# How Should Large Loads Be Charged for the Electricity They Use?

urbanplanadvisor.com · September 25, 2026

> The Direct Answer Large-load electricity cost allocation should assign each customer the costs it causes, while protecting the system from both subsidy...

## The Direct Answer

Large-load electricity cost allocation should assign each customer the costs it causes, while protecting the system from both subsidy and arbitrary “big-user” penalties. For a data center, artificial intelligence campus, industrial plant, housing development, or other electricity user above roughly 50 megawatts, that ordinarily means paying for its contracted supply, metered consumption, interconnection, required network upgrades, and an appropriate share of capacity and reliability expenses. As of September 26, 2026, there is no single national formula because electricity tariffs, market rules, and regulatory jurisdictions differ substantially. The defensible principle is causation: costs caused by a new large load should generally be recovered from that load, while genuinely shared network benefits should be spread among the customers that receive them. A uniform rate increase imposed on every residential customer would not pass that test.

**Also worth reading:** [How Should Cities Zone AI Data Centers for Growth, Infrastructure, and Community Protection?](https://urbanplanadvisor.com/knowledge/how_should_cities_zone_ai_data_centers_for_growth_infrastructure_and_community_protection.php) · [How Should Cities Control Risk When Procuring AI Planning Systems?](https://urbanplanadvisor.com/knowledge/how_should_cities_control_risk_when_procuring_ai_planning_systems.php) · [What Are the Municipal AI Procurement Rules Cities Must Follow in 2026?](https://urbanplanadvisor.com/knowledge/what_are_the_municipal_ai_procurement_rules_cities_must_follow_in_2026.php)

The calculation must also distinguish social cost from customer price. A new data center may increase generation and transmission requirements even when short-term bids appear inexpensive, while its contribution to tax receipts, employment, and local economic activity may create political pressure to subsidize it. Neither fact settles the tariff question. Regulators should compare expected bills, congestion, reliability risk, tax benefits, infrastructure commitments, and the possibility that projected demand fails to materialize before approving special treatment. “Pay your fair share” is directionally useful, but it is not a complete rule until every cost category has been identified and assigned.

## What Large-Load Cost Allocation Means

Cost allocation is the process of deciding which customers pay for electricity infrastructure. Historically, many distribution systems recovered network investment through per-kilowatt or per-kWh rates, spreading fixed costs across all meters. That model works reasonably well when loads are small, similar, and predictable. It becomes less precise when one customer requests tens or hundreds of megawatts, has a different utilization profile, needs remote transmission reinforcement, or may operate for only a few years. Allocating every upgrade through flat residential rates can transfer risk to small users, but assigning a new customer every network cost can also overcharge it for assets that would have been built anyway.

A large-load review normally separates four categories. The first is direct service cost, including the utility's connection work, metering, dedicated facilities, and capacity reservation. The second is the energy cost associated with the electricity actually consumed. The third is network cost, such as substations, feeders, transmission lines, transformers, and related upgrades. The fourth is policy cost, including curtailment, demand-response, tax incentives, retail-rate support, and other obligations imposed by a commission. This separation prevents an administrative tariff from hiding energy, infrastructure, and political choices in one number.

The relevant test is incremental and counterfactual. Regulators should ask what expenditure would be avoided if the large customer did not connect, what expenditure is required because it connects, and what benefit existing or future customers receive. Costs that exist independently should remain with the general rate base, while identifiable incremental costs can be assigned through a contract, special tariff, deposit, minimum bill, or periodic rate adjustment. Allocation should be reviewed as forecasts become actual operating data rather than frozen permanently at interconnection.

## Why Data Centers Are Testing Traditional Tariffs

Data centers are a prominent case because their electrical demand can arrive quickly and operate continuously. The research record supplied for this answer identifies reporting by RMI, the Earthjustice organization, the Center for Global Electricity Projects, Harvard Kennedy School’s Belfer Center, and other institutions on how data centers affect affordability, infrastructure, and electricity prices. Some jurisdictions are considering whether growing data-center load warrants changes to ratemaking, return on equity, curtailment, or cost responsibility. The same debate can involve semiconductor factories, battery plants, green hydrogen facilities, crypto facilities, and concentrated commercial developments.

