Who Pays for Data-Center Grid Expansion? The Contract Risk Behind the Megawatts

The financing consequence is a mismatch between two payment schedules. A developer can owe minimum electricity-service charges before its tenants begin paying enough to support them. Greater revenue protection for the utility does not automatically protect the campus owner's return. The important variables are the start dates, minimum payments and credit support in both contracts.

This article uses Ohio Power as a bounded case study, not a description of nationwide rules. The November 2026 tariff book available on November 13 marks DCT Sheets 223-1 through 223-8 effective August 28, 2026. The original 2025 approval announcement alone is insufficient to explain a current bill. Current tariff book

Utility revenue protection needs a matching customer contract

The Public Utilities Commission of Ohio's July 2025 announcement identified the risk that investments built for data centers would be underused and their costs shifted to other customers. That is a regulatory objective, not a guarantee that residential electricity bills cannot rise. Generation procurement, transmission and other network costs require separate analysis. PUCO approval announcement

There are at least three parties in the commercial chain: the utility providing network service, the developer signing the electricity agreement, and the tenant buying hosting or compute. Minimum-demand charges reduce some uncertainty for the first party. If the tenant's contract starts later, expires earlier or depends on equipment acceptance, the developer retains the difference. This is Atlas's contract-analysis framework, not a claim about an undisclosed customer's terms.

This framework does not establish that every obligation has been transferred in a particular project. A financing assessment should build separate timelines from the electricity agreement, customer lease, construction contract and financing documents. A group-level backlog number cannot reconcile those schedules.

The 85% rule concerns demand, not an annual energy purchase

For New Load, the current DCT distinguishes the ramp from the period that follows it. During the ramp, the floor is 85% of agreed ramp contract capacity. Afterward, billing must consider measured monthly demand, 85% of the highest previously established monthly billing demand over the preceding 11 months, and the tiered minimum. Existing Load has separate provisions. DCT Sheets 223-3 and 223-4

Consider a single 100MW New Load contract without other affiliated loads in the calculation. Its post-ramp tiered minimum is 57.5+(100−75)=82.5MW, below the cap of 85% of contract capacity, or 85MW. If the historical billing-demand high reached 100MW, the historical floor would instead be 85MW. If current measured demand reaches 90MW, billing cannot fall below 90MW. Thus, 82.5MW is a conditional floor in this example, not the billable quantity in every month.

Describing this as a requirement to buy 85% of contracted annual energy confuses kW with kWh. Demand measures the capacity relevant to network service; energy measures consumption over time. A low-utilization campus can consume less electricity while retaining substantial capacity-related charges. Generation, transmission, reactive demand and applicable riders still need to be added.

Contract capacity is also different from GPU nameplate power. A campus must serve cooling and other facilities alongside computing. No power-usage-effectiveness assumption is used here to turn 100MW into a GPU count, and no electricity minimum is treated as guaranteed cloud revenue.

A 100MW sensitivity shows the cost of a slow ramp

The following illustration assumes primary-voltage service, a final New Load contract of 100MW, the published $8.08 per kW monthly distribution charge and a $186.30 monthly customer charge. The voltage category is selected to demonstrate the mechanism; it is not a claim that 100MW campuses normally connect at that voltage. Transmission-voltage service has a different rate structure. All results exclude generation, transmission, other riders, reactive charges and tax, and are not all-in electricity quotes. DCT Sheet 223-3

To isolate the ramp, the first four years use the minimum agreed ramp percentages stated in the tariff, assuming measured demand remains below each corresponding billing floor. The post-ramp tiered case separately assumes the historical-demand floor is no higher. Monthly base distribution cost equals billing demand in MW multiplied by 1,000 and $8.08, plus $186.30.

Source: DCT rates and ramp definitions, with Atlas calculations. Four years is the ramp chosen for this illustration; the tariff allows a ramp of no more than four years. It does not establish a four-year ramp for every customer. Ramp-period floors and the post-ramp tiered floor must not be interchanged.

Suppose post-ramp actual demand is 40MW and the historical floor is no higher than the tiered minimum. The monthly base distribution cost remains $666,786.30. Billing solely on 40MW at the same rate and customer charge would produce a comparison of $323,386.30. The difference is $343,400 monthly, or $4,120,800 if those conditions persist for 12 months. This is an annualized sensitivity, not a reported loss or an assumption that rates must remain unchanged for a year.

