How this is built

Every company that appears as a developer in the US interconnection queues, with portfolio statistics recomputed directly from project-level records: capacity, project counts, status split, technology mix, and state footprint. Covers CAISO, PJM, MISO, ERCOT, NYISO, SPP, ISO-NE plus non-ISO West and Southeast.

Queue developer fields are noisy — they are full of single-purpose project entities like “Blue Jay Solar II” that are projects, not companies. Of 6,311 candidate names, 2,798 were filtered out as project SPVs and 1,909 fell below the inclusion bar (3+ projects or 250+ MW), leaving 1,599.

The value chain the queues cannot see: software, manufacturers, storage, nuclear, geothermal, minerals, EPCs, capital and offtakers. A software vendor never files an interconnection request, so it would be invisible in queue data no matter how central it is to the industry.

748 companies currently carry a written profile. These are hand-entered and not independently verified.

A project resubmitted across queue cycles appears as many rows carrying the same name, state and capacity. Summing them inflates everything: one developer had a single 600 MW project counted fourteen times, and across the whole directory the raw sum came to 6,129 GW against 3,951 GW of genuinely distinct projects.

So every figure here is computed over one row per project — deduplicated on developer, project name and state, keeping the most recent submission so the status reflects where the project stands today. This makes our capacity numbers materially smaller than a naive read of the same source, and correct.

Interconnection queues record several parties per request, and not all of them built anything. Two of the upstream ISO loaders were populating the developer field from the grid side of the connection: MISO from transmissionOwner, ISO-NE from “Interconnecting Entity”. The result was that pure transmission companies — METC, ITC Midwest, American Transmission Company — appeared as some of the largest developers in the country, and a single offshore wind project was credited with a 2007 pumped-hydro station and four gas plants.

Those two sources are now excluded from developer attribution entirely. ERCOT and SPP use the same field name but there it genuinely resolves to the project company, so they stay.

For utilities, transmission owners and public power, what remains is labelled an interconnection footprint rather than a queue footprint: those projects connect to their system, which is a real relationship but not authorship.

• Queued capacity is requested, not built. Most interconnection requests are eventually withdrawn. A large queue number signals ambition and pipeline, not an operating fleet — use the operational and withdrawn splits on each profile.
• Names are matched, and matching is imperfect. Companies file under subsidiaries, project entities and inconsistent spellings. Where a curated company shares a name with an unrelated queue developer, the two are kept as separate records and labelled.
• Absence of queue data is not absence of activity. Software, capital and supply chain companies show no capacity by design.

Queue data snapshot: 2026-03-20. Directory rebuilt: 2026-09-21. The queue layer is regenerated from source rather than edited by hand, so a rebuild replaces it wholesale.