The central issue is timing. Generation, transmission, and distribution assets require investment years before demand fully materializes, while large projects can cancel, relocate, or operate below announced capacity. If a developer obtains service using a general residential tariff and later leaves, stranded network costs may be shifted to remaining customers. Conversely, requiring a project to guarantee an unusually broad set of infrastructure costs may make some worthwhile investment uneconomic. Utility planners therefore need staged commitments, deposits, milestone-based payments, and periodic true-ups rather than relying on either unlimited corporate credit or unconditional residential backing.

Large loads are also unusual consumers rather than inherently bad ones. A data center may have a high load factor, offer demand-response potential, locate near lower-cost generation, pay substantial property taxes, or support infrastructure used by other customers. It can also cause disproportionate peak demand, depend on a small number of firm transmission paths, or require redundant supply that would be unnecessary for ordinary service. A responsible policy recognizes both the amount consumed and the network characteristics it creates, avoiding the simplification that every large customer should pay a punitive rate.

## The Main Cost Allocation Options

Regulators and utilities can combine several allocation methods. No option is universally correct, and the strongest packages usually blend them. A contract is useful during planning, a tariff governs ongoing service, and a later ratemaking case reconciles actual costs and benefits. The key is transparency: the same cost should not appear in both the customer contract and the general base, and credits for shared benefits should be applied consistently.

| Feature | General-Rate Approach | Large-Load-Specific Approach |
| --- | --- | --- |
| Basic principle | Spreads costs across all rate classes | Assigns identifiable costs to the requesting load |
| Near-term risk | May shift infrastructure risk to residential customers | May require deposits, guarantees, or minimum commitments |
| Treatment of network upgrades | Often recovered through general rates | Can be assigned through a special tariff or contract |
| Benefit to existing customers | No new revenue from the project | May receive congestion, tax, or reliability benefits if credited fairly |
| Administrative complexity | Lower | Higher because costs, benefits, and forecasts must be tracked |
| Main weakness | Subsidy risk and slow cost recovery | Risk of overcharging or treating a large customer as uniquely risky |
| Best application | Mature load growth with limited network impact | New projects with large, discrete, or remote infrastructure needs |

A large-load-specific approach can include an initial contribution, a cost-of-service study, a minimum monthly capacity charge, a non-refundable deposit, or an obligation to fund upgrades as construction proceeds. These mechanisms are not interchangeable. A deposit protects against default; a capacity charge compensates for reserved service; a contribution funds construction; and a tariff rider distributes ongoing network costs. Combining them without explaining each purpose can produce a bill that appears punitive and difficult to audit.

## Practical Rules for Assigning Infrastructure Costs

The first practical step is to create a transparent load-specific cost study. It should identify the proposed service voltage, contracted capacity, expected load factor, ramp schedule, location, redundancy, expected operating life, and any non-firm or interruptible service. For illustration, a project connected at 50 MW should not be treated exactly like one requesting 500 MW, even if both receive the same rate per kWh. Thresholds should be based on planning materiality and system impact rather than a political definition of “data center.”

Second, regulators should divide projects into ordinary and extraordinary components. The ordinary component may include service connection and the share of the substation or feeder that would have been built for ordinary growth. The extraordinary component can include a new line, remote substation, unusually large transformer bank, or additional transmission capacity caused mainly by the project. An example threshold might assign projects requesting at least 50 MW, or projects expected to cause more than 10% of a feeder’s firm capacity, to a detailed review. Those numbers are policy screening tools, not universal legal standards.

Third, estimate benefits as carefully as costs. If the project improves substation utilization, enables a later industrial district, or pays taxes that fund local resilience investments, those effects may justify a credit. However, uncertain future development should receive little present value, and community-wide economic claims should not erase measurable network burdens. Cost causation and economic development are separate tests. A project does not need to prove that every local resident benefits, but it should not receive an open-ended subsidy simply because it advertises jobs.

