Ax3.ai
Home / News / The Ax3 take
Utility DivePowerSeptember 3, 2026

FERC puts grid operators on the clock for large-load interconnection

Six show-cause orders. Sixty days. Federal regulators are forcing RTOs to answer for how gigawatt-class loads get connected — and co-location with behind-the-meter generation just moved from workaround to sanctioned path.

Following an October 2025 directive from the Department of Energy, FERC issued show-cause orders in June to six grid operators — CAISO, ISO New England, MISO, NYISO, PJM and SPP — requiring them to address how large loads such as AI datacenters connect to the grid. The orders demand answers on five fronts, including transparent cost allocation, transmission service for flexible loads, and, notably, co-location and behind-the-meter generation options.

Responses were due on a 60-day clock, and the docket (RM26-4) is now live — PJM floated its own large-load framework in August. FERC’s preliminary findings lean toward extending the co-location transmission services developed in PJM’s proceeding to more load types and more regions.

Translation: the regulatory system is being rebuilt around loads that bring their own answers — their own generation, their own flexibility, their own interconnection already in hand. What it is not doing is making the queue shorter for anyone arriving empty-handed. Multi-year waits for new large-load interconnection remain the base case.

The Ax3 take

What it means if you’re buying: Power in hand is the moat. A site that is already connected and flowing — or holds a transferable power contract — sidesteps the exact process regulators are still arguing about. Behind-the-meter and IPP-direct structures just gained institutional legitimacy.

What to do about it: This is the thesis behind the Ax3 powered-site book: ALTAIR 12 (12 MW connected and flowing, SW Kansas), PROCYON 24 (transferable power contract through 2027, Georgia), and SPICA 285 (285 MW IPP direct supply, North Dakota, scalable to GW). Megawatts today beat megawatts in 2029 — and the paperwork now agrees.

More from the Pulse

Related coverage

NVIDIA NewsroomNVIDIASept 3, 2026

Vera Rubin ramps to full production — first shipments begin this fall

NVIDIA says production shipments of the Vera Rubin platform start this autumn, with first allocation going to CoreWeave, Lambda, Oracle Cloud and Azure.

Industry analysesAISept 3, 2026

Hyperscaler AI capex barrels past $600B for 2026

Independent estimates now cluster between $600B and $690B of hyperscaler infrastructure spend this year — absorbing GPUs, transformers and grid queues at unprecedented scale.