Why Interconnection Queues Are the Real Bottleneck for AI Infrastructure

Before a new large electrical load (or a new generator) can connect to the grid, the utility or grid operator has to study what that connection does to the rest of the system — whether nearby infrastructure can handle it, what upgrades are needed, and who pays for them. That study process, and the backlog of projects waiting for it, is the interconnection queue.

It's not a formality. A large new load can require substation upgrades, new transmission segments, or protection system changes elsewhere on the grid — real engineering work with real lead times, on top of the administrative queue itself.

Two things are happening at once: a wave of new generation (much of it renewable, which typically requires more interconnection studies per megawatt than traditional generation) is trying to connect to the grid, and a wave of new large loads — AI data centers chief among them — is trying to draw from it. Research from groups like Lawrence Berkeley National Laboratory has tracked queue times climbing across nearly every major U.S. grid region over the past several years, and the pattern shows no sign of reversing on its own.

The practical result: in many markets, requesting a brand-new interconnection for a large greenfield data center today can mean a multi-year wait before power is actually available — regardless of how fast the building itself could be constructed.

A site that already has an active or recently active interconnection — a retired power plant, a former smelter, a closed industrial facility with a live substation — isn't starting that process from zero. In many cases, the existing interconnection can be modified or reactivated far faster than a new one can be studied and approved from scratch, because the grid impact of the connection point itself is already understood.

This is the entire premise behind treating powered industrial sites as a distinct, valuable category rather than generic real estate: the hardest, slowest, most unpredictable part of standing up new AI infrastructure is frequently already solved on these properties.

If you're a compute buyer racing a deployment timeline, a greenfield build's interconnection queue position should be one of the first things you ask about, not a detail you discover later — it can single-handedly determine whether a 'fast' build is actually fast. A powered site with existing interconnection sidesteps that risk almost entirely.

If you're a property owner, this is the specific reason a retired industrial site with real transmission-level power infrastructure can be worth pursuing even if it looks, on the surface, like a demolition candidate. The interconnection is very often the most valuable thing left on the property.