Fourth, use milestones and true-ups. A utility may reasonably require an initial payment at application, a larger payment before equipment procurement, and reconciliation when the energized capacity and final cost are known. Contracts should address cancellation, delay, relocation, curtailment, changes in load factor, and the treatment of upgrades used by subsequent customers. Reviews every three to five years can update cost allocation without forcing a full tariff proceeding after every minor adjustment.

## How Pricing, Reliability, and Affordability Interact

The cheapest possible wholesale energy price is not necessarily the lowest system cost. A new large load can raise transmission congestion, require voltage support, or encourage generation that is useful only when the load is present. Those expenses can appear in retail bills long after the project is operating. Allocating the energy cost while ignoring system impacts does not make those impacts disappear; it merely makes them less visible.

A fair pricing framework should therefore report both the customer’s bill and the system’s cost. A data center may pay a low bilateral energy price but a high network charge, or a high all-in rate while avoiding years of proposed residential infrastructure. Comparisons should use the same assumptions, including contract length, utilization, financing, and backup supply. Per-MW and per-kWh figures should not be mixed because a fixed network cost becomes more affordable to a high-load-factor customer than to a small intermittent one.

Reliability value is another disputed component. A large customer with flexible operations may help balance intermittent generation, but continuous critical computing load may also require redundancy and firm capacity. Regulators should not assume that interruptibility is available merely because a project says it can reduce demand. The value should be demonstrated with measurable operating limits, penalties for failure, and an alternative for the system when the customer is unavailable. Conversely, dedicated backup assets should not automatically be socialized when they serve one private facility rather than the wider grid.

Affordability protections matter most for small customers, who have less bargaining power and fewer ways to relocate. States responding to load growth may use targeted general rate funding, tax revenue, grants, or special large-load charges, but each tool has fiscal and legal constraints. A politically attractive annual rebate can conceal a structurally unfair cost shift. The better benchmark is whether residential bills, service quality, and utility finances would have improved without subsidized expansion, after accounting for taxes, jobs, and infrastructure that other customers can use.

## Common Mistakes in Large-Load Electricity Policy

A frequent mistake is dividing the entire requested system upgrade by the project’s initial capacity and calling the result a “fair share.” Later customers may use the same line, future projects may rely on the same substation, and the original project may cancel. In such cases, the developer should fund what it actually caused, but remaining users may eventually be assigned unavoidable residual assets through ordinary ratemaking. Both propositions can be true at different times.

Another error is assuming that high taxes equal payment for higher electricity costs. Property or payroll taxes can fund public services, but they do not necessarily compensate a regulator for a specific utility investment. A claimed tax offset should identify which customer bill, rate, or public expenditure it actually replaces. Similarly, job creation should be evaluated separately rather than used to waive tariffs that make other households finance the project. Political benefits are real, but they are not a substitute for cost accounting.

A third mistake is setting one threshold for all regions. A 30 MW request in a constrained urban network may trigger an immediate upgrade, while 100 MW near planned generation and spare substation capacity may cause little incremental strain. Review triggers should be tied to voltage, capacity, feeder penetration, congestion, redundancy, and remaining headroom. A national or state screening threshold can be useful, but final allocation should follow the local system plan.

The final error is promising exceptionally low power rates without reconciling who absorbs the difference. Some states are examining special tariffs or other mechanisms as data-center demand grows, but a discount that relies on future residential customers, deferred grid investment, or an assumption that generation will remain cheap is not free. A credible proposal should identify the subsidy source, the duration, the termination condition, and the consumer-protection rule if the load fails to arrive.

## When to Act, and What Cities Should Do

Action is warranted before a project receives final interconnection approval, not after construction. Municipalities, planners, utilities, and developers should exchange load forecasts early, check zoning and substation plans, and identify whether land use permits could create incompatible expectations. By the time a planning application mentions only annual kWh, it may be too late to distinguish ordinary growth from a project requiring a new high-capacity corridor. A city should therefore request peak demand, power quality, backup generation, expected ramp-up, water and cooling dependencies, and a schedule for phases.

