PJM Moves to Redefine Behind-the-Meter Power for AI Data Centers

PJM Interconnection is moving to rewrite how behind-the-meter power is treated across its grid, signaling a major shift as AI-scale data centers push electricity demand into territory the current regulatory framework was never designed to handle.

For years, PJM’s retail behind-the-meter generation rules allowed customers with onsite generation to “net” their load, reducing the amount of demand counted for transmission and other grid-related charges. The framework dates back to 2004, when behind-the-meter generation was typically associated with smaller industrial facilities or campus-style energy systems.

PJM now argues that those assumptions no longer hold. The arrival of very large co-located loads, particularly hyperscale and AI data centers seeking hundreds of megawatts of power on accelerated timelines, has exposed gaps in how the system accounts for and plans around those facilities.

In February 2026, PJM asked the Federal Energy Regulatory Commission to approve a tariff rewrite that would sharply limit how new large loads can rely on legacy netting rules. The move reflects a broader challenge facing grid operators as the rapid expansion of AI infrastructure begins to collide with planning frameworks built for a far slower era of demand growth.

The proposal follows directly from a December 18, 2025 order from FERC finding that PJM’s existing tariff was “unjust and unreasonable” because it lacked clear rates, terms, and conditions governing co-location arrangements between large loads and generating facilities.

Rather than prohibiting co-location, the commission directed PJM to create transparent rules allowing data centers and other large consumers to pair with generation while still protecting system reliability and other ratepayers. In essence, FERC told PJM not to shut the door on these arrangements, but to stop improvising and build a formal framework capable of supporting them.

Why Behind-the-Meter Power Matters

Behind-the-meter arrangements have become one of the most attractive strategies for hyperscale and AI data center developers seeking to bypass grid bottlenecks. Across the PJM Interconnection footprint, particularly in Northern Virginia and other Mid-Atlantic markets, the queue for new grid service has become a strategic constraint.

Developers are trying to energize multi-hundred-megawatt campuses on accelerated timelines, while utilities and regional planners face transmission limits, slow generation additions, and growing debate over who should bear the cost of serving massive new loads. Those tensions have increasingly drawn the attention of regulators, policymakers, and consumer advocates.

PJM’s Board acknowledged in January 2026 that data centers account for a substantial share of forecast large-load additions and that their scale and speed are unlike anything the system has previously experienced. At the same time, PJM has made clear it does not intend to slow that growth. Instead, it wants the data center sector to participate more directly in addressing the reliability and affordability challenges its demand is creating.

At the center of the debate is the original purpose of behind-the-meter generation. PJM argues the framework was designed for relatively small industrial or campus-style systems and not for giant single-point loads sitting behind power plants.

Under existing rules, behind-the-meter generation allows a facility to net its load against its generation, which can leave large co-located loads effectively outside PJM’s planning assumptions. In materials presented at the May 2025 PJM Large Load Additions Workshop, PJM stated bluntly that “large load behind generation was not intended” under the current rules. The workshop also warned that such configurations may not be fully captured in forward-looking planning models, and in some cases may not even be fully visible to PJM if the generation is classified as retail rather than a PJM resource.

The concern is not theoretical. In testimony submitted to the Federal Energy Regulatory Commission in late 2024, PJM Chief Operating Officer Stu Bresler argued that co-located behind-the-meter loads are effectively excluded from PJM’s load forecasts. Electrically, they sit behind a generator, but the system’s planning models do not treat them as load the broader grid must ultimately be prepared to serve.

That disconnect creates a potential reliability risk. If a co-located generator fails or retires, the regional grid may suddenly be expected to supply a large load it was never planned to support. Bresler’s argument is not that such arrangements are impossible, but that the current framework creates a mismatch between physical reality, system planning obligations, and the political expectations that emerge once a major facility is operating.

In the era of AI-scale infrastructure, PJM argues, those mismatches become increasingly difficult to manage. For grid planners, the concern is simple. A data center sitting behind a power plant may appear invisible in the planning models ... until the day the grid is expected to supply it.

Proposed 50 MW Threshold

PJM Interconnection’s February 2026 filing attempts to resolve the planning mismatch by drawing a clear line between traditional behind-the-meter generation and the new scale of data center demand.

