Do Data Centers Raise Electricity Bills? How Utilities Decide Who Pays

Data centers can put upward pressure on household electricity bills, but a new facility does not automatically make residential rates rise. The outcome depends on what generation and grid upgrades are needed, which costs the data center pays directly, and which costs regulators allow the utility to recover from broader customer groups.

That answer varies by location because retail electricity rates are largely shaped through state, local, or utility-specific proceedings. A tariff approved in Ohio may use different thresholds and safeguards from one in Oregon or Missouri. Even within a state, a municipal utility, electric cooperative, and investor-owned utility may not follow the same process.

The useful question is therefore not simply, “Does a data center use a lot of electricity?” It is: What new costs does the project cause, and who remains responsible for those costs if its plans change?

Why data centers can require major grid investment

A data center is not just another building connected to a nearby power line. Large facilities may operate around the clock and request tens or hundreds of megawatts of capacity. Serving that demand can involve several layers of the power system:

1. Generation: power plants, contracted energy, storage, or other resources capable of supplying the added demand.
2. Transmission: high-voltage lines and network upgrades that move bulk electricity into the region.
3. Substations: equipment that connects circuits and changes voltage between parts of the grid.
4. Distribution or dedicated service facilities: the local lines, transformers, and other equipment that deliver power to the site.
5. Planning and reserves: capacity the utility or regional grid must arrange so supply remains reliable during peaks and outages.

The U.S. Energy Information Administration’s grid overview explains the same generation-to-transmission-to-distribution chain for ordinary consumers. A large new load can affect one layer or several; it does not follow that every project requires a new power plant or the same package of upgrades.

The scale of the issue is growing. Lawrence Berkeley National Laboratory estimated that U.S. data centers used 176 terawatt-hours of electricity in 2023, about 4.4% of national consumption. Its scenario range for 2028 was 325 to 580 terawatt-hours, or roughly 6.7% to 12% of U.S. electricity use. Those are official projections, not measured 2028 consumption, and the range is wide because future equipment, deployment, and efficiency are uncertain. The figures come from the 2024 United States Data Center Energy Usage Report, summarized by the Department of Energy.

Who normally pays for utility infrastructure?

Regulated utilities recover approved costs through rates. In a simplified process, a regulator determines how much revenue a utility is allowed to collect, assigns costs among customer classes, and approves rates intended to collect each class’s share.

Residential, commercial, and industrial customers do not necessarily pay the same kind of rate. Large users commonly pay both for the electricity they consume and for their peak demand—the maximum amount of capacity they require during a billing period. Rates can also distinguish generation, transmission, distribution, and customer-specific facilities.

The guiding concept is often called cost causation: costs should, as far as reasonably possible, be assigned to the customers or activities that cause them. Applying that principle is not mechanical. A new transmission project might serve one customer initially but provide wider reliability benefits later. A power plant built earlier than planned for a data center might eventually serve everyone. Regulators must decide what is customer-specific, what benefits the shared system, and when those judgments should be revisited.

This is why a data center’s effect cannot be read directly from its electricity use. The consequential step is the regulatory decision that translates investment into charges.

Why data centers create a special cost-allocation problem

Utilities have long served factories, mines, steel mills, and other large users. Data centers intensify several familiar planning risks:

• Size: one project can represent a large addition to a utility’s forecast.
• Speed: a customer may want service faster than generation and major transmission can be built.
• High utilization: many data centers expect to draw substantial power through most hours of the year.
• Uncertain development: multiple proposed projects may compete for sites and grid access, while not all are ultimately built.
• Ramp-up risk: a completed facility may take years to reach its contracted demand.
• Technology and business risk: the amount of computing at a site can change before utility investments are fully recovered.

The utility may need to reserve capacity or begin construction before the customer reaches full operation. If the forecast proves too high, other customers could be exposed to an underused asset—unless the tariff or service contract assigns that risk back to the large-load customer.

How utilities try to keep other customers from paying

Berkeley Lab’s August 2026 technical brief, Electricity Rate Designs for Large Loads, reviewed a sample of 55 large-load tariffs, contracts, and frameworks. It found several recurring protections. Their exact design differs, and not every tariff uses every mechanism.

Minimum contract terms

A long service commitment gives the utility more time to recover an investment. In Berkeley Lab’s sample, standardized minimum terms ranged from one to 20 years. The median among tariffs proposed since January 2025 was 12 years, compared with five years among earlier proposals in the sample.

This is not a promise that the customer will use its full capacity every hour. It is a period during which contractual payment obligations can remain in force.

