State public utility commissions across at least 12 jurisdictions have opened formal proceedings in the 2024 to 2026 window to redesign the rate structure under which hyperscale data centers take service. The collective scope of those proceedings covers approximately 65 GW of announced large-load interconnection requests and roughly $40 to $80 billion of utility transmission and distribution upgrade capex tied directly to those requests. The instrument that is emerging in each docket is structurally similar: a dedicated large-load tariff class, a minimum demand threshold of 25 to 100 MW, a long-term contract requirement of 10 to 15 years, and a cost-causation allocation that places the full marginal upgrade burden on the triggering customer.
The shift is a direct response to the bill impacts produced by the PJM Interconnection 2024 to 2025 Base Residual Auction, which cleared at approximately $269.92 per MW-day across most of the footprint, roughly nine times the prior auction's clearing price. The pass-through to residential customers across Maryland, New Jersey, Ohio, Virginia, Pennsylvania, and Illinois has been quantified at 10 to 20 percent on the supply portion of the monthly bill, depending on utility and rate period. Commissioners, governors, attorneys general, and consumer advocate offices have arrived at the same conclusion in independent proceedings: the residential class will not continue to absorb the cost of large-load build-out.
Between 2026 and 2030, the rate-class re-architecture currently in progress will reallocate $5 to $15 billion of transmission and capacity-related cost from residential ratepayers to hyperscale and large industrial customers across PJM, MISO, SERC, and the Southwest Power Pool. The pricing assumptions used in 2023 to underwrite hyperscale campus development now sit inside a tariff window that will close. Projects that reach commercial operation in 2027 and later will do so under a service agreement structurally different from the one their counterparts received in 2020.
How The Old Tariff Let The Cost Socialize
The data center build-out of the 2010s ran on a tariff inheritance designed for a load mix that no longer exists. The standard large general service rate, the Schedule LGS or GS-3 nomenclature varying by utility, was written for industrial loads in the 1 to 20 MW range with diversified load shapes, modest growth assumptions, and a transmission and distribution cost-allocation methodology that spread upgrade costs across the embedded customer base. The Northern Virginia Electric Cooperative, Dominion Energy Virginia, Georgia Power, AEP Ohio, ComEd, and Duke Energy Carolinas each served their early hyperscale customers under variants of this tariff structure.
The cost-allocation mechanics worked through a process called the Average and Excess Demand method, or the Coincident Peak method, depending on jurisdiction. Both methods produce the same effect in a low-growth environment: large customers pay an embedded-cost rate that reflects historical system investment, and incremental upgrade costs caused by new connections are recovered from the system as a whole through the rate base process at the next general rate case. The mechanism is rational when load grows at 0.5 to 1.5 percent annually and individual customer additions are small relative to the system.
The mechanism breaks when a single customer requests 300 to 1,000 MW on a system whose existing peak is 12,000 MW. The transmission upgrade required to deliver that load, the generation capacity needed to serve it under reliability standards, and the distribution buildout to support the campus all become attributable to that one customer. Under the inherited tariff, those costs nonetheless enter the rate base. They are then allocated across the entire customer pool, including residential ratepayers who did not request the service and receive no benefit from it.
Dominion Energy Virginia's transmission rate base grew by approximately 70 percent between 2018 and 2024, against retail load growth of roughly 10 percent. The mismatch was absorbed quietly through annual rider adjustments and base rate proceedings. The data center industry's transmission cost-of-service was, by Dominion's own filings, materially below the cost-of-service allocated to it under the embedded methodology. The differential was paid by other customer classes.
The American Electric Power Ohio subsidiary, AEP Ohio, filed a similar pattern in its 2024 Electric Security Plan docket before the Public Utilities Commission of Ohio. The utility identified approximately $2.2 billion of transmission upgrade obligations directly attributable to data center interconnections in the Columbus and central Ohio cluster, with a recovery mechanism that would have spread roughly half of that cost across residential and small commercial customers in the standard service offer.
The Indiana Utility Regulatory Commission's docket 45911, opened in 2024 to address Indiana Michigan Power's large-load requests in northeast Indiana, identified an analogous structure. The utility's load forecast added approximately 2,800 MW of expected data center demand against a 2024 system peak of roughly 4,800 MW. The cost recovery methodology in effect would have allocated meaningful portions of the resulting transmission build to general service customers.
