Why Do Data Centers Pay for Power They Don’t Use?

Data centers can be required to pay for electrical capacity they do not fully use because utilities may reserve grid capacity and build expensive infrastructure before the customer reaches its expected load.

That structure is often described as a take-or-pay arrangement. In utility tariffs, the more precise term may be minimum billing demand.

The key point is that a data center is not simply paying for unused kilowatt-hours. It may be paying for the capacity the utility reserved and built to serve it.

What Is a Data Center Take-or-Pay Electricity Contract?

A large-load electricity contract can contain two different billing components:

  • Energy charges: based on kilowatt-hours actually consumed.
  • Demand charges: based on the level of electrical capacity used or billed.

A minimum billing demand creates a floor under the demand portion of the bill.

Assume a data center signs for 100 MW of contract capacity and the tariff uses a 75% minimum billing demand.

100 MW × 75% = 75 MW minimum billing demand

If the facility’s actual monthly peak is only 40 MW, the demand-charge calculation may still use at least 75 MW.

That does not mean the customer is automatically billed for 75% of energy it never consumed.

Demand in kW or MW and energy in kWh or MWh are different billing concepts.

Why Utilities Require Minimum Payments

A 100 MW or 500 MW data center can require dedicated substations, larger transformers, transmission upgrades, switchgear and additional grid capacity.

The financial risk appears if the utility builds for the expected load and the project is later delayed, reduced or canceled.

Without contractual protections, part of that investment could become a stranded cost.

Minimum billing demand, deposits, collateral and long contract terms are designed to keep more of that risk with the large customer that caused the investment.

As an eBay Partner, I may be compensated if you make a purchase.

What Duke Energy Is Proposing in North Carolina

On October 7, 2026, Duke Energy announced a settlement with North Carolina Public Staff and several large customers, including Amazon, Google, Meta and Microsoft.

If approved by the North Carolina Utilities Commission, the agreement would expand large-load protections to new customers of 50 MW or more with an 80% load factor that signed electric service agreements after June 1, 2026.

The settlement includes nonrefundable upfront payments for customer-specific facilities, deposits and security guarantees for shared grid upgrades, and required service under a High Load Factor rate schedule.

North Carolina’s existing Duke high-load-factor schedules use a minimum billing demand equal to 75% of contract capacity.

That 75% is a demand-billing floor.

It should not be confused with a requirement to physically consume 75% of the contracted electricity.

The settlement is not final yet. Duke said a North Carolina Utilities Commission decision is expected by mid-November 2026.

How Minimum Billing Works in Practice

ItemExample
Contract capacity100 MW
Minimum billing demand75%
Minimum billed demand75 MW
Actual monthly peak40 MW
Demand used for minimum-bill calculationAt least 75 MW

→ Related Guide: Do Data Centers Raise Your Electricity Bill?

How Other States Handle the Same Risk

AEP Ohio’s Data Center Tariff uses minimum demand charges, collateral requirements and long contract terms.

Depending on customer size and prior billing demand, the minimum monthly billing demand can reach 85% of contract capacity.

Michigan uses similar protections.

In October 2026, regulators approved a special contract for a Google data center with an 80% minimum billing demand, a 20-year term, early termination payments and credit or collateral requirements.

→ Related Guide: Michigan’s $295M Power Cost Fight

What Happens if a Data Center Cancels Its Project?

Minimum billing demand mainly deals with underuse after service begins.

Cancellation creates a different risk.

Utilities can use upfront contributions, deposits, parent-company guarantees, collateral, long minimum contract terms and exit payments.

AEP Ohio, for example, requires customers to reimburse 100% of certain buildout costs if a project is canceled or delayed by more than 12 months before energization.

This is why a project can create financial obligations even before it begins consuming large amounts of electricity.

Minimum Billing Demand Is Not Transmission Cost Allocation

These issues are related, but they are not the same.

Minimum billing demand determines how much a large customer must pay under its retail electricity agreement.

Transmission cost allocation determines how broader regional grid costs are divided among customers, utilities or zones.

That distinction matters because the same data center can be affected by both mechanisms.

→ Related Guide: Pennsylvania’s New CPTR and Data Center Transmission Costs

Do These Contracts Protect Household Electricity Bills?

They can reduce the risk that households and small businesses are left paying for infrastructure built primarily for a new data-center project. To test the household budget effect of a rate change you choose, try the Electricity Rate Increase Calculator. It models household kWh and fixed charges, not data-center minimum billing demand, and does not predict a rate increase caused by a particular project.

But they do not guarantee lower residential bills.

Electric rates are affected by generation investment, transmission expansion, fuel costs, storm recovery, financing and overall load growth.

Large-load contracts are better understood as a cost-allocation tool.

Minimum billing demand, collateral and exit fees are intended to keep more of the financial risk with the customer that caused the investment.

Bottom Line

Data centers do not simply pay for electricity that never flowed through the meter.

Instead, many large-load contracts require them to pay for a minimum level of reserved electrical capacity.

A 100 MW contract with a 75% minimum billing demand can create a 75 MW demand-charge floor even if actual demand is much lower.

That distinction between MW of reserved capacity and MWh of actual electricity consumption is the key to understanding data center take-or-pay electricity contracts.

About the author

I have practical experience in electrical construction and electrical project coordination in South Korea. For U.S.-focused guides, I research official codes, government data, utility documents, and manufacturer specifications to explain electrical and energy topics as accurately and practically as possible.

Sources & Official References

Duke Energy — October 7, 2026 settlement announcement. The proposed agreement expands protections to qualifying 50 MW+ high-load-factor customers, requires upfront infrastructure payments and remains subject to North Carolina regulatory approval. Duke Energy | News Center

North Carolina Energy Policy Task Force — Duke Energy’s existing high-load-factor schedules use a 75% minimum billing demand based on contract capacity. NC Governor

AEP Ohio — Data Center Tariff. The current tariff includes load-ramp requirements, collateral, minimum demand charges and project cancellation protections. AEP Ohio

Michigan Public Service Commission — October 1, 2026 Google data-center contract approval, including an 80% minimum billing demand, 20-year term and early-termination protections.

As an eBay Partner, I may be compensated if you make a purchase.

Is there something you’d like to see?

Suggest an electrical topic, practical guide, or calculator you would like to see on JHOTAS-LAB. Your suggestion may help shape future content.