Behind-the-Meter Power: Shortcut or New Constraint?

For data-center developers, behind-the-meter power can look like the cleanest answer to an overloaded grid queue: put generation next to the load, control the schedule and stop waiting for a utility upgrade. The appeal is understandable. It is also incomplete.

On 29 September 2026, Reuters reported that 29.6 GW of new behind-the-meter gas-fired capacity is planned in the United States through 2030, with data centers accounting for 88% of that pipeline. The same report highlights equipment scarcity and cost pressure. That is the current fact pattern: onsite generation is expanding, but the shortcut has its own queue.

My view is that behind-the-meter power for data centers should not be treated as a substitute for grid due diligence. It is a different development pathway, with different dependencies. The correct question is not whether onsite generation is “faster” in the abstract. It is whether the complete power system — fuel, permits, equipment, redundancy, controls and grid interface — can reach commercial operation before the grid alternative, at an acceptable lifecycle cost.

Behind-the-meter power changes the critical path

A conventional grid-led project starts with a connection position, network studies, reinforcement scope, commercial terms and energisation milestones. Power is not secured because a developer has received a positive conversation or a high-level capacity indication. That distinction is central to powered-land due diligence and to how we model data-center development readiness.

Behind-the-meter generation removes some of those dependencies, but it adds others. The campus may need a firm gas supply, interconnection to a pipeline, air permits, noise studies, water or alternative cooling arrangements, black-start capability, emissions controls, fuel-price hedging and an operating model for assets that are no longer merely emergency backup.

This matters because “onsite” is not one technical configuration. A plant can be fully islanded, grid-connected with limited export, or operated alongside a utility supply for resilience and optimisation. Each structure changes protection design, metering, dispatch rights, reserve requirements and the extent to which the campus still relies on the public system.

The Federal Energy Regulatory Commission’s 2024 technical conference on large loads co-located with generation framed exactly these questions around reliability, ancillary services, backup supply and cost allocation. Those are not theoretical details. They shape tariffs, contracts and the bankability of the power architecture.

The apparent speed advantage has four gates

The first gate is fuel. A gas turbine without firm delivery capacity is no more “powered” than a substation without upstream reinforcement. Developers need to know pipeline proximity, pressure, lateral timing, curtailment exposure, contracting structure and who funds the connection. A map pin for a gas main is not evidence of available firm transport.

The second gate is equipment. Turbine availability has become a scheduling constraint of its own. Capacity, efficiency, emissions performance and service support vary substantially between large turbines, aeroderivatives, reciprocating engines and modular packages. A developer that selects equipment only after land and permitting work may discover that the delivery slot, not the planning approval, is now the critical path.

The third gate is permitting. Continuous generation is regulated differently from standby systems. US projects can face federal and state air-quality requirements; the US Environmental Protection Agency maintains specific standards for stationary combustion turbines. Local authorities may also examine noise, visual impact, fuel infrastructure and community exposure. A technically feasible scheme can still fail the entitlement test.

The fourth gate is operating economics. Developers should compare fuel, variable operations and maintenance, carbon exposure, emissions-control consumables, overhaul reserves and availability guarantees against the delivered grid tariff. The comparison must use consistent boundaries. It is not valid to compare a wholesale gas input with an all-in grid price while excluding the capital cost and operating risk of the plant.

That same discipline applies to grid-upgrade cost allocation. Behind-the-meter generation may avoid one category of network contribution while creating a new set of privately funded assets. The cost has moved; it has not disappeared.

Grid connection still matters

Even a campus designed around onsite generation may need the grid for commissioning, maintenance periods, restart, redundancy, future expansion or the sale and purchase of energy. The relevant measure is therefore not a binary “grid or no grid” label. It is the secured capacity by date, operating mode and contingency.

This is why data-center capacity is not one number. A 200 MW generation headline does not automatically equal 200 MW of deliverable IT load. Auxiliary consumption, derating, redundancy, ambient conditions, maintenance philosophy and cooling all sit between gross generation and usable IT capacity.

The US Department of Energy’s December 2024 data-center energy report explicitly includes onsite generation and storage among the tools for meeting demand growth. It does not present them as isolated from the wider power system. The IEA’s Energy and AI report, published in April 2025, similarly treats supply, networks, flexibility and energy security as connected parts of the same problem.

In practice, the strongest schemes preserve optionality. They advance the utility connection while developing onsite power, then decide which configuration should provide baseload, backup or transitional capacity as evidence matures. That is not always possible, but it is usually better than prematurely collapsing the project into a single pathway.

A decision test for developers and investors

I would ask five questions before describing a behind-the-meter scheme as development-ready.

First, what is the evidence for fuel delivery — not just proximity — and when does it become binding? Second, are equipment slots reserved under terms that match the project schedule and performance envelope? Third, is the permitted operating profile compatible with continuous generation rather than emergency use? Fourth, how does the plant perform under the project’s actual redundancy and maintenance philosophy? Fifth, what grid services remain necessary, and are they contractually available?

Those answers belong in a structured readiness record alongside land control, planning, cooling, fiber and demand. The PowerlandMap methodology separates reported capacity from verified milestones, while market intelligence connects project events to the underlying asset. Developers can then compare the onsite route with utility-led alternatives across market coverage, rather than relying on one headline schedule.

For investors, the key inference is that an onsite-power announcement should change the diligence questions, not close them. For operators, it should prompt an examination of availability, service contracts and operating responsibility. For utilities, it may create a more flexible customer — or a large contingency whose backup requirements still matter.

The emerging pipeline is commercially important, and it deserves to be tracked. But the commissioning-readiness test remains the same: can each dependency be evidenced, sequenced and funded through operation?

Behind-the-meter power for data centers can be a credible route to capacity. It is not free of constraints; it replaces a familiar grid constraint with a coordinated fuel, equipment, permitting and operating programme. That can be faster. It can also be more complex than the first diagram suggests.

To compare project readiness, power evidence and delivery risk across markets, request access to PowerlandMap.