Electricity Rates and Power Costs for Data Center Sites

Large data centers are among the steadiest electricity customers on any system. They run near their peak most hours of the year, which gives them a high load factor. That profile is why they usually take service on industrial or large-power tariffs rather than commercial ones, and why the industrial sector average is the most useful public benchmark when comparing states.

EIA publishes average prices by state and sector each month in its Electric Power Monthly.⁷ Using 2025 EIA data, the U.S. Chamber of Commerce reported a U.S. industrial average of 8.62 cents per kWh, with Oklahoma at 6.15 cents, Florida at 8.84, Delaware at 9.64 and the District of Columbia at 14.78.¹ EIA’s Short-Term Energy Outlook forecast U.S. industrial prices of roughly 8.3–8.9 cents per kWh by quarter in 2026.⁸

A state average is a starting point, not a quote. It is total industrial revenue divided by total industrial sales, so it mixes small factories on distribution service with large plants at transmission voltage, and customers with very different load shapes. A campus served at transmission voltage, with a high load factor, will often pay a different effective rate than the average suggests. Our guide to tax incentives vs. power costs shows how the two compare over a project’s life.

Most large-power tariffs bill several components separately. The names differ by utility, but the structure is similar.

• Energy charges, in cents per kWh or $/MWh, for the electricity consumed. Many utilities pass fuel and purchased-power costs through a separate fuel or energy adjustment that changes over time.
• Demand charges, in $/kW, based on the customer’s highest demand in the billing period. The basic arithmetic is kW of billing demand times the $/kW rate.⁵
• Transmission and distribution charges, sometimes inside the demand charge and sometimes separate, depending on the market and utility.
• Riders: separate line items that recover specific costs or programs, such as fuel, transmission, environmental compliance, energy efficiency or storm costs.
• Contract terms: minimum bills, minimum demand percentages, contract length and exit fees, set in the tariff or an electric service agreement.²³

Demand charges usually apply to commercial and industrial customers, who have higher peak loads than households.⁵ Some tariffs bill on the customer’s own monthly peak; others also charge for demand at the time of the utility’s system peak. Many large-power tariffs also apply a ratchet, so billing demand cannot fall below a share of a prior peak or of the contract demand. Confirm which methods apply with the utility.

In that example the campus uses about 58,400 MWh a month, so a $4.8 million bill works out to roughly 8.2 cents per kWh. A facility with the same peak but a lower load factor would pay the same demand charges over fewer kWh, and its effective rate would be higher. That is the main reason high-load-factor data centers often see a lower all-in rate than their tariff’s energy charge plus demand charge might suggest at first glance.

The biggest change in data center rates since 2024 is not the price per kWh. It is the arrival of rate classes that make large loads commit to pay for the capacity they request, whether or not they use it.

In Ohio, the Public Utilities Commission approved AEP Ohio’s data center tariff in a July 9, 2025 order. New large data center customers must pay for at least 85% of their subscribed electricity usage, regardless of actual use, for up to 12 years.² The commission found that the approach aligned costs with the customers causing them.² The terms include a four-year ramp period, proof of financial viability and an exit fee if a project is canceled or cannot meet its obligations.⁹

In Virginia, the State Corporation Commission’s November 2025 final order in Dominion Energy’s biennial review created a GS-5 class for customers requesting 25 MW or more, effective January 1, 2027.³¹⁰ Those customers must sign 14-year contracts to pay at least 85% of contracted transmission and distribution demand and at least 60% of generation demand, even if they use less.³

For site selection, these terms change the economics of speculative capacity. A developer reserving 300 MW for a campus that ramps slowly can owe minimum bills on capacity it is not yet using. Model the ramp against the minimum-demand schedule, and read our guide on who pays for data center load for the policy context behind these tariffs.

Where a site sits in the market shapes what the bill contains. In a vertically integrated state, one utility sells generation, transmission and distribution under a single approved tariff. In a retail choice state or in ERCOT’s competitive areas, the customer typically buys energy from a retail provider or the wholesale market and pays a wires utility separately for delivery. Our guides on ISOs, RTOs and utility territories and regulated vs. competitive sites explain those structures.