The most consequential element is a proposed 50 MW threshold. Under the filing, new behind-the-meter loads larger than 50 MW would no longer qualify for the legacy netting treatment that allows customers to offset load with onsite generation.

PJM proposes a three-year transition period, and existing behind-the-meter arrangements would be grandfathered for the life of their contracts. Backup generation would not count toward the threshold. In practical terms, the rule is aimed squarely at the emerging class of mega-loads rather than smaller industrial or campus-style systems.

For data center developers, the change is significant. Netting has been attractive because it reduces the amount of load that appears on the transmission system, lowering exposure to certain grid-related charges. PJM now wants to reserve that treatment for smaller facilities while pushing large new loads into explicit transmission-service categories.

Alongside the filing, PJM proposed three new service constructs for co-located loads:

• Interim Network Integration Transmission Service

• Firm Contract Demand Transmission Service

• Non-Firm Contract Demand Transmission Service

The rates and detailed terms for those services will be addressed in a subsequent filing, but the intent is already clear. PJM wants co-located data centers to move out of the gray zone between “on-system” and “off-system” status and instead select a defined service model with corresponding obligations and costs.

The proposal goes well beyond billing mechanics. It is part of PJM’s broader 2026 strategy for managing the surge of large-load development. In January, the PJM Board outlined a six-part framework that includes improved load forecasting, a larger role for state regulators, pathways for large loads to bring their own generation, a “connect-and-manage” model that allows earlier grid access with potential curtailment risk, an accelerated interconnection track for some resources, and a backstop procurement process to address near-term reliability gaps.

The underlying message to the market is plain. The legacy framework cannot accommodate a wave of multi-hundred-megawatt data centers arriving faster than the grid can add generation and transmission. Under PJM’s proposed rules, large loads will increasingly need to bring supply with them, accept operational limits, or pay more directly for the grid services they require.

In effect, PJM is signaling that AI-scale infrastructure can no longer rely on regulatory ambiguity to secure power access. For the next generation of AI campuses, access to power may depend less on proximity to generation and more on how transparently the load participates in the grid.

The Politics of Behind-the-Meter Power

The broader architecture PJM is proposing matters not only for grid planning but also for the politics surrounding large data center development.

For years, behind-the-meter arrangements were often presented as a way for data centers to reduce strain on the grid by supplying some of their own power. PJM’s recent posture is more cautious. While proximity to generation can offer operational benefits, PJM now argues that physical location alone does not resolve the system-level obligations created by very large loads.

A facility may sit next to a power plant, but it can still create planning, reliability, and cost-allocation challenges if the arrangement allows it to avoid paying for transmission infrastructure while remaining dependent on the broader grid when outages or operational disruptions occur.

Those concerns came into sharper focus during the high-profile dispute over a proposed arrangement between Amazon Web Services and Talen Energy involving co-located data center load at the Susquehanna nuclear facility in Pennsylvania. The Federal Energy Regulatory Commission rejected an amended interconnection service agreement that would have expanded behind-the-meter power use for the data center, citing concerns about reliability and potential cost impacts on other customers.

For regulators, the case became a clear illustration of the risks associated with large co-location arrangements. PJM’s latest filing can be interpreted as an effort to prevent future disputes of that kind from being resolved one project at a time.

At the same time, the proposed reforms are generating pushback from other industrial energy users. Trade groups such as the Industrial Energy Consumers of America and the PJM Industrial Customer Coalition argue that the changes could undermine the economics of long-standing behind-the-meter operations at manufacturing facilities.

Industry advocates warn that stricter treatment of co-located loads could make hundreds of megawatts of existing or planned industrial behind-the-meter generation uneconomic, particularly for combined heat and power systems that have historically relied on the netting framework.

That tension helps explain why PJM chose a 50 MW threshold and a transition period rather than eliminating behind-the-meter netting altogether. The filing appears designed to preserve the traditional framework for smaller facilities while separating very large data center loads from a tariff structure PJM believes was never intended to support them.

What This Means for Large-Scale Data Center Development

For developers of AI-scale campuses, the commercial implications are substantial. PJM’s broader large-load strategy effectively pushes the market toward two emerging models.

