When Power Is Not Power
There is a quiet assumption sitting underneath most conversations about data centers.
If you can secure power, you can operate.
The PJM Interconnection is about to prove that assumption wrong.
Across 13 states in the Mid-Atlantic and Midwest, PJM is trying to solve a problem that did not exist at this scale even five years ago. Data centers are no longer just large customers. They are system-level loads that can reshape entire regional grids. Tens of gigawatts of projected demand are now competing for access.
The proposed solution sounds straightforward. Allow data center developers to colocate generation with their facilities. Build power next to the load. Reduce dependence on the broader transmission system. Move faster.
On paper, this looks like the answer.
In reality, it breaks down.
Under the proposed framework, even if a developer builds or secures enough generation to meet its own demand, that power is not guaranteed to serve the data center. It is treated as part of the broader network. It can be redirected. It can be curtailed. It can be taken away in the name of system reliability.
So you can bring your own power and still not have power.
And the timeline for resolving this is not immediate. The framework is not expected to be fully in place until around 2029. That means years of uncertainty in a market that is being driven by immediate demand.
This is where the model collapses for anyone still thinking in traditional development terms.
You can solve for land. You can solve for capital. You can even solve for generation.
But you cannot bypass the rules of the grid.
In PJM, the constraint is no longer whether power exists. It is whether you have the right to use it when you need it.
That is a very different problem.
The Separation of Power and Control
What PJM is exposing is not just a regulatory issue. It is a structural shift in how infrastructure works in the AI economy.
We are watching the separation of power from control.
For decades, the assumption was simple. If you generated electricity, you controlled its use. If you built capacity, you could rely on it. Ownership and access were closely linked.
That link is breaking.
In modern grid systems, especially under stress, electricity is not allocated based on ownership. It is allocated based on system rules designed to preserve reliability across millions of users. Those rules sit above any single project, no matter how large or well capitalized.
That means a data center developer can do everything right and still face curtailment.
This is the part that is hard for the real estate mindset to absorb.
In real estate, control is tied to ownership. If you own the land, you control what happens on it within the bounds of zoning and regulation. In digital infrastructure, control is conditional. It is mediated by systems you do not own and rules you do not set.
This is why power is no longer the asset people think it is.
Access to power is the asset. More specifically, predictable access under defined conditions.
That predictability is now shaped by transmission rights, interconnection agreements, and operating rules that most developers only encounter after they have already committed capital.
This is where Digital Dirt becomes more precise.
Digital Dirt is not land with power nearby. It is not even land with power secured on paper. It is land where power can be delivered, used, and relied upon without interruption under the rules of the system it sits within.
PJM shows what happens when that last piece is missing.
The next generation of developers will not win by simply locking up megawatts. They will win by understanding how those megawatts move through the system, who has priority, and under what conditions they can disappear.
Because in this market, having power is no longer enough.
You have to have control over when and how it shows up.