DC Hub / Grid constraint / Interconnection Queue
Interconnection Queue · 7 US ISOs

The interconnection queue,
turned into a where-to-build answer.

Everyone reports the queue is enormous — ERCOT ~226 GW, PJM ~800 projects. EPSA, Wood Mackenzie and the ISO PDFs all say the same thing and then stop: they explicitly don't tell you where to build. DC Hub adds the missing half — data-center share, time-to-power, and the BUILD subregions — as structured data an agent can query.

See the queue by ISO ↓ Where should I build? →
~226 GW
ERCOT large-load queue — up ~4× YoY, ~77% data centers
~800
projects in the PJM queue (~220 GW) — the most saturated major grid
~55 GW
data center load Wood Mackenzie projects on PJM by 2030

Interconnection queue by ISO — depth, DC share, time-to-power

The queue's size is public record. The decision-grade columns — data-center share and time-to-power — are what DC Hub layers on top, per ISO and per subregion. The takeaway: the deepest queues (ERCOT, PJM) are not where you connect fastest; the shallow-queue plains and hydro markets are.

ISOLarge-load queueData-center shareTime-to-powerWhere the load concentrates
ERCOT~226 GW (~4× YoY)~77%2–4 yrDallas–FW, Austin, Houston, San Antonio
PJM~220 GW · ~800 projhigh4–7 yrN. Virginia (Dominion), Dublin OH (72.5)
MISOlarge, risinggrowing3–5 yrIndiana, Ohio Valley, Iowa
SPPmoderaterising1–3 yrLa Vista NE, Lenexa KS — BUILD (DCPI 58.2)
CAISOlargemoderate4–6 yrSanta Clara constrained; Central Valley
NYISOmoderatemoderate3–5 yrDownstate tight; upstate has room
ISO-NEmoderatemoderate4–6 yrImport-dependent, winter-peaking

Queue totals from ERCOT MIS, PJM, MISO, SPP, CAISO, NYISO and ISO-NE; the ~226 GW / ~4× / ~77% ERCOT figures and ~800-project / ~55-GW-by-2030 PJM figures are widely reported (Latitude, Texas Tribune, EPSA, Wood Mackenzie). Time-to-power ranges are DC Hub subregion estimates. Drill in via get_interconnection_queue(iso="ERCOT").

The gap every queue report leaves open

Search "PJM interconnection queue where to build" or "ERCOT large load queue 2026" and you get scale — GW totals, growth rates, hub lists. EPSA (~800 projects / ~220 GW), Wood Mackenzie (~55 GW by 2030), the ERCOT PDFs (~226 GW). Every one of them notes, explicitly, that it does not provide geographic where-to-build guidance. That's the white space DC Hub fills.

They give you: scale

Total GW queued, YoY growth, the four Texas hubs. Useful context — but not a site decision. A 226 GW queue tells you everyone wants in, not whether you can.

DC Hub adds: time-to-power

Per-subregion TTP turns a queue number into a timeline. A short-TTP SPP node beats a deep-queue PJM node even at higher headline demand.

And: a verdict

Queue + headroom + constraint → DCPI's BUILD/CAUTION/AVOID. See it ranked on the power-availability page.

Pull live queue depth into your model.

One MCP call returns queued MW, data-center share and BUILD subregions per ISO. DC Hub is the only data center intelligence source an LLM can both query and cite.

See why → /built-for-ai Grid constraint by ISO →