The financial commitment can come before the electricity.
A customer that asks a utility to reserve or build capacity for a data center, manufacturing plant or other large load may now face minimum bills, long contracts, collateral, exit charges and direct responsibility for certain infrastructure costs before the promised load fully materializes.
The question is no longer only: Who needs the power?
It is increasingly: Who financially guarantees the power if the projected demand does not fully materialize?
That is the shift this article maps.
The documented lengthening of minimum terms
The 2026 LBNL/Brattle research analyzed a sample of 55 large-load tariffs and frameworks. Among the tariffs proposed since 2025, the median minimum contract duration rose to 12 years, compared with five years among earlier tariffs.
That is a median, not an average. It is also a finding about tariffs and frameworks in the research sample, not a national requirement and not proof that every new data-center contract lasts 12 years.
The significance is directional: regulators and utilities are increasingly using longer contractual commitments as one tool for dealing with the uncertainty surrounding large new loads.
Side-by-side: the financial terms large customers face
| Jurisdiction | Threshold | Customer commitment | Duration | Security | Exit exposure | Status |
|---|---|---|---|---|---|---|
| Virginia — Dominion GS-5 | ≥25 MW | 85% minimum transmission/distribution demand; 60% minimum generation demand | 14-year structure, including ramp provisions | Enhanced collateral requirements | Remaining minimum obligations under applicable terms | Approved; effective Jan. 1, 2027 |
| Ohio — AEP Ohio DCT | >25 MW | Minimum demand can reach 85% | Up to 4-year ramp plus minimum post-ramp term | Enhanced collateral for qualifying customers | Exit/reallocation mechanisms under DCT | Approved July 9, 2025; effective July 23, 2025 |
| Kansas — Evergy LLPS | ≥75 MW | 80% of contract demand as minimum monthly bill | 12-year minimum + optional ramp up to 5 years | 2 years of minimum monthly bills | Remaining minimum bills under termination terms | Approved Nov. 6, 2025 |
| Wisconsin — We Energies VLC | ≥100 MW | Full-Benefits model assigns 100% of associated generation-resource costs | 15-year minimum initial term | Credit/security requirements | Specific exit treatment remains distinct from a simple minimum-bill formula | Approved April 24, 2026 |
| Oregon — PGE Schedule 96 | Large-load/data-center framework | Predictable charges when infrastructure built for the customer is not fully used | Specific detailed terms depend on the adopted schedule | Customer obligations under Schedule 96 | Exact exit formula not independently established here | Approved 2026 |
| Pennsylvania | ≥50 MW individual / ≥100 MW aggregate | Statewide model framework rather than one uniform minimum-billing percentage | Utility-specific implementation | Utility-specific implementation | Utility-specific implementation | Model tariff adopted May 12, 2026 |
These mechanisms are not interchangeable.
A minimum-billing requirement creates a payment floor.
A long-term contract extends the period over which that obligation applies.
Collateral provides security against non-payment or other contractual exposure.
An exit fee raises the financial cost of terminating or reducing a commitment.
A direct resource-cost allocation assigns a particular category of generation cost to the customer.
An interconnection security requirement is a separate mechanism tied to connecting the load to the grid.
And a project-qualification process is not itself a financial guarantee.
That distinction becomes especially important in Texas.
Virginia: a new large-load rate class
Virginia's State Corporation Commission created the GS-5 rate class for large-scale energy users at 25 MW or more, with the new class becoming effective January 1, 2027. The Commission also approved minimum charges of 85% of contracted distribution and transmission demand and 60% of generation demand for applicable large-scale customers.
The structure is designed around a specific problem: a utility can commit substantial resources to serve a large customer before the customer's projected demand becomes fully realized.
The SCC has explicitly framed the new class as a way to make large customers bear costs associated with serving them and reduce cost shifting to other customer classes.
But 85% is not "85% of the risk."
That distinction matters. The 85% and 60% figures describe minimum billing obligations. They do not, by themselves, establish what percentage of Dominion's total infrastructure investment is covered or what dollar amount of residual exposure remains elsewhere.
Ohio: minimum demand plus a longer commitment
AEP Ohio's Data Center Tariff was adopted by the Public Utilities Commission of Ohio on July 9, 2025, with the compliance tariff becoming effective July 23, 2025.
The tariff establishes minimum-demand requirements that can reach 85%. AEP's published tariff materials also require qualifying customers that do not meet specified credit and liquidity tests to provide a guarantee or collateral equal to 50% of the total minimum charges for the full contract term.
