Skip to main content

The data center industry has a power problem. Not a future problem – a right now problem.

Across key U.S. markets, demand driven by AI and high-density compute is outpacing the grid’s ability to respond. Interconnection queues are years long. Transmission upgrades are slow. And in response, developers have started going directly to the source- striking deals with power generators to secure capacity outside of the traditional grid process.

That approach, commonly referred to as “co-location” or “behind-the-meter” power, is now under scrutiny.

Recent actions involving PJM Interconnection and the Federal Energy Regulatory Commission signal a turning point. The rules that allowed data centers to quietly secure power advantage are being rewritten in real time.

This is not a minor policy adjustment. It is a structural shift that will reshape how data centers get built, and who gets to build them.

What’s Changing

 

At the center of the issue is a simple question: Who pays for power and who carries the risk?

Large data centers have increasingly pursued direct connections to generation assets-nuclear plants, gas facilities, or dedicated on-site systems-to bypass grid constraints. While efficient in theory, these arrangements raise two major concerns:

  1. If a data center secures dedicated power but still relies on the grid for backup or balancing, other ratepayers may end up subsidizing that reliability.
  2. These arrangements sit partially outside traditional grid planning, making it harder for operators to forecast demand and maintain system stability.

Regulators have stepped in to address both.

PJM is now proposing a new framework that tightens rules around “behind-the-meter” power-particularly for large loads above roughly 50 MW-while introducing new service structures that more clearly define reliability and cost responsibility. Most importantly, it pushes large users toward a “pay your way” model, where access to power must be matched with corresponding investment in generation or infrastructure.

The message is clear:

You don’t get to bypass the system without paying for it anymore.

Not Everyone Agrees

 

This shift is not happening quietly.

Large power producers-including companies like Vistra Corp. and Constellation Energy-have pushed back on how PJM is approaching these changes. Their argument is straightforward-overly restrictive rules could slow down the very infrastructure the grid needs most.

From their perspective, co-location deals are not a loophole-they are a solution. They allow new demand (data centers) to be paired directly with existing or new generation, reducing strain on the broader transmission system and accelerating deployment timelines.

There is also concern that forcing all large loads fully into traditional grid structures could create bottlenecks, delay projects, and ultimately limit the ability of the U.S. to keep pace with AI-driven infrastructure demand.

On the other side, regulators and consumer advocates are holding firm. Their position is that without clear rules:

  • Costs could shift unfairly onto existing ratepayers
  • Grid reliability could be compromised during peak demand events
  • And planning visibility would continue to degrade as large loads operate partially outside the system

In short, one side is optimizing for speed and capital deployment. The other is optimizing for system integrity and fairness.

Both are valid. But they are fundamentally in tension.