Minimum billing demand and minimum bills

A minimum billing demand means the utility calculates part of the bill using at least a specified share of contracted demand, even if measured demand is lower. Across the tariffs Berkeley Lab reviewed, the median floor was 80% of contracted demand.

A related minimum bill establishes a revenue floor that may include demand charges, customer charges, energy-related components, and approved riders. These structures are sometimes described informally as “take-or-pay”: the customer commits to a minimum payment even when it takes less power than expected. The actual obligation is whatever the approved tariff and contract say, not the informal label.

Direct assignment and upfront contributions

Some costs can be billed directly to the new customer rather than placed into the shared rate base. A contribution in aid of construction, often shortened to CIAC, is an upfront customer payment for specified infrastructure. Direct assignment can also cover dedicated substations, interconnection work, transmission upgrades, or generation capacity procured for the project.

The boundary matters. Directly assigning every upgrade may overlook benefits to the wider system; assigning too little may leave existing customers subsidizing a project-specific investment.

Collateral and credit requirements

Letters of credit, cash, parent-company guarantees, or other collateral protect the utility if the customer cannot meet its obligations. Some tariffs reduce collateral after years of on-time payment; others connect it to projected minimum charges or the cost of new resources.

Collateral does not itself decide whether an investment was prudent. It is financial security for obligations already established elsewhere in the tariff or service agreement.

Exit fees and capacity resizing

An exit fee can require payment if a customer cancels service, closes early, or materially reduces contracted demand. Some frameworks also let utilities reassign unused capacity to another customer, which can reduce the amount still owed by the original customer.

These provisions are especially important when construction has started but the expected load disappears.

Separate classes, demand charges, and hold-harmless tests

A regulator may place very large users in a separate customer class, approve demand charges designed around their cost profile, or require a study showing that existing customers are not made worse off. These approaches make cost tracking more explicit, although the details still determine how effective they are.

Stranded-cost risk, in plain English

Imagine a utility builds a dedicated substation and commits to new generation because a data center says it will need 100 megawatts. The project later opens at half that level, is delayed for years, or is abandoned. The equipment and financing obligations do not vanish.

The unrecovered portion is a stranded-cost risk: spending incurred for expected service that the expected customer no longer pays enough to support. If no contract, collateral, exit charge, or other protection covers it, the utility may ask to recover the remaining cost from other customers. A regulator could approve, reduce, reallocate, or reject that request depending on the facts and applicable law.

Stranded cost is therefore a risk, not proof that households have already been billed. The protective provisions are intended to determine who bears that risk before it becomes an unpaid balance.

Three current approaches—and what they demonstrate

The following examples were current as of August 2026. They illustrate mechanisms, not a single national template.

State and utility	Status	Selected mechanism	What it demonstrates
Oregon: Portland General Electric Schedule 96	Approved by the Oregon Public Utility Commission in 2026	A distinct rate class for qualifying data centers and other large loads, with contract and cost-allocation protections	A state can require a separate large-load structure aimed at preventing household and small-business subsidies.
Missouri: Evergy and Ameren large-load tariffs	Approved; the Missouri PSC reported both in place in 2026	Long minimum service commitments, minimum demand payments, financial security, and other safeguards	A commission can combine several protections rather than rely on one fee.
Pennsylvania: statewide model tariff framework	Final order released in May 2026; guidance for later utility filings	Interconnection fees, contributions in aid of construction, study requirements, and customer construction options	A commission can establish a model first, while utility-specific terms still require later proceedings.

In Oregon, the PUC’s approval of Schedule 96 created a distinct classification for large data-center loads served by Portland General Electric. The commission described the objective as making the class pay the full and fair cost of infrastructure needed for its service. The approved structure also addresses contract commitments and clean-electricity requirements. It is an adopted utility schedule, not a nationwide rule.

The Missouri Public Service Commission’s large-load tariff guide says approved Evergy and Ameren structures include minimum service contracts and other provisions intended to prevent unjust or unreasonable large-load costs from being charged to residential and commercial customers. Missouri’s approach shows why consumers should read the complete package: contract duration, minimum payments, collateral, exit provisions, and cost assignments work together.

Pennsylvania took a different procedural route. Its Public Utility Commission released a final large-load model tariff order in May 2026. The framework addresses customers at or above specified load thresholds and includes CIAC treatment for interconnection facilities and network improvements. The model guides future utility filings; it should not be mistaken for identical final rates already applied by every Pennsylvania electric company.