The tariff did not break because of a defect in its drafting. It broke because the load characteristics it assumed no longer described the load that arrived.
What PJM 2024 Made Visible
PJM Interconnection's 2024 to 2025 Base Residual Auction, conducted in July 2024 for the delivery year beginning June 2025, cleared at approximately $269.92 per MW-day in the RTO-wide zone, with locational clearing prices in the Mid-Atlantic Area Council zone reaching $466.35 per MW-day. The prior auction had cleared at $28.92 per MW-day. The roughly ninefold increase in capacity revenue obligation translated directly into the capacity charge component of every load-serving entity's wholesale supply cost, which in turn flowed through to retail bills under the supply portion of customer rates.
The Maryland Public Service Commission, in Case No. 9707, quantified the residential bill impact at approximately 24 percent on the supply portion and roughly 15 percent on the total bill across the Baltimore Gas and Electric, Potomac Electric Power Company, and Delmarva Power service territories. The New Jersey Board of Public Utilities documented comparable increases across Public Service Electric and Gas, Jersey Central Power and Light, and Atlantic City Electric in its 2025 basic generation service procurement proceeding. The Illinois Commerce Commission's review of ComEd's 2025 procurement produced similar findings.
The cause of the auction outcome is not contested. PJM's capacity supply curve shifted left because of accelerated coal retirements, slow new gas and storage entry, queue reform that constrained new resource entry, and a demand curve that was repositioned by PJM's Independent Market Monitor to reflect higher reliability requirements. The demand curve shift alone added approximately 7,000 to 9,000 MW of cleared capacity obligation across the RTO. The supply-side constraint and the demand-side reform produced the price spike jointly.
The political consequence was immediate. Governor Wes Moore of Maryland sent a public letter to the PJM Board of Managers in February 2025 calling the auction outcome a transfer from residential ratepayers to data center developers. Governor Josh Shapiro of Pennsylvania filed a formal complaint at the Federal Energy Regulatory Commission, Docket EL25-12-000, challenging PJM's capacity auction parameters and seeking a rehearing. The attorneys general of New Jersey, Maryland, Delaware, and Illinois joined or filed parallel pleadings.
FERC's response, in its order on the complaints issued in mid-2025, did not vacate the auction but accepted PJM's proposal to recalibrate capacity demand parameters for the 2025 to 2026 and 2026 to 2027 auctions. The recalibration is expected to moderate clearing prices but will not return them to pre-2024 levels. The structural conditions that produced the spike, principally the load-growth contribution from data centers against a retiring thermal fleet, remain in place.
The political consequence at the state level, separately, was the immediate opening of dockets to address what state regulators began to call the cost-causation gap. By the end of 2025, formal proceedings had been opened in Ohio, Virginia, Maryland, New Jersey, Pennsylvania, Indiana, North Carolina, South Carolina, Georgia, Tennessee, Mississippi, and Texas. The dockets vary in scope, but the common animating premise is identical: residential customers should not be paying for transmission and capacity costs caused by large-load connections.
The States That Moved First
Ohio has moved most rapidly. The Ohio General Assembly's House Bill 15, enacted in 2024 and refined through Sub. HB 15, established a framework directing the Public Utilities Commission of Ohio to require utilities to file a dedicated large-load tariff class for customers above a defined demand threshold. AEP Ohio's compliance filing in PUCO Case No. 24-508-EL-AIR established the structural template: customers above 25 MW take service under a separately designed tariff with 15-year minimum commitment terms, dedicated cost-causation allocation, and explicit obligations for transmission upgrade funding through a customer-specific deferred recovery mechanism. The first filings under this framework cleared PUCO review in late 2025.
Georgia's Public Service Commission, in Docket 56002, approved an interim rate rider for large loads in early 2025 that applies to Georgia Power customers with demand above 100 MW. The rider includes a long-term contract requirement, dedicated capacity cost recovery, and a step-up demand charge structure. Georgia Power's 2025 Integrated Resource Plan filing, Docket 56010, projected approximately 8,200 MW of incremental data center load by 2031. The interim rider establishes the cost-allocation rules under which that load will be served.
