When the Meter Reads Far Less Than the Contract
Data centers with behind-the-meter generation sign interconnection agreements for large blocks of capacity, yet their utilities’ meters can record substantially less than the contracted capacity on a typical day—the rest is produced on-site. That gap between contracted capacity and delivered electricity is where the commercial tension between large customers and their utilities begins. It sits inside a broader surge in data center electricity demand that utilities are already working to plan around.
Behind-the-meter generation covers any plant installed on the customer side of the utility point of delivery—gas turbines, fuel cells, solar, batteries, or combinations—operated in parallel with the grid or islanded during outages. The technology mix matters less than the change in the customer’s position: the facility becomes both load and source, while the utility continues to provide reliability, backup, and often the only path back when on-site plant fails.
The International Energy Agency’s Electricity 2025 analysis identifies data centre load as one of the fastest-growing sources of new electricity demand. That growth is prompting a closer look at how behind-the-meter generation changes utility planning and cost-recovery assumptions, including how much grid capacity to hold and how to recover fixed costs when a customer generates part of its own power.
The Economics That Make the Move Attractive
For a developer, the appeal often begins with the interconnection queue. In many markets, new large loads face multi-year waits for transmission and distribution upgrades, and available capacity may remain uncertain until studies complete. On-site generation shifts part of the timeline from the utility’s study process to the developer’s own procurement and construction schedule. It does not remove the grid, but it changes which parts of the project are gated by the utility.
Electricity availability and grid constraints now shape location decisions as much as fiber connectivity or land cost. A site that can host its own generation can reduce the amount of grid capacity required at a site where the local network has limited spare capacity, subject to the utility’s requirements for backup and contingency conditions, because the data center does not need the full contracted capacity from the utility on day one. This shift is visible in how developers evaluate sites in regions where grid expansion has not kept pace with load growth.
Whether behind-the-meter generation avoids transmission and distribution charges depends on the applicable tariff and billing structure. Under many tariff designs, energy generated behind the meter does not incur those charges because it never crosses the utility network, and the customer may also reduce demand charges based on peak consumption from the grid. Over the operating life of a data center, those avoided charges can change project economics, even when the on-site plant has a higher levelised cost than grid power. The customer is choosing which network services to pay for, rather than simply purchasing delivered energy.
The appeal is not confined to grid-constrained sites. Some developers use behind-the-meter gas generation as a bridge while waiting for a grid upgrade, or as a way to secure interim capacity in markets where transmission build-out is slower than load growth. The result is a portfolio approach to power supply, in which the grid becomes one supply option among several rather than the default.
What the Utility Sees
Utilities plan transmission and distribution around expected load. When a data center contracts for a large block of capacity, the utility studies the substation, feeders, and upstream transmission needed to serve it, and the utility may invest in infrastructure whose costs are recovered through rates or customer-specific charges. If the customer later installs behind-the-meter generation and draws only a fraction of the contracted capacity, the utility may have committed to infrastructure that is then underused. How the fixed costs of that infrastructure are recovered depends on the utility and the applicable tariff: they may be recovered through the large customer’s own charges, through broader rate allocation across the customer base, or through other mechanisms.
This is the core of the regulatory tension. The utility’s obligation to serve and its right to recover prudent investment were designed around customers that consume delivered electricity. A large behind-the-meter plant disrupts that model, because the customer can appear as part of the system for planning purposes while withdrawing from it for billing purposes. For other ratepayers, the result can be a cost shift: they may cover network costs that a large customer no longer pays, depending on how regulators allocate those costs.
The issue becomes more visible as data center load grows across multiple markets. A single behind-the-meter arrangement may be manageable; dozens in one utility territory change the planning picture. Once a utility’s largest loads begin generating on-site, the fixed network costs built to serve them do not disappear, but fewer ratepayers may remain to pay for them. That concentration of load turns an accounting question into a structural one.
Behind-the-meter plant also makes load forecasting more difficult. Historically, a large commercial customer presented a predictable load shape. On-site generation makes that shape contingent on equipment availability, fuel supply, and market prices. A substantial share of the customer’s peak may or may not appear on the system at any given hour, complicating both short-term operations and longer-term resource planning.
