 16 tariff filing, watch FERC’s response in the second quarter, and watch the state PUC reactions through the summer. By the third quarter of 2026, the deal structure for the next decade of US data center power will be set. The decisions made in those filings are larger than any single announced project.
Sources
- Federal Energy Regulatory Commission, Order on Show Cause Proceeding, Docket EL25-49-000, issued December 18, 2025, ferc.gov/news-events/news/fact-sheet-ferc-directs-nations-largest-grid-operator-create-new-rules-embrace.
- Federal Energy Regulatory Commission, Order Rejecting Amended Interconnection Service Agreement, Docket ER24-2172-000, issued November 1, 2024 (the Susquehanna order).
- PJM Interconnection, Informational Report on Reliability Considerations for Co-Located Load, filed January 19, 2026, Docket EL25-49-001.
- PJM Interconnection, Public stakeholder materials on the Co-Located Load tariff development process, January and February 2026 Members Committee briefings.
- Talen Energy Corporation, Request for Rehearing, Docket ER24-2172-001, filed December 2024.
- Day Pitney LLP, client alert, “Federal Energy Regulatory Commission Orders Major PJM Tariff Reforms for Co-Located Load and Behind-the-Meter Generation,” December 2025.
- Baker Botts LLP, client publication, “FERC Issues Order Providing Guidance for Co-locating Power Plants with Data Centers within PJM,” December 2025.
- Gibson Dunn LLP, client alert, “FERC Orders PJM to Revise Tariff to Permit New Transmission Services for Co-Located Loads,” December 2025.