Queued Up: 2026 Edition (LBNL / Interconnection.fyi)

Abstract

Queued Up is Lawrence Berkeley National Laboratory's annual snapshot of trends in generator/storage requests for interconnection to the US bulk-power transmission grid, produced in partnership with Interconnection.fyi and funded by DOE (Contract No. DE-AC02-05CH11231). The 2026 edition summarizes data through the end of 2025 from all seven ISOs/RTOs plus 50 non-ISO utilities — collectively ~98% of installed US generating capacity.

Key Contributions / Findings

• Total active queue: ~2,061 GW seeking interconnection as of end-2025 (1,312 GW generation + 749 GW storage) — a 10% YoY decrease in total active volume.
• Generator mix shifted sharply toward gas. Natural gas capacity in queues increased 86% in 2025, while solar (-19%), storage (-16%), and wind (-19%) all declined — the clearest queue-level signal yet of AI/datacenter-driven firm-power preference over intermittent renewables.
• Withdrawals now exceed new requests for the second year running: over 750 GW of capacity withdrew from queues in 2025 vs. ~600 GW of new requests submitted.
• Completion rates remain low and timelines are lengthening: only ~19% of projects (13% of capacity) that requested interconnection 2000-2020 had reached commercial operation by end-2025. The median project completed in 2025 took 61 months from interconnection request to commercial operation, up from 36 months in 2015 and 22 months in 2008.
• Processing volume hit records in several regions: CAISO, ERCOT, SPP, and MISO each executed more interconnection agreements (IAs) than in any prior year; CAISO and ERCOT each executed IAs summing to >50% of their peak-load capacity in 2025.
• New in this edition: coverage of resource-adequacy "fast-track" programs — MISO's Expedited Resource Addition Study (ERAS), SPP's ERAS, and PJM's Reliability Resource Initiative (RRI) — which carve gas, storage, and some nuclear/other capacity out of the standard cluster-study queue for faster interconnection.

Methodology

Empirical analysis of publicly available ISO/RTO and utility interconnection-queue data, supplemented by non-public data provided directly to LBNL by transmission providers to fill gaps. Covers generation and storage resources connecting to the transmission grid only — large-load (datacenter) interconnection queues are tracked separately and are explicitly out of scope for this report.

Limitations

• Report reflects data through end of 2025 only; any 2026 developments are not captured.
• Valid in-service dates were available for only 73% of operational projects; withdrawal-year data available for only 63% of withdrawn capacity (so the time-series undercounts total withdrawn capacity).
• The report is explicitly descriptive — "neither directly comments on nor recommends any specific policies."
• Large-load (datacenter) interconnection queues are a separate, un-covered dataset — this report cannot be used directly for datacenter-specific power-availability timelines.

Why It Matters (lens: energy — grid constraints)

This is the single most authoritative annual dataset on whether the US grid can physically keep pace with new demand (LENS.md Required Number: "US interconnection queue backlog (GW) | Annual LBNL Queued Up | LBNL"). The 86% gas surge alongside renewable declines is the clearest empirical confirmation of the AI-power thesis's core claim — that firm, dispatchable capacity (gas now, nuclear later) is winning the queue race against intermittent renewables, even as overall queue volume shrinks. The lengthening request-to-operation timeline (61 months and rising) is the hard evidence behind "when does my datacenter get power" — the central question this LENS names.

Full Content

See extracted findings above (source: 74-page slide-deck PDF, sections: High-Level Findings, Active Queue Requests, Operational & Withdrawn Requests, Duration Trends, Conclusions). Full dataset and interactive tools available at https://emp.lbl.gov/queues.

Source: Queued Up: 2026 Edition — Lawrence Berkeley National Laboratory / Interconnection.fyi, June 2026. Fetched via curl (declared User-Agent); WebFetch returned HTTP 403.