Bring Your Own Generation (BYOG).
Under this approach, a data center developer or a utility partner arranges new generation to arrive with the load. PJM’s Board explicitly endorsed voluntary BYOG and directed staff to develop an expedited interconnection track for projects that bring new supply online alongside large loads as a way to mitigate curtailment risk.

Connect-and-Manage.
Under this framework, a large load gains grid access more quickly but accepts the possibility of curtailment during stressed system conditions.

Both models are more explicit (and likely more expensive) than relying on legacy behind-the-meter netting rules. Analysts say PJM’s plan could accelerate direct partnerships between data center developers and independent power producers.

According to reporting by Reuters, the package could encourage more private pairings between large loads and new generation because the alternative may involve a more constrained and potentially more costly path through the regional grid framework. In the near term, that dynamic could favor gas-fired generation, which remains one of the few dispatchable resources capable of being deployed quickly at the scale the market is demanding, even as nuclear, storage, and other technologies remain part of the longer-term supply mix.

Redefining Behind-the-Meter Power in the AI Era

At a deeper level, PJM is attempting to redefine what “behind the meter” means in an era of AI-scale infrastructure.

Historically, the term implied a relatively simple arrangement: onsite generation supplying a facility that had limited dependence on the broader grid. PJM now argues those assumptions break down when the load in question is a 200 MW, 500 MW, or even 1 GW data center campus.

At that scale, a facility sitting behind a generator is no longer just a customer with onsite power. It becomes a regional planning issue, a capacity issue, a transmission issue, and potentially a political issue if the arrangement fails and the grid is expected to step in.

Previous tariff structures often obscured that reality. PJM’s new proposal attempts to force it into the open.

When the Federal Energy Regulatory Commission evaluates the filing, it will need to balance three competing priorities:

1. Enabling faster interconnection for large loads and supporting economic growth.

2. Protecting existing ratepayers from subsidizing bespoke power arrangements for the largest new customers on the grid.

3. Preserving legitimate industrial behind-the-meter models that have long supported manufacturing and combined heat-and-power systems.

PJM’s proposal attempts to thread that needle by grandfathering existing arrangements, carving out smaller facilities, and pairing behind-the-meter reforms with new transmission service options and a broader large-load integration strategy.

This is not a minor tariff adjustment. It is one of the clearest signals yet that regional grid operators are moving from reacting to data center growth to structurally redesigning market rules around it.

For developers, the message is increasingly obvious. Securing fast access to power will no longer be as simple as locating next to a generator and labeling the arrangement “behind the meter.”

Large loads will increasingly need to demonstrate how they are supplying power, what level of grid service they expect, what curtailment risk they are willing to accept, and how the costs of those choices will be allocated.

In the PJM region, home to the largest concentration of data center development in the world, that represents a major shift. It may also become a template for how other markets respond as AI infrastructure collides with power systems that were never designed for the speed or scale of this new demand.

In the emerging gigawatt era of digital infrastructure, the question is no longer whether power can be found. The question is how transparently and at what cost the grid is expected to support it.

What Data Center Developers Should Watch

For the data center industry, PJM’s proposal is about more than a technical tariff revision. It reflects a broader shift in how grid operators are responding to the rapid emergence of AI-scale infrastructure.

Three implications are particularly important for developers and operators planning large campuses in PJM territory.

• Behind-the-meter netting may no longer function as a shortcut to large-scale power access. The proposed 50 MW threshold effectively draws a line between traditional industrial behind-the-meter arrangements and the new generation of hyperscale and AI data center loads. Future projects may increasingly need to adopt formal transmission service structures rather than relying on legacy netting rules.
• Bring-your-own-generation models could become more common. As PJM develops expedited interconnection pathways for loads paired with new supply, developers may find it advantageous to pursue direct partnerships with power producers. Gas-fired generation is likely to play a near-term role because of its scalability and development timelines, even as nuclear, storage, and other technologies remain part of the longer-term mix.
• The PJM approach may influence other regional power markets. As large AI campuses emerge across the United States, grid operators in other regions - including Midcontinent Independent System Operator (MISO), Electric Reliability Council of Texas (ERCOT), and California Independent System Operator (CAISO) - are grappling with similar questions about how massive new loads should connect to the grid.

If approved by the Federal Energy Regulatory Commission, PJM’s proposal could become an early template for how U.S. power markets adapt to the accelerating demands of AI infrastructure.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.