Berkeley Lab’s broader review also documents approved structures in states including Ohio, Florida, Virginia, Kentucky, and Wisconsin, alongside proposals still awaiting decisions. That distinction matters: a utility filing describes what the utility wants; a commission order establishes what was approved.

Does a new data center mean your bill will increase?

Not by itself. Several steps normally stand between a project announcement and a change in a residential rate:

Project request → utility load forecast → engineering and resource studies → infrastructure plan → proposed tariff or contract → regulatory review → approved cost allocation → later rate recovery

A household bill is more likely to be affected when a project causes material shared-system spending and the regulator allows some of that spending to be recovered from residential customers. The risk is lower when the large-load customer pays customer-specific costs upfront, remains responsible for minimum payments over a sufficiently long term, and secures its obligations if the project shrinks or exits.

There can also be benefits. A large customer may contribute revenue toward existing fixed costs, finance useful infrastructure, or support resources that later serve other users. Whether those benefits exceed the costs is utility- and project-specific. Claims of either guaranteed savings or inevitable increases deserve scrutiny.

Residential rates can rise for many unrelated reasons, including fuel prices, storm damage, aging equipment, new power plants, transmission work, and financing costs. The EIA’s electricity-price guide provides that broader context. A rate increase occurring near a data-center announcement is not, by timing alone, evidence that the data center caused it.

How to check what is happening where you live

Start with the regulator, not a project press release.

1. Find your state utility commission. Search its docket system for your utility’s name plus “large load,” “data center,” “tariff,” “special contract,” or “rate case.” Municipal utilities and cooperatives may instead answer to a city board or member-elected board.
2. Identify the document’s status. Is it an application, staff recommendation, settlement, proposed order, final order, or effective tariff sheet?
3. Check the eligibility threshold. Large-load protections may apply only above a specified number of megawatts or load factor.
4. Look for cost assignment. Search for “CIAC,” “direct assignment,” “interconnection facilities,” “network upgrades,” and “incremental generation.”
5. Check the downside protections. Find the minimum contract term, minimum billing demand, collateral, exit fee, cancellation terms, and rules for reducing capacity.
6. Read testimony from more than one party. Utility witnesses, commission staff, consumer advocates, customers, and grid operators may disagree about forecasts and risk.
7. Follow the next rate case. Even an approved large-load tariff may interact with a later proceeding that allocates shared generation or transmission costs.

Tax incentives are a separate question from electric rates. For the local-government side of the issue, see our guide to Texas school-district tax breaks for data centers and power plants. That article examines incentive decisions; it does not determine who pays a utility’s grid costs.

Bottom line

Data centers can contribute to higher household electricity costs when serving them requires new spending and regulators assign some of that spending to the broader rate base. But the result is not automatic. Large-load tariffs, long service commitments, minimum payments, direct cost assignments, collateral, and exit provisions can place more of the risk on the customer that caused it.

For consumers, the most revealing documents are not project announcements. They are the effective tariff, the service obligations it requires, the commission order approving it, and the later rate cases that decide how shared costs are recovered.

Frequently asked questions

Do data centers pay the same electricity rate as homes?

Generally no. Residential and large commercial or industrial users are usually placed in different customer classes or rate schedules. Very large loads may also be subject to a dedicated tariff or negotiated service agreement approved under local rules.

What is a large-load tariff?

It is a published schedule of prices, eligibility rules, and service conditions for customers above a specified demand threshold. It may include demand charges, minimum payments, contract terms, collateral, cost assignments, and exit provisions. For regulated utilities, the relevant regulator generally must approve it.

What does “take-or-pay” mean for a data center?

The phrase commonly refers to an obligation to pay for a minimum level of service even when actual use is lower. The legally meaningful details are the tariff’s minimum billing demand, minimum bill, and service agreement—not the shorthand phrase.

Can a data center lower other customers’ rates?

It is possible if its revenue exceeds the incremental costs it causes and contributes to costs that would otherwise be collected from existing customers. That outcome depends on the approved rate design, actual usage, infrastructure spending, and how regulators allocate revenue and costs. It should not be assumed in advance.

Who regulates data-center electricity rates?

For investor-owned utilities, state public utility commissions commonly regulate retail tariffs. Municipal utilities and cooperatives may have different governing bodies. The Federal Energy Regulatory Commission oversees specified interstate wholesale and transmission matters, so a project can involve more than one jurisdiction.

How can I tell whether a tariff is actually in effect?

Look for a final commission order and an effective tariff sheet. A utility application, press release, testimony, or proposed settlement is not by itself proof that the requested terms were approved.