The tariff also provides mechanisms for assigning some contract capacity to another customer rather than simply leaving unused capacity stranded. Importantly, any such reassignment must satisfy conditions designed to prevent stranded investment from being recovered from other ratepayers.
Ohio's approach therefore combines a payment floor with security and mechanisms intended to make unused capacity transferable rather than automatically becoming someone else's problem.
Kansas: a 12-year floor with a five-year ramp
Kansas regulators approved Evergy's Large Load Power Service plan on November 6, 2025 for new or expanded facilities at 75 MW or more.
The contract has a minimum term of 12 years, with an optional load-ramp period of up to five years. Customers pay a minimum monthly bill based on 80% of contract demand, regardless of actual monthly usage, and provide collateral equal to two years of minimum monthly bills.
The KCC also states that early termination produces an exit obligation based on the minimum bills that otherwise would have been charged.
Kansas therefore uses several layers at once:
- a long contractual commitment;
- a minimum billing floor;
- collateral; and
- an exit mechanism.
The purpose is explicit: protect existing customers from bearing an unfair share of incremental costs associated with the new load.
Wisconsin: moving beyond a minimum-billing percentage
Wisconsin takes a different approach.
On April 24, 2026, the Public Service Commission approved a Very Large Customer tariff for We Energies, extending the minimum initial term to 15 years and lowering the eligibility threshold from a proposed 500 MW to 100 MW.
The Commission also removed a proposed capacity-only option that would have allowed data centers to pay only 75% of the cost of generating facilities. Instead, it approved the Full-Benefits resource model, under which the VLC customers pay 100% of their associated generation-resource costs.
That is materially different from saying "customers pay 100% of all grid risk."
They do not.
The documented rule assigns a specific category of generation-resource costs. It does not establish that every transmission, distribution, market or system-wide financial exposure disappears.
Oregon: the contract changes before the meter does
Oregon's Public Utility Commission approved Schedule 96 for large data centers and other large-load customers served by Portland General Electric.
The Commission says the structure is intended to ensure that large customers pay the costs associated with serving their loads rather than shifting those costs to residential and commercial customers.
The new structure also includes contract requirements designed to create predictable charges if a data center does not use infrastructure PGE has built to serve it. The Commission says sufficient clean-energy resources must also be available before a large customer can take service.
The Oregon source examined for this article does not independently establish every detailed numerical term previously attributed to Schedule 96, including a specific 90% minimum or a 10-to-30-year range. Those figures are therefore not presented here as independently verified facts.
That is an important distinction because the existence of a tariff does not automatically verify every number reported about it.
Pennsylvania: a framework, not a national-style mandate
Pennsylvania's approach is different again.
The Pennsylvania Public Utility Commission adopted a model large-load tariff for customers at or above 50 MW individually or 100 MW in aggregate, with the order entered May 12, 2026.
The Commission describes the action as a statewide framework that electric distribution companies can adopt or adapt.
That means Pennsylvania should not be placed in the same category as a state tariff that establishes one mandatory statewide minimum billing percentage for every qualifying customer.
The model provides guidance on issues including interconnection costs, minimum contract terms, exit fees and collateral, but utilities retain an important role in translating that framework into their own service arrangements.
So Pennsylvania belongs on the map — but as a framework, not as a uniform binding tariff with one statewide set of financial terms.
Texas: two different mechanisms
Texas is where the distinction between project qualification and financial exposure becomes especially important.
Track A — ERCOT Batch Zero
ERCOT's Batch Zero process is an interconnection-study and project-qualification mechanism for large loads.
ERCOT's current Large Load Integration page says entities seeking to interconnect a facility of 75 MW or greater through Batch Zero should follow the PGRR145 eligibility path.
In September 2026, ERCOT also began issuing verification requests for information to large-load entities conditionally included in Batch Zero. The requests require supporting documentation and attestations, and successful completion of verification is a condition for inclusion.
That is important because Batch Zero addresses whether a project qualifies to move through the interconnection process.
It is not itself a retail tariff that guarantees the utility will recover all of its investment.
Track B — §37.0561 and large-load interconnection commitments
Texas separately created a statutory framework under PURA §37.0561 for large-load interconnection standards.
The Texas Register's March 2026 proposal for §25.194 states that the rule is intended to minimize potential stranded infrastructure costs while maintaining reliability. The proposed rule covers financial security, interconnection fees, direct interconnection costs and consequences for withdrawal or failure to meet milestones.