The Regulatory Gray Zone
Regulators are still working through what a utility may charge a customer that generates behind the meter but remains connected for backup. The options include standby rates, minimum bills, reservation or capacity charges, and other customer-specific charges. Each seeks to recover the cost of maintaining the grid capacity the customer still depends on, but each creates different incentives for how much generation to install and how large a grid connection to retain.
A standby rate charges for the capacity made available when the on-site plant is down. A minimum bill sets a floor on the customer’s monthly charge regardless of consumption. Reservation charges recover the cost of reserving grid capacity even when it is not used. The choice matters because a data center can respond to any of these by changing the size of its behind-the-meter plant, the size of its grid connection, or both.
The treatment of backup power is especially contested. A data center may want a grid connection large enough to replace its entire on-site plant during an outage, which means the utility must reserve capacity that may be used only rarely. Some regulators view that as a legitimate cost of providing firm backup; others view it as charging for capacity that was already built under existing rates.
Several jurisdictions are examining whether these mechanisms should apply to large loads with behind-the-meter generation. In parts of Asia’s power grids, where AI-related load is testing utility planning assumptions, similar questions are emerging. The Federal Energy Regulatory Commission has opened a technical conference on large loads co-located at generating facilities, raising the question of how co-located and behind-the-meter arrangements should be classified. The unresolved question centres on who pays for the grid capacity that remains necessary when the on-site plant fails or is unavailable.
Operational Frictions: Load Steps, Islanding, and Fuel Supply
Behind-the-meter generation also changes the utility’s operational relationship with the customer. A large data center with on-site generation may operate islanded during grid disturbances, but if that on-site plant trips, the entire load can return to the grid in a single step. A load step of that size can produce voltage sags on distribution feeders, especially where the local substation was not designed for rapid changes of that magnitude. The transfer between islanded and grid-connected modes carries its own switching challenges, as described in IEEE Std 1547.4-2011.
Grid-connected generators typically go through interconnection studies that model their effect on the surrounding system. Behind-the-meter plant serving only its host may not face the same level of scrutiny, because it is classified as load rather than generation. Yet from the utility’s perspective, the operational impact can be similar when the plant disconnects or when the customer switches between islanded and grid-connected modes.
Fuel supply adds another layer. A behind-the-meter gas turbine may be sized for continuous operation but still depend on natural gas deliveries that are not firm. During cold weather or pipeline constraints, the turbine may not be able to run, and the data center may draw full load from the grid for extended periods. Maintaining reliable service therefore means being prepared for that load even though the customer’s normal grid consumption is near zero.
The Signals to Watch
In some jurisdictions, regulatory discussions and utility proposals are introducing or expanding standby rates, minimum bills, and reservation charges for large behind-the-meter customers. Some developers have pushed back, arguing that behind-the-meter plant reduces pressure on the grid and should be encouraged rather than penalised. The debate now reaches the same questions in markets with rapidly growing AI load.
The debate also touches on a larger issue. Many data center operators pursuing clean energy goals are already navigating the limits of power purchase agreements and 24/7 carbon-free accounting. Behind-the-meter generation can help match load with dedicated carbon-free resources, but it also complicates the relationship with the grid that remains essential for reliability. On-site generation often serves as one piece of a broader procurement strategy rather than a replacement for grid service.
The most useful signals to monitor are the specific tariff cases before state commissions, especially those proposing standby rates or reservation charges for data centers. The outcomes of FERC’s technical conference on large loads co-located with generating facilities may shape how co-location and behind-the-meter arrangements are classified. Reliability standards could also be updated to treat large behind-the-meter plants more like grid-connected generation for operational planning. Each of those developments affects how a customer class that is partly inside and partly outside the traditional grid relationship is treated.
References
- International Energy Agency — Electricity 2025 (context on data centre demand growth)
- Federal Energy Regulatory Commission — Technical Conference on Large Loads Co-Located with Generating Facilities, Docket AD24-11 (regulatory treatment of behind-the-meter and co-location arrangements)
- IEEE Std 1547.4-2011 — IEEE Guide for Design, Operation, and Integration of Distributed Resource Island Systems with Electric Power Systems (islanding, reconnection, and switching dynamics)
