Behind the meter: a data center’s own power supply

Understand on-site generation, grid connections and microgrids—and the questions a private power plant still has to answer.

7 min read

A data center can arrange to generate electricity near its servers rather than purchase every unit through the usual utility connection. This is often described as behind-the-meter power. The phrase identifies an electrical and commercial arrangement; it does not make fuel, permits or reliability problems disappear.

It also does not necessarily mean the facility is completely disconnected from the grid. Some arrangements combine local resources with utility service. Others aim to operate independently. Before judging a proposal, find out which arrangement is actually being proposed and what it expects from the wider power system.

Which side of the meter?

The meter marks a point where electricity service is measured. Behind-the-meter generation serves a customer on the customer side of that service boundary. The generating equipment can be owned by the customer or supplied through another commercial arrangement. Ownership and electrical location are related questions, not identical ones.

The Congressional Research Service describes self-supply as electricity sourced from on-site plants, potentially owned by a data center or a third party. A power purchase agreement for a distant plant is a different arrangement. Buying electricity commercially does not by itself put generation behind a site’s meter.

Think of the distinction as a building’s water connection and its own water system. Knowing that a resource is available on the property does not tell you its reliability, capacity or relationship with public infrastructure. The electrical engineering is different, but the analogy helps separate location from complete independence.

The generation technology still matters

Behind the meter is not a fuel type. A project can involve combustion equipment, fuel cells, renewable generation, storage or combinations of resources. Each has different operating characteristics. The phrase cannot tell you emissions, water needs, fuel requirements or when electricity will be available.

Gas-fired engines and turbines require fuel and generate exhaust. Fuel cells make electricity through electrochemical processes; the fuel and complete system still matter to their environmental assessment. The site’s fuel-cell news brief distinguishes local stack operation from emissions across the broader supply chain.

Wind and solar depend on resource conditions, while a battery stores energy supplied from somewhere else. Storage can move electricity from one time to another, subject to its power and energy limits. Calling a battery a power source without explaining charging can conceal part of the system.

For every proposal, ask for the resource mix and intended operation. Is equipment supplying the normal daily load, supporting peaks or reserved for emergencies? A description of standby backup should not be used as though it describes continuous generation, or the other way around.

A microgrid adds coordination

DOE defines a microgrid as connected loads and local resources within a clear electrical boundary, acting as a controllable system with respect to the grid. It can be designed for grid-connected and islanded operation. Islanded means operating separately from the wider grid.

The controller has a real job: balancing local production, storage and demand while meeting the system’s operating requirements. Merely putting a generator and battery next to a building does not explain how they will behave during a disturbance. The project must specify controls and electrical protection.

DOE also describes barriers such as upfront cost, complex design and regulation. Those qualifications matter when an announcement presents a microgrid as an effortless shortcut. Local control can offer useful capabilities without making the engineering simple.

Ask whether the facility can actually separate from the grid, what remains powered when it does and how it reconnects. Different designs can support different amounts of load for different periods. “Resilient” is a goal until the design and tests establish what it can withstand.

Moving the bottleneck is not removing every bottleneck

Developers may consider local generation when a utility connection cannot meet their desired schedule. That moves attention to generating equipment, fuel delivery, site infrastructure and permits. An earlier route to one resource can create a different set of dependencies.

An on-site plant still needs electrical equipment that makes its output usable for the facility. Transformers, switching equipment and protection systems connect generation with the loads it serves. The building’s cooling and computing equipment need a dependable supply, not merely an installed nameplate rating.

Fuel also has a delivery system. For a proposed gas plant, ask whether the supply and transport arrangements cover the intended operation. For other technologies, examine their own inputs and operating needs. A purchase order for the main generating equipment is not the whole project.

This is a distinction between a shorter route and a completed route. A developer can have a credible reason to pursue on-site power while still facing construction and commissioning work. Report the milestones actually reached rather than treating the strategy as proof of an opening date.

Permits and neighbors remain part of the project

CRS’s review of data-center energy infrastructure discusses federal permitting considerations across generation and associated infrastructure. Actual requirements depend on the project, location and applicable authorities. A private site does not create a general exemption from public rules.

For residents, the relevant questions include normal operating hours, emissions, noise, fuel infrastructure and the applicable permit documents. An emergency generator operated for testing is not the same local exposure question as generating equipment intended to run routinely. The proposal should make that difference visible.

Grid relationships need equal attention. If a site normally self-supplies but expects utility backup, the utility may still need capacity and infrastructure to serve it. Whether and how those costs are allocated depends on the service arrangement. A low annual purchase total does not automatically mean zero system responsibility.

Nor is exporting electricity automatic. Sending power into the grid can introduce additional operating and interconnection requirements. The project’s plan should distinguish serving its own equipment, importing power and exporting power. Each direction has implications beyond the physical presence of a generator.

A checklist for the operating boundary

First identify the electrical boundary and draw the expected flow of power. Does the campus import from the grid during normal operation? Does it expect backup service? Is export proposed? A simple description of these flows is more informative than calling the site independent without qualification.

Then list the equipment that must remain available for the plan to work. Include the fuel or charging source, generating equipment, storage where used, controls and the path to the computing load. If one item is unavailable, ask what replaces it and for how long. The answer belongs in the reliability plan.

Finally, distinguish an engineering capability from a commercial right. Equipment might be physically capable of importing or exporting power, while the service agreement limits when and how it may do so. Likewise, a contractual entitlement to service does not mean every piece of construction is finished. The proposal needs both a workable physical design and the appropriate arrangements.

These are questions for reading public documents, not instructions to operate electrical equipment. Actual switching, protection and commissioning require qualified personnel and a project-specific design.

Why it matters

Behind-the-meter generation can change how a data center obtains electricity, but it does not abolish the need for a complete power system. It shifts which equipment, contracts and approvals deserve attention. The environmental outcome depends on the technology and operation, not on the meter’s location.

This article goes deeper into that operating arrangement than the existing overview of power sources. Read the backup-power explainer alongside it to separate keeping servers alive during interruptions from meeting their normal long-term electricity demand.

When the next private-power announcement arrives, ask what is operating, what is proposed, what runs routinely and what happens when a resource is unavailable. Then ask what the facility still expects from the grid. Those answers show whether the project has a full supply plan or only the first piece of one.

Sources and further reading

Sources checked Oct 24, 2026. Worked examples and analogies are explanatory, not measurements of a real project.