For projects below a material threshold, the default can remain an ordinary utility connection with standard review. Around 10 MW, a local system may absorb growth within existing plans; around 50 MW, a dedicated study becomes more defensible; above 100 MW or when a new transmission line is required, detailed allocation is almost unavoidable. These are planning examples rather than fixed legal cutoffs. Regions with severe congestion may set lower triggers, while unusually robust systems may set higher ones based on a defined share of available capacity.

Cities should negotiate public-benefit commitments separately from regulated cost allocation. Possible terms include construction employment, local hiring, tax agreements, energy-efficiency targets, on-site generation, demand-response capability, or contributions to resilience projects. Such commitments can be valuable, but they should be enforceable and discounted for realistic probability. A developer should not receive a lower system-cost allocation merely for offering uncertain jobs, and a city should not substitute tax revenue for its own infrastructure plan.

Before approval, publish at least four figures: the estimated direct connection cost, the project’s network investment obligation, the expected annual tariff and energy charge, and the identified public benefits. Explain whether deposits are refundable, how cancellation is treated, and when a regulator will conduct a true-up. These disclosures make it harder for either side to claim that a subsidized project is “cost-neutral” merely because construction spending, taxes, and ratepayer impacts are examined separately.

## The Best Default for 2026 and Beyond

The best default is a graduated large-load framework rather than a binary “ordinary customer versus data center” tariff. Small and median users continue paying ordinary retail rates. Projects with high capacity, unusual network needs, or material feeder impact submit standardized studies and may fund identifiable incremental infrastructure, subject to a transparent process and credit for demonstrable shared benefits. The utility receives recovery aligned with construction and payment risk; the developer receives clear rules; existing customers receive protection from unsupported cost transfers.

That framework should be reviewed after actual operating experience. If a project uses far more power than forecast, charges should reconcile to actual cost and capacity. If it consumes less, the developer should not bear every sunk cost that later became useful to other users. If it relocates, contractual commitments must operate without stranding the community. Regular review—perhaps at least every three years for fast-changing markets—also allows regulators to adapt thresholds as generation, transmission, storage, and customer behavior change.

The definitive conclusion is that large electricity users should pay the full cost of the service they require and the incremental system burdens they cause, but they should not be charged twice or treated as a license to impose all system risk on the public. No single percentage of a data center’s bill can be declared universally fair. The defensible answer is a documented, project-specific calculation using cost causation, expected use, measurable benefits, staged payments, and periodic reconciliation, administered through rules that protect both competition and residential affordability.

## Quick answers

### Should data centers pay more per megawatt-hour than ordinary customers?

Not automatically. They may owe more if they require unusually expensive infrastructure, consume scarce capacity during peak periods, or create disproportionate network investment. A fair comparison must include the same service commitments, load factor, contract duration, and network benefits, not merely compare headline energy prices.

### What is a reasonable threshold for a large-load tariff review?

There is no national threshold, but projects around 50 MW or those expected to consume more than roughly 10% of a feeder’s firm capacity are plausible review triggers. Utilities should also examine remote substations, dedicated transmission, voltage requirements, and redundancy because impact matters more than project type.

### Can local taxes offset the electricity costs caused by a data center?

Taxes can fund public services, but they do not automatically compensate a utility or regulator for a specific network investment. An offset should be transparent and linked to an identifiable cost or public benefit rather than used to make an otherwise subsidized power price appear cost-neutral.

### Who pays if a large-load project cancels after construction begins?

The contract should specify deposits, milestone payments, cancellation charges, and treatment of equipment that can be reused. A developer may appropriately fund unavoidable work caused by its cancellation, while assets that provide continuing value to other customers should be reallocated through transparent ratemaking.

### Do cheaper energy contracts make large-load development affordable?

A cheap energy rate can still produce high retail costs when transmission, substation, capacity, and reliability charges are included. Both developers and regulators should compare full delivered cost over the project’s life, including backup service, financing, curtailment exposure, and the cost of dedicated infrastructure.

Canonical: https://urbanplanadvisor.com/knowledge/how_should_large_loads_be_charged_for_the_electricity_they_use.php
Markdown: https://urbanplanadvisor.com/knowledge/how_should_large_loads_be_charged_for_the_electricity_they_use.php/index.md