# Large Load Tariff A rate structure created specifically for very large electricity customers — in practice, data centers — that conditions service on long-term commitments, minimum billing regardless of consumption, and financial security against the customer failing to materialize. It is the instrument American utility regulators converged on between 2025 and 2026 to answer the cost-shifting question, and it is a more interesting object than a rate increase because **its central term is not price but risk**. ## The problem it solves A conventional rate charges for power consumed. That works when a customer's arrival does not require new capital. It fails when a single customer's arrival requires a billion dollars of generation and wires, because the utility must build before the customer draws a watt, and the customer may not show up. If it does not show up, the asset remains in the rate base and its cost is recovered from everyone else — the stranded-cost failure mode described in [[Cost Allocation]]. The large load tariff answers this by making the commitment binding in advance. ## The standard components | Component | What it does | |---|---| | **Threshold** | Defines who is captured. Commonly 25 MW to 100 MW depending on jurisdiction | | **Minimum take** | Customer pays on a floor percentage of contracted demand whether or not it consumes — the core term | | **Long contract** | 12 to 15 years is typical, so the commitment outlasts the depreciation of the assets built | | **Ramp period** | Several years before the minimum bites, acknowledging that campuses energize in phases | | **Exit fees** | Charges for reducing contracted capacity below a threshold | | **Collateral** | Deposits or credit support sized to cover the infrastructure if the project fails | | **Curtailability** | Increasingly, an obligation to reduce demand under system stress | ## What states have adopted - **Indiana**, February 2025 — loads above 70 MW at one site or 150 MW aggregate: minimum 12-year term, exit fees above a 20% capacity reduction, minimum monthly demand charge of at least 80% of contract capacity. - **Ohio**, order 2025-07-09, effective 2025-07-23 — the AEP Ohio settlement: payment on a minimum of 85% of contracted capacity, four-year ramp, with tiered minimums above 75,000 kW. - **Virginia**, approved November 2025, effective 2027-01-01 — Dominion's GS-5 class for customers at 25 MW and above: at least 85% of contracted distribution and transmission demand, at least 60% of generation demand, minimum 14-year contracts. In August 2026 the SCC further ordered Dominion to develop a tariff directly assigning transmission costs to large loads. - **Georgia**, 2025 — minimum billing requirements and longer terms for customers at 100 MW and above, though early termination requires only two years of minimum bills rather than the full remaining term. - **Pennsylvania**, 2026-04-30 — PUC framework requiring interconnection upgrade costs be recovered from the large-load customer, with deposits and collateral sized to the infrastructure. ## The public-power version [[TVA]] produced the sharpest instrument, because a federal corporation's board can act in one meeting where a commission needs a year. On **2026-08-20**, meeting in Memphis, the board moved data centers out of the manufacturing service rate — the discounted industrial rate they had been sharing with factories — into a dedicated class running roughly **10% higher**, phased over three fiscal years. Then the part with teeth: effective **2026-10-01**, new or expanding data-center load above **5 MW** pays a **Capacity Commitment Charge of approximately $1.5 million per megawatt**, payable over three to five years. A 300 MW campus therefore commits to roughly $450 million as a condition of connecting. The FY2027 budget contained **no rate increase for residential customers**, which is the political point the board was making. TVA's Integrated Resource Plan projects a need for 11–32 GW of additional generation, with more than $13 billion of generation and transmission spending through FY2029. > [!note] Attribution correction > Coverage of the TVA decision has circulated the rationale — that the charge reflects incremental capacity "not covered in the base rate" — attributed variously to Tom Rice as CEO. Rice is TVA's **chief financial officer**; the interim president and CEO at the time was Mike Skaggs. Sources conflict on the title and the CEO attribution appears to be an error worth not propagating. ## Pennsylvania: permitting as leverage Governor [[Josh Shapiro]]'s **Executive Order 2026-05**, signed 2026-08-18, applies to data center projects above 25 MW and takes a different route — it uses permits rather than rates. Binding provisions: the Department of Environmental Protection reviews an application only if the developer has made a legally binding commitment to the state's GRID requirements and already holds local approval; all data center projects are **removed from the Permit Fast Track Program**; and **non-disclosure agreements are impermissible** for agencies under the governor's jurisdiction. The cost-allocation piece is weaker than headlines suggested and worth stating accurately. The order does not by itself mandate that data centers pay the full cost of new generation. It directs the governor's special counsel to **advocate before the PUC** for tariff rules charging data centers for PJM backstop auction costs, preventing shifts of those costs to other customers including on data-center insolvency, and ensuring data centers pay Commission-jurisdictional interconnection costs. Advocacy to the rate regulator, not a self-executing rule — because rates are the PUC's, not the governor's. The order's own recitals cite the PJM market monitor's $29.4 billion figure, and describe the federal Ratepayer Protection Pledge as **non-binding** — the state's stated reason for acting itself. ## The voluntary alternative The **Ratepayer Protection Pledge**, announced by President [[Donald Trump]] in the February 2026 State of the Union and formalized in early March 2026, has hyperscalers commit that data-center growth will not raise household bills, including covering the full cost of delivery infrastructure upgrades and paying agreed rates whether or not power is used. Seven companies signed initially — [[Amazon]], [[Google]], [[Meta]], [[Microsoft]], [[OpenAI]], Oracle and xAI. On 2026-07-23 it expanded to more than 200 utilities, developers and cooperatives plus 23 governors. Its commitments closely track the large load tariff's terms. The difference is enforceability, and Pennsylvania's executive order says so on the record. ## The reading Watch which instrument a jurisdiction reaches for, because each carries a different theory of the problem. A **rate class** says the issue is price. A **minimum take** says the issue is risk. A **capacity commitment charge** says the issue is that new capacity has a cost that base rates never contemplated. A **permitting condition** says the issue is that the rate regulator is too slow or too captured to act. A **voluntary pledge** says the issue is reputational. The instruments are not interchangeable, and the choice reveals what the actor thinks is actually wrong. ## See Also - [[Public Utility Commission]] — the venue for most of these - [[Cost Allocation]] — the underlying question - [[Rate Case]] — the slower general proceeding these were designed to bypass - [[PJM Interconnection]] — the wholesale regime whose enforcement these tariffs carry