Queue position is the real asset

Site selection for a data center used to weigh land, fiber, tax treatment and labor. Those factors still appear on the checklist and no longer decide anything. The decision is made by whichever utility controls the interconnection, and by when it says power can be delivered.

Developers describe power availability as having moved from a diligence item to the primary constraint, with advisers noting that time-to-power now separates viable sites from ones that merely look good on every other measure.

The differences are not marginal. Analysis of the queues describes secondary markets delivering grid connection in a year and a half to three years while the most contested primary markets run to five, six or seven, which is longer than the useful life of the hardware being planned for.

A queue is also not a list a developer can pay to skip. Position is procedural, established by application, and a site with a good position is worth more than a comparable site without one, which has made queue position an asset traded in its own right.

The backlog behind those timelines is enormous. Reporting on the interconnection problem describes a volume of generation and storage capacity waiting in American queues that exceeds the country's entire installed capacity, which is a study result rather than a forecast.

That backlog is not primarily a data center problem, which is why it will not be solved on a data center timeline. Queues fill with generation and storage projects, and the studies that clear them are sequential, interdependent and slow by design.

The result is a map redrawn by electricity. Regions that were unremarkable for computing have become attractive because a substation has headroom, and established hubs have become difficult for the same reason in reverse.

Operators have responded by bringing their own power rather than waiting. On-site generation, co-location next to existing plants, and bridge arrangements that energize a facility before full grid service all trade cost and complexity for years of schedule.

Those workarounds carry their own permitting and emissions exposure, which is the part rarely mentioned when they are described as solutions. A generator installed to beat a queue is a fuel contract, an air permit and a local political question.

The strategic consequence is that capacity planning now runs on utility timescales rather than product ones. A company committing to serve inference demand in a particular region is making a decision whose constraint was set by a study filed years earlier.

The buildout has not slowed. It has relocated to wherever the electricity already is, which is a different industry than the one that chose sites by tax abatement and fiber routes.