The important point for this article is the mechanism, not an unsupported fixed dollar figure.
The rulemaking record has changed during the process, and the source set examined here does not provide a sufficiently clean final primary-source basis to present one particular per-MW amount as the definitive September 2026 figure.
So the defensible statement is narrower:
Texas has established a statutory framework for large-load interconnection, while the detailed financial requirements must be tied to the applicable adopted rules rather than treated as final based on an earlier proposal.
That is stronger journalism than repeating an attractive number that belongs to a different stage of the rulemaking.
What these mechanisms transfer — and what they do not
Line up Virginia, Ohio, Kansas, Wisconsin, Oregon, Pennsylvania and Texas and the same basic problem appears in different forms.
Utilities and regulators are asking large customers to put money behind projected demand earlier.
They are using:
- minimum bills;
- long contractual terms;
- collateral;
- exit obligations;
- direct assignment of particular resource costs;
- interconnection financial commitments; and
- project qualification and verification.
The objective is straightforward: if a customer asks the system to prepare for a large load and that load arrives late, ramps slowly, shrinks or disappears, the customer should bear more of the financial consequences.
But that does not mean the customer has assumed every dollar of risk.
A minimum-billing percentage is not a percentage of stranded-cost risk.
An amount of collateral is not the same thing as a utility's total investment.
Paying 100% of a particular generation-resource cost is not paying 100% of all system costs.
And a project that passes Batch Zero verification has not thereby guaranteed that every infrastructure investment made around it will be economically recovered.
The evidence establishes risk-allocation mechanisms.
It does not establish a universal dollar-for-dollar measure of how much residual exposure remains with utilities or other customers.
That residual remains the unresolved financial question.
This isn't new — it's bigger
There is another important qualification.
Utilities did not invent long-term contracts, minimum commitments, collateral or exit provisions because of artificial intelligence.
Large industrial customers have faced forms of these mechanisms for decades.
What is different is the combination of speed, scale and uncertainty.
A large data-center project can request an enormous amount of capacity while the infrastructure needed to serve it may take years to plan and construct. A project can also change its expected load, delay construction or fail to materialize.
That makes the old contractual tools more important.
The novelty is therefore not that utilities discovered minimum commitments.
It is that regulators are applying familiar financial tools to a new generation of very large, rapidly growing and sometimes uncertain electricity demand.
Who bears the financial risk?
The answer is becoming clearer, but it is not absolute.
In the jurisdictions examined here, the large electricity customer is increasingly required to put financial commitments behind the capacity it asks the utility to reserve or build.
That can mean a minimum bill even when actual usage is lower.
It can mean a multi-year contractual commitment.
It can mean collateral.
It can mean an exit charge.
It can mean direct responsibility for specific generation or interconnection costs.
The intended effect is to prevent existing customers from automatically becoming the backstop for every cost created by a new large load.
But the evidence does not establish that existing customers are now fully insulated.
The remaining financial exposure depends on what costs the particular mechanism covers, what costs remain in the utility's rate base, how infrastructure is shared, what happens if demand changes, and what the applicable regulatory cost-allocation rules ultimately recover.
The unresolved financial gap
The new rules answer one part of the problem:
Who must put money behind the forecast?
Increasingly, the answer is the large customer requesting the capacity.
The harder question remains:
Exactly how much of the total financial exposure has actually moved away from utilities and existing ratepayers?
The available tariff terms do not provide one national answer.
They were never designed to.
They are different mechanisms applied to different cost categories in different regulatory systems.
That is why the most defensible conclusion is narrower than the headline rhetoric around AI's electricity demand:
America is changing the contract before it changes the grid.
The financial commitment increasingly begins before the electricity is used.
What remains unmeasured, state by state and project by project, is exactly how much risk is still sitting behind that contract.
- LBNL / Brattle — 2026 Large-Load Rate Design Research
- Virginia SCC — Dominion GS-5 / PUR-2025-00058 material
- AEP Ohio — Data Center Tariff
- Kansas Corporation Commission — LLPS approval
- Wisconsin PSC — We Energies VLC decision overview
- Oregon PUC — Schedule 96 / large-load decision
- Pennsylvania PUC — Final Order / Model Tariff, M-2025-3054271
- ERCOT — Large Load Integration / Batch Zero
- ERCOT — PGRR145
- Texas Register — Proposed §25.194 / Large Load Interconnection Standards
Comments
Post a Comment