Who Pays for the New Power Lines Serving AI Data Centers?

Conceptual architectural illustration of a transmission pylon and substation connecting a data-center building and a distant residential neighbourhood. Servers
The utility building a power line is not necessarily the party that ultimately pays for it. US transmission rules and Ohio’s data-center tariff show how shared-network costs and the risk of unused capacity are handled separately.

A utility can build a new transmission line for a data-center expansion without sending the entire construction bill to that data center. The company that finances the equipment and the customers who ultimately repay it are not necessarily the same party.

The answer depends on which facilities are being built, which network rules apply and how the customer’s service agreement handles a project that arrives late or never arrives. A dedicated connection, a shared transmission reinforcement and the electricity subsequently delivered through them can have different charging mechanisms.

In the United States, that distinction has become a concrete tariff issue. Ohio provides an instructive example: long-term payment commitments are being used to reduce the risk that other customers inherit infrastructure costs after an expected data-center load fails to materialise.

Follow the cost beyond the site fence

The visible connection is only the last part of a larger electrical system. Supplying a large new load can require work at its connection point and reinforcements elsewhere, where existing equipment would otherwise become a bottleneck.

An upfront customer contribution is one way to recover a particular cost. Recovery through regulated charges over time is another. “The utility is building it” identifies the builder, while “the customer pays its electricity bill” says little about how that bill recovers the network investment.

This is the physical background to AI servers arriving before their electrical supply is ready. Hardware delivery does not create upstream capacity. A connection study must establish what work is needed before a service promise becomes usable power.

One electrical route, different cost questions

Customer connection

What work serves this site, and what contribution or contractual security is required?

Shared network

Which customers benefit from the reinforcement, and how do the applicable rules allocate its costs?

A conceptual boundary, not a universal ownership map. Actual projects, voltage levels and tariffs determine where facilities and charges belong. The arrow shows an electrical relationship, not a percentage split.

A regional line is not automatically a private extension

Transmission planning can identify projects serving several purposes and customers. A reinforcement prompted by new load might also relieve an existing constraint or support a broader reliability need. Its allocation cannot be inferred simply from which building appears nearest on a map.

FERC’s explanation of Orders 1920, 1920-A and 1920-B describes cost allocation for selected long-term regional transmission facilities. Costs must be distributed in a way roughly commensurate with estimated benefits. The process includes regional planning and a role for state entities.

That framework is not a universal formula for every new wire serving a data center. It applies to facilities selected through the planning processes covered by those orders. Local connections and other categories of network work have their own applicable arrangements.

FERC also distinguishes transmission regulation from distribution rates, which are generally handled by state or local regulators. A national claim that “data centers pay for their own power lines” therefore needs both a project boundary and a regulatory boundary before it becomes meaningful.

Ohio’s problem: equipment built for a forecast that disappears

New load can make an investment look recoverable. If the customer contracts for substantial capacity and then uses much less—or abandons the project—the planned revenue changes while the physical equipment remains.

In its July 9, 2025 announcement, the Public Utilities Commission of Ohio said its AEP Ohio data-center tariff decision addressed the risk of shifting underused infrastructure costs onto other customers. That is the regulator’s stated purpose, not a measurement showing that every future cost shift has been eliminated.

The service arrangement changes the financial consequences of a demand forecast. It makes a developer’s commitment more than an estimate submitted during a connection enquiry. The company needs to support its requested capacity with contractual obligations.

This is related to, but distinct from, the question of whether a proposed data center is real enough to plan around. Credible demand helps determine what to build. Contract terms help determine who bears the loss if that demand later falls short.

A minimum demand charge is not a purchase of unused energy

AEP Ohio’s published tariff summary says Schedule DCT became effective on July 23, 2025. Its initial contract term consists of the load-ramp period, capped at four years, plus eight years.

The summary specifies minimum monthly billing demand using contract capacity and earlier billing demand. Its size-dependent formula matters: the widely repeated “85%” shorthand is not a flat minimum for every customer size.

Demand is a capacity-related billing quantity, expressed in kW or MW. Energy is electricity used over time, expressed in kWh or MWh. A minimum demand obligation is not evidence that unused megawatt-hours were delivered.

A demand charge gives the utility revenue associated with keeping capacity available, even when usage is below a contractual threshold. That serves a different purpose from charging for the energy that actually passes through the meter.

The same AEP summary describes a separate pre-energisation protection: its Letter of Agreement requires reimbursement of buildout costs if the customer cancels or delays the project by more than 12 months before the target energisation date. The obligation expires once the project is completed. This addresses a different stage of risk from the subsequent minimum-demand commitment.

“Protected from cost shifting” still needs a boundary

A tariff can strengthen a customer’s payment obligations without proving that its operation has no effect on any other electricity customer. Construction-cost recovery is one question. Wholesale energy costs, generation capacity and other shared-system charges are separate questions.

Likewise, a renewable-energy purchase does not by itself specify who pays for a local substation or a shared network reinforcement. The energy contract and the connection arrangement must be read on their own terms.

For a particular project, the decisive evidence is the approved cost allocation and the enforceable service agreement: the facilities covered, the required payments and the consequences of cancellation or lower demand. The most revealing question is what the developer still owes if its servers never use the capacity that was built for them.

Sources

Rate article