Virginia's State Corporation Commission opened Case PUR-2024-00170 to address Dominion Energy Virginia's industrial rate structure under the pressure of approximately 21,000 MW of disclosed data center interconnection queue in the Northern Virginia cluster. The Commission's order in 2025 directed Dominion to file a new large-load tariff class, with structural elements paralleling Ohio: minimum demand threshold, contract term, dedicated cost recovery, and explicit treatment of transmission upgrade obligations. Dominion's compliance filing, in proceedings continuing through 2026, will set the operating tariff for the largest data center cluster in the United States.
The New Jersey Board of Public Utilities, the North Carolina Utilities Commission, and the Indiana Utility Regulatory Commission have each opened parallel proceedings. The New Jersey docket addresses Public Service Electric and Gas's pending large-load requests in the central New Jersey corridor. The North Carolina docket addresses Duke Energy Carolinas and Duke Energy Progress's 2025 IRP filings, which included approximately 17,000 MW of incremental load forecast through 2035, the bulk of it tied to data center customer commitments. The Indiana docket continues the work begun in 45911.
South Carolina, Mississippi, Tennessee, and Texas are moving more slowly but in the same direction. The Tennessee Valley Authority, although not subject to state PUC jurisdiction, has filed analogous tariff language in its Industrial Service contracts for large-load customers. ERCOT's House Bill 5066 in the Texas Legislature, enacted in 2025, established a framework for cost causation in interconnection studies that operates outside the traditional PUC tariff process but produces similar economic effects.
The Maryland Public Service Commission's order in Case No. 9712, addressing data center rate design across Potomac Electric Power Company and Baltimore Gas and Electric, is expected to issue in mid-2026. The Pennsylvania Public Utility Commission's investigation, opened under Docket I-2025-3024156, addresses the same question across PPL Electric Utilities, Duquesne Light Company, and the FirstEnergy Pennsylvania operating companies.
The proceedings are concurrent and the design principles are converging.
What The New Rate Class Actually Contains
The dedicated data center rate class taking shape across the jurisdictions above contains six structural elements that recur with minor variation.
Minimum demand threshold. The threshold for eligibility, and in most filings the threshold for mandatory class assignment, is 25 MW in Ohio, 100 MW in Georgia, 50 MW in Virginia, and similar ranges elsewhere. The threshold is designed to capture hyperscale and high-performance computing campuses while leaving smaller commercial customers under the standard general service tariff.
Long-term contract requirement. Contract terms run 10 to 15 years, with most filings landing at 12 years. The term is calibrated to align with the depreciation life of the transmission assets being constructed to serve the load. The contract includes minimum demand commitments, ramp schedules, and explicit termination liability if the customer fails to take service at the committed level.
Dedicated cost-causation allocation. Transmission and distribution upgrade costs attributable to the interconnection are recovered through a dedicated rate component charged to the triggering customer, not the general body of ratepayers. The allocation methodology varies, but the most common form is a customer-specific deferred asset recovered through a fixed monthly charge over the contract term.
Generation capacity attribution. The capacity charge component reflects the customer's coincident peak contribution, calculated against PJM, MISO, SERC, or SPP capacity market clearing prices, with no embedded-cost averaging across customer classes. The customer pays the marginal capacity cost of its own load.
Collateral and credit support. The tariffs require parent guarantees, letters of credit, or escrowed capital deposits sized to the projected transmission upgrade cost. Georgia Power's filing specifies collateral at approximately 100 percent of the customer-specific deferred asset balance. AEP Ohio's framework is similar.
Demand reduction and ride-through obligations. Several filings include provisions requiring large-load customers to curtail under specified grid stress conditions, accept ride-through requirements during transmission events, or contribute to flexibility services. The provisions are calibrated to mitigate the reliability burden that large concentrated loads impose on the system.
The six elements together constitute the rate class. The variation across jurisdictions is on the calibration of each element, not on the principle that the elements are necessary. The structural template is settled. The litigation and negotiation in each docket is about thresholds, terms, and collateral requirements.
The Projects That Reprice
The financial close documents executed in 2022 and 2023 for hyperscale campuses across PJM, MISO, and SERC included tariff assumptions that no longer hold. The campuses underwritten on the standard large general service rate with embedded-cost transmission allocation are now in service or under construction inside jurisdictions whose tariff frameworks are being rewritten around them.
