Interconnection queue position: how the line is ordered

You cannot connect a power plant — or a gigawatt-scale data center — to the grid without the grid operator studying the request first. When you file an interconnection request, the operator assigns it a queue number: your ordered place in the line of projects waiting to be studied and approved.

That position is more than a ticket. The operator studies each request to determine how the new project affects power flows, what transmission upgrades the grid needs to accommodate it, and who pays for them. Where you sit in the line historically shaped both when your studies run and how much of the upgrade cost lands on you versus the projects around you.

For decades the queue ran first-come, first-served. Projects were studied serially, roughly in the order they arrived, and each project’s position — and its cost allocation — depended on everything ahead of it. That created a brittle chain: when a project near the front withdrew, the studies behind it often had to be redone, cascading delays down the whole line.

That model is being replaced. Under FERC Order 2023 (issued in 2023, with regions transitioning through roughly 2024–2025), the process shifts to first-ready, first-served cluster studies. Instead of studying requests strictly one-by-one by arrival, the operator groups projects in a geographic area into a cluster and studies them together, allocating shared upgrade costs across the group. The reform also adds readiness and commercial milestones and financial deposits so speculative projects can’t hold a good position indefinitely and clog the line.

The regional grid operators run their own variants of this reform: PJM, MISO, CAISO, and ERCOT each administer queue processes tuned to their footprint, so the exact rules for a given position depend on which ISO/RTO territory the project sits in.

Even a strong queue position does not buy a fast connection. The line moves slowly for structural reasons: many projects need transmission upgrades that are expensive and slow to build, withdrawals upstream can trigger restudies, and the sheer volume of requests overwhelms the capacity to study them.

The scale is the real story. Berkeley Lab’s Queued Up analysis found more than 2,000 GW of generation and storage capacity waiting in the nation’s queues — more than the entire existing US power fleet — with the median project taking about four to five years from request to commercial operation (LBNL). Moving up a few places in a line that long changes the finish date by months, not years.

For an AI operator, queue position is often beside the point. If you need power in 2027, it does not matter whether you are near the front of a line that clears in 2030 — the timeline still misses the build schedule.

This is why large loads increasingly bypass the queue entirely with behind-the-meter generation: building power on-site, on the customer side of the utility meter, so the facility runs without waiting for a grid connection. It converts a queue-and-permitting problem into a procurement-and-construction one — the trade being that you now own the generation and its timeline. Fast-start gas turbines are the common way to energize a site while a grid connection or larger plant is still years out.