AI data centre grid interconnection + FERC + ISO

TOPIC HUB · AI DATA CENTRE GRID INTERCONNECTION

AI data centre grid interconnection

The reference on grid interconnection for AI-density data centres: FERC Order 2023 + ISO queue reform, generator vs load queue interaction, cluster study + fast-track processes, hyperscaler queue positioning, and the specific interconnection constraint that gates AI infrastructure buildout.

01Direct answer

AI data centre grid interconnection is the specific process by which AI-density facilities connect to the electric transmission or distribution grid. It is the pacing constraint for the AI buildout in most US markets. FERC Order 2023 (July 2023, implemented 2024-2025) restructured the generator interconnection queue process across the seven US ISOs (PJM, MISO, ERCOT, CAISO, NYISO, SPP, ISO-NