Power Is No Longer an Input in Data Centers. It Is the Product.

A data center site can have land, fiber, approvals and a customer ready to deploy, and still be nearly worthless. The reason is changing what developers are actually searching for.

Power scarcity is changing the economics of development.

On power grid operator PJM, capacity prices jumped from $28.92 per megawatt-day for 2024-2025 to $269.92 for 2025-2026, then climbed again to $329.17 for 2026-2027. Power availability is increasingly becoming the asset around which everything else is assembled.

Land and fiber still matter. But neither creates a viable project if electricity cannot arrive when the customer needs it.

That changes the site-selection conversation. Where is the substation? How much capacity is actually available? When can the first tranche of power be delivered?

A hyperscaler may ultimately need hundreds of megawatts, but not all on day one. If 50 megawatts can be energized now and another 100 can arrive 18 months later, that site may be more valuable than one promising 300 megawatts several years from now.

Power is not valuable simply because it exists on a utility map. It is valuable when it can be delivered.

That distinction is forcing developers deeper into the energy business. They are pursuing utility relationships, transmission access, substations and gas infrastructure with the same intensity they once reserved for land and entitlements.

They are also confronting a mismatch in timelines. AI companies plan deployments in months. Major transmission projects, substations and generation assets can take much longer.

The result is a growing market for solutions that bridge the gap. Behind-the-meter generation, co-generation and direct energy partnerships are increasingly part of the conversation. The question is no longer simply whether a utility can serve a project, but whether and how the developer can assemble an energy system until the grid catches up.

This changes the meaning of site readiness.

A hundred acres beside a fiber route may look attractive. But if meaningful power will not arrive for five years, it may be little more than a land position. A less obvious site with existing electrical infrastructure and a path to phased energization may be worth far more.

It also changes how we should think about existing infrastructure.

Across the country are former manufacturing centers, industrial corridors and legacy energy sites with transmission infrastructure, substations, gas connections or power allocations that are now underutilized.

The buildings may be obsolete, but the infrastructure beneath them may not be. I think of these places as a new form of digital dirt. Powered land describes access to electricity. Digital dirt is broader: land whose value comes from an inherited infrastructure stack that would be difficult, expensive or slow to re-create.

Paducah, Ken., offers an example. The former gaseous diffusion plant there once consumed enormous amounts of power and water. Today, the site is being redeveloped for a major AI data center and energy campus in part because its transmission capacity, water infrastructure, fiber connectivity and industrial land remain valuable.

The building closed, but the infrastructure did not.

That distinction matters because conventional real estate values land based on location, density and the economic activity it can support. AI infrastructure adds another layer. The value of the dirt increasingly depends on the energy infrastructure connected to it and how quickly that infrastructure can become usable capacity.

This changes the meaning of location.

For an office building, location might mean proximity to workers. For logistics, it might mean highways and population centers. For AI infrastructure, location is increasingly defined by the geography of deliverable electricity.

Power does not eliminate the importance of land, fiber, water or approvals. It increasingly determines where the search begins.

Developers are no longer simply finding sites and asking utilities to power them. They are increasingly finding power and asking what can be built around it.

That reversal has implications beyond the data center itself. It changes the value of industrial land, increases the strategic importance of utility relationships, and gives new economic relevance to infrastructure built for another purpose.

The next major AI infrastructure project may not begin with cheap land or generous tax incentives. It may begin with an old substation, an underused transmission corridor or a former industrial site that already has what everyone else is waiting to build.

Geography will increasingly follow power. When power determines where the project can go, electricity is no longer simply an input.

It is the product.

Suhail Y Tayeb is clinical assistant professor at New York University’s Schack Institute of Real Estate and director of the Center for the Sustainable Built Environment.