New York's 12-gigawatt interconnection queue carries a 77% historical dropout rate
NYISO's large-load queue has grown twelvefold since 2022, but attrition history and a tepid regional demand outlook complicate the bullish read.
NYISO imported 52 gigawatt-hours of electricity from Canada on Friday (2026-07-03) — the largest single-day cross-border flow since January 2025 — with some of that power traveling along the newly commissioned Champlain Hudson Power Express line, according to EIA data published on Monday (2026-07-20).6
The milestone arrived as New York's large-load interconnection queue had grown from six projects totaling roughly 1,045 megawatts in 2022 to 48 proposals covering approximately 12 gigawatts as of December, a nearly twelvefold increase in proposed capacity driven largely by data center and AI infrastructure interest. Most market participants have taken that queue expansion as direct evidence of surging demand. The arithmetic is more complicated.
From 2000 through 2019, only 13% of the capacity that submitted interconnection requests across the US had reached commercial operations by the end of 2024; 77% had been withdrawn, according to reporting in QZ published in May 2026 (2026-05-19). Interconnection filings are cheap to enter and expensive to complete. Most projects that enter a queue encounter transmission upgrade costs and permitting timelines that erode their economics long before construction begins.1
That historical pattern has not obviously changed with the current cycle. Grid Strategies projected the US data center market would add at least 65 gigawatts of demand by 2029, and potentially as much as 90 gigawatts, according to analysis published in May 2026 (2026-05-19). The filing activity across multiple queues reflects that expectation. But history suggests most of it will not complete.2
New York's regional demand outlook provides a second complication. ICF, in a report released in June 2026 (2026-06-26), projected NYISO demand growth of just 14% from current levels through 2035. PJM, which covers the Mid-Atlantic and parts of the Midwest, is expected to grow 43% over the same period. The Southeast and New York could face supply constraints in the coming years, ICF analysts said, but the demand trajectory in New York is materially softer than in regions where the data-center build-out is most concentrated. Projects in the NYISO queue priced for aggressive load growth may not find the demand to justify them.5
The distinction between nameplate capacity and deliverable capacity adds a further layer. ICF expects roughly 445 gigawatts of new nameplate capacity to come online across the US from this year through 2030, but estimates only about 191 gigawatts will actually contribute to meeting peak load — given the capacity characteristics of solar and storage, which dominate the pipeline. A queue heavy in those technologies would not deliver the supply buffer that raw gigawatt figures imply.5
CHPE's commissioning changes New York's supply equation in ways the interconnection queue count does not capture. Friday's (2026-07-03) import record gave NYISO operators a direct alternative to local generation during high-demand periods. If Canadian hydro availability holds through summer stress events, the scarcity premiums some queue projects depend on for their investment case may not materialise with the consistency those projects require. Whether that cross-border flow can be replicated under tight grid conditions, rather than remaining a single-session outlier, will become clearer as late-summer peaks arrive.6
FERC's show-cause orders to all six RTOs and ISOs, issued in June 2026 (2026-06-18), directing each to justify or rewrite its large-load tariff under Section 206 of the Federal Power Act, add regulatory uncertainty to already long interconnection timelines. NERC's 2026 Summer Reliability Assessment, released in early June 2026 (2026-06-03), separately flagged that large computational loads present specific operational challenges; AI data center demand is harder to predict than traditional industrial load, and unexpected disconnections have already complicated grid planning.4,3
The demand case for the queue isn't fabricated. IEA data show global electricity demand from data centers grew 17% in 2025, with AI-focused consumption surging 50% in the same year. In the US, data centers now account for roughly half of all incremental demand growth, per IEA. Yet those national aggregates obscure the regional variation that ICF's forecasts are beginning to price in, and NYISO is not the geography where the demand shock is hitting hardest.1
The withdrawal rate in the NYISO large-load queue over the next 12 to 18 months — as projects confront FERC's revised tariff requirements and full interconnection study costs — is the figure that would confirm or falsify the contrarian read. A dropout rate approaching historical norms would mean the 12-gigawatt headline overstates committed supply by a wide margin. Sustained retention near half the queue would indicate this cycle genuinely differs. Friday's (2026-07-03) CHPE import record is a reminder that the supply side of New York's power balance has factors the queue count alone does not reflect.6,1,4