Kortex · US interconnection queues · LBNL end-2024 vintage · calculated live in Kortex

PRODUCTIONevidence: transformed (source-reported records)

The queue is not a pipeline

Nearly two terawatts of generation are waiting to connect to the US grid; routinely discussed as if they were future power plants. Historically, queue capacity behaves like an option set with heavy attrition: most megawatts that enter a queue never operate. Here is every request since 2000, cohort by cohort, with the still-active ones honestly marked undecided rather than counted either way.

Region

Technology

Basis

completion is quoted only for cohorts old enough to have resolved (2000–2018); active projects are right-censored, never counted as failed

What happened to each year's applicants

operational withdrawn suspended still in queue (undecided)

Completion by technology: resolved cohorts only (2000–2018)

share of submitted capacity that reached operation · width of the faint bar = capacity submitted

Queue congestion in a siting screen A clean year can have borrowed hours ↓ CSV; all cohorts

Method & receipt. LBNL "Queued Up" interconnection-request records, end-2024 vintage (the end-2025 release is the declared update path); Kortex's transformation of this dataset reproduces LBNL's published rate statistics within ~1 percentage point (see validation study 2). Cohort = queue-entry year. Active projects are right-censored; reported as undecided, excluded from completion rates; completion is quoted for 2000–2018 cohorts, old enough for most outcomes to have resolved. Hybrid projects count under their first listed technology; secondary-component capacity (mw2/mw3) is excluded; a known under-collection stated by the source. Durations use request-to-operation dates where both exist. This is historical attrition, not a forecast: queue reform (FERC Order 2023) is explicitly changing the process it measures. GET /v2/spatial/queue-cohorts · sources.