The customer-specific deferred asset structures now appearing in compliance filings shift the order of magnitude of all-in delivered power cost for a 500 MW campus by approximately $4 to $9 per MWh. Against an underwritten delivered cost of $55 to $70 per MWh, the increment of $4 to $9 represents a 6 to 16 percent total cost increase, concentrated entirely in the transmission and capacity components of the bill. The economic effect over a 15-year contract life is approximately $260 to $590 million per gigawatt of installed capacity.
Projects with executed Electric Service Agreements that pre-date the new tariff filings are in most cases grandfathered, although the grandfathering provisions vary. The Ohio framework grandfathers customers under existing ESAs through their contract terms but applies the new tariff to any service expansion above 25 MW from the date of PUCO approval. The Georgia interim rider applies to new service requests from the docket opening date forward. The Virginia framework is expected to follow a similar pattern.
The projects most exposed are those that signed letters of intent or non-binding term sheets in 2023 and 2024 but had not converted to executed ESAs before the tariff filings cleared. The hyperscaler customer pipeline behind those LOIs runs into the tens of gigawatts across the affected jurisdictions. Each of those projects now reprices against a tariff that did not exist when the LOI was signed.
The repricing effect is asymmetric across the customer pool. Hyperscalers with internal cost-of-power tolerance above $80 per MWh continue to underwrite the projects with the new tariff incorporated. Hyperscalers operating against $55 to $65 per MWh thresholds are renegotiating site selection, requesting tariff carve-outs, or moving allocated capacity to jurisdictions where the rate-class redesign has not yet cleared.
The campuses that reach commercial operation in 2027 and beyond will, in most cases, have executed under the new framework. The cost stack underwritten in their financial models will be approximately 6 to 16 percent higher than the cost stack that supported comparable projects placed in service in 2024. The differential will be material to project IRR, to lease pricing offered to hyperscaler tenants, and to the competitive position of campuses sited in jurisdictions that have completed the tariff redesign versus those that have not.
What Becomes True By 2030
By 2030, every state with material hyperscale load growth will have a dedicated large-load rate class on file, approved, and operational. Ohio, Georgia, Virginia, Indiana, North Carolina, Maryland, New Jersey, Pennsylvania, Illinois, South Carolina, Tennessee, Mississippi, and Texas will each have completed the tariff redesign cycle. The remaining states with smaller exposure, principally in WECC and the upper Midwest, will follow with derivative frameworks designed against the templates established in the lead jurisdictions.
The cumulative cost shift from residential ratepayers back to large-load customers will measure $5 to $15 billion across the affected RTOs and ISOs over the 2026 to 2030 window. The exact magnitude depends on the rate of incremental hyperscale interconnection and on the calibration of cost-causation methodologies in each jurisdiction, but the order of magnitude is consistent across independent estimates produced by the National Regulatory Research Institute, the Lawrence Berkeley National Laboratory's electricity markets group, and the consumer advocate offices of the affected states.
The tariff-design competence required to close a hyperscale campus financing will become a senior development discipline. Campuses sited in jurisdictions with settled tariff frameworks will close on predictable cost structures. Campuses sited in jurisdictions with open dockets will face financing delays, tenant repricing, and in several cases project cancellations. The capability to read a PUC docket, anticipate the tariff outcome, and structure a customer-specific deferred asset before the rate case clears will be priced into senior development team valuations.
The cost-causation principle will spread outside the data center rate class. Industrial customers with announced expansions above the relevant demand thresholds in steel, ammonia, hydrogen, semiconductor fabrication, and battery manufacturing will be assigned to the same or analogous rate classes. The instrument originally designed for hyperscale data centers will become the default tariff structure for any large-load customer connecting to the grid in the second half of the decade.
The political durability of the rate-class architecture is high. The coalition supporting it spans state attorneys general, governors of both parties, consumer advocate offices, residential ratepayer associations, environmental justice groups, and several utility regulators who view the redesign as a defense of the embedded customer base. The hyperscaler customer interest in reversing the framework is meaningful but politically isolated. The rate class is unlikely to be unwound in the 2026 to 2030 window or beyond.
The map of competitive AI campus development is moving in response. The campuses that close in 2027 and later are those whose tariff design is settled, whose customer-specific deferred asset is sized and collateralized before financial close, and whose service agreement reflects the cost-causation principle in its base economics. The campuses underwritten on the socialized-cost assumptions of 2020 to 2023 are repricing in real time. The ratepayer re-class is the most consequential change to large-load utility economics in the United States since deregulation.