For a separate unit-cost comparison, assume a 720-hour month with a flat 40MW load. Energy consumption is 28.8 million kWh, making base distribution cost approximately 2.315 cents per kWh. At a flat 100MW load, including 100MW measured demand, the comparable monthly charge is $808,186.30 over 72 million kWh, or approximately 1.122 cents per kWh. These are allocations of base distribution costs, not generation prices. Converting reserved capacity into paying load can improve both revenue and the share of this cost borne by each unit of service sold.

Credit support can consume funding before revenue arrives

The initial New Load term is at least the ramp plus eight years. The current tariff also sets credit and liquidity conditions for security, allowing qualifying corporate guarantees, standby bank letters of credit or cash. The security requirement is tied to full-term minimum charges. It should not be described as a universal requirement to deposit the same amount of cash upfront. DCT Sheets 223-5 through 223-7

The financing effects differ. Cash collateral reduces money available for equipment and construction. A letter of credit can use bank capacity and incur fees. A parent guarantee transmits project exposure to the group. These are different cash and credit commitments, not interchangeable operating expenses. Without a customer's security documents, the simplified distribution model cannot establish its complete collateral obligation.

Exit terms also determine how quickly a weaker-demand project can adjust. A reduction in compute demand does not necessarily remove minimum electricity charges at the same time. A useful stress test combines tenant departure, continuing electricity commitments and a potential request for additional security. Testing interest rates alone misses this contractual liquidity risk.

On-site generation creates a different set of obligations

Behind-the-meter generation can change the mix of grid capacity and energy purchases. It does not establish an unconditional right to draw full replacement grid capacity when the on-site source fails. For customers electing to net generation against contract capacity, DCT sets associated rapid-curtailment requirements and preserves safety and reliability remedies. DCT Sheets 223-7 and 223-8

The comparison must include two complete systems: the minimum cost of retaining more grid capacity, and the capital, fuel, maintenance, backup and curtailment arrangements for on-site supply. If a generation outage forces IT load reductions, service credits and customer retention also matter. A fuel-cell or turbine equipment quote cannot by itself establish the cheaper power strategy.

Phasing electricity contracts alongside tenant commitments offers another adjustment. It can reduce early fixed obligations but may sacrifice queue position, purchasing leverage or future connection certainty. Customer commitments and local network conditions determine the trade-off; the available evidence does not support one universal solution.

Protecting other customers and lowering total costs are different tests

The strongest case for minimum charges starts with the utility's timing problem: networks require capital before usage produces revenue. Enforceable, collectible payment commitments can reduce some recovery risk if a customer cancels or underuses infrastructure. Developers can also benefit when clearer terms allow a connection agreement that would otherwise remain unavailable.

The countercase is that those commitments raise the funding threshold and delay projects without strong credit or secured tenants. If uncertain demand still drives network planning, or guarantors weaken in the same industry downturn, contractual minimums may not cover all actual costs. Performance should be assessed through delivered load, collected cash, network expenditure and loss allocation, rather than signed megawatts alone.

The legal outcome must remain separate. The Ohio Supreme Court's public docket for Case 2025-1458 was open when read for this article. A November 10 entry scheduled oral argument for January 5, 2027. That is a procedural event, not a final merits ruling. This analysis uses the publicly posted tariff, does not forecast the appeal and is not project-specific legal advice. Court docket

Over the next 12–36 months, compare customer coverage with power commitments

Atlas's conditional view is that enforceable minimum electricity charges, combined with insufficient tenant coverage, make credit support and cash planning more important competitive advantages for new AI campuses. Demand can keep growing while an early start to electricity obligations lengthens the period during which sponsors fund the project. Equal megawatts can therefore produce different equity returns.

Evidence supporting that view would include fuller disclosure of customer minimums, prepayments, guarantees, connection dates and acceptance milestones, followed by collections that match network commitments. Rapid load absorption, effective contractual pass-through or material changes in tariffs and risk allocation would weaken it. These observations should precede judgments about an appropriate debt spread or equity return.

For utilities, the corresponding questions are when investment becomes recoverable, whether customer credit support is collectible and how the allowed return compares with financing cost. Minimum charges improve visibility; they do not make every dollar of incremental cash available to shareholders. This article provides no security rating, price target or trading recommendation.