Market rules also decide how some costs are allocated. In ERCOT, transmission costs for large customers are allocated using their load during four coincident system peaks (4CP), so a large customer that cuts consumption during those intervals can reduce its transmission bill for the following year.⁶ The PUCT’s March 2026 draft report on transmission cost recovery, prepared under Senate Bill 6, noted concern that some sophisticated customers can reduce their obligations this way without a matching reduction in the system costs they cause.⁶ In July 2026 the commission approved for publication a proposed rule that would replace 4CP with a 12 coincident peak (12CP) method measured on 30-minute intervals.¹¹ As of October 2026 that is a proposal, not a final rule, so check its status before modeling transmission costs for a Texas site.

In PJM, capacity costs flow into retail rates and have risen sharply with load growth, which our guide to PJM capacity prices covers. Sites that can shift or curtail load may earn credits or avoid peak-based charges; see flexible data center loads.

Riders can move a bill as much as base rates. Fuel and purchased-power riders follow natural gas and wholesale prices. Transmission riders follow the utility’s transmission investment. Because riders can be reset outside a full rate case, an all-in rate that looks attractive today can change within a year.

Economic development riders run the other way: temporary discounts for new or expanded load. In June 2025, El Paso Electric applied to the Public Utility Commission of Texas for an economic development rate rider for a new data center in El Paso.¹² In Colorado in 2023, a discounted Xcel Energy rate proposed for the QTS data center in Aurora needed commission approval, and both commission staff and the state’s Office of Utility Consumer Advocate argued the deal was too favorable.¹³

Utilities are also narrowing eligibility. Indiana Michigan Power modified its Economic Development Rider 2 to make data centers ineligible for service under the rider.¹⁴ Treat any discount as a negotiated, time-limited item and check whether data centers qualify at all.

Retail prices have climbed. Utility Dive, reporting on the Lawrence Berkeley National Laboratory and Brattle study, said nominal industrial rates rose 27% from 2019, and that utility rate increase requests reached about $18 billion in 2025, the highest in decades.⁴ The 2026 edition of the LBNL study extends the data through 2025 and looks at transmission and distribution costs, capacity market outcomes and the mixed and uncertain relationship between load growth and prices.¹⁵

The increases are uneven. EIA data for February 2026, reported by EUCI, showed industrial prices up about 8.6% year over year to roughly 9 cents per kWh, and across all sectors Virginia (26.3%) and Ohio (21.9%) posted the largest year-over-year increases in average revenue per kWh.¹⁶ Both are major data center markets, and both now have large-load tariffs.²³

• Fuel prices, especially natural gas, which flow through fuel riders.
• Transmission and distribution investment, including upgrades for new large loads.¹⁵
• Capacity costs in organized markets with tight reserve margins.¹⁵
• Rate case outcomes and the share of new costs assigned to each customer class.⁴

For a long-lived campus, the trend matters as much as today’s level. A low-rate state with heavy planned transmission spending and rising capacity costs may close the gap on a higher-rate state within a contract term.

1. 01Identify the serving utility and the tariff the load would take, including any large-load class and its size threshold.
2. 02Get the tariff sheets and riders, and model a full year at your expected peak, load factor and ramp, not a single average rate.
3. 03Read the minimum-demand, contract-length, collateral and exit terms, and model the bill if the ramp slips.
4. 04Check whether energy is bundled or bought separately, and how transmission and capacity costs are allocated.
5. 05Ask about economic development riders and whether data centers qualify.
6. 06Look at pending rate cases and rider filings that could change the numbers within a few years.

Treat the result as an estimate to confirm with the utility and an energy advisor. Power cost is one input among many: a cheap tariff on a site with no near-term capacity is worth less than a moderate one with a firm date. Pair it with power purchase agreements where those are an option. When we screen a site, the serving utility and its tariff are part of the power review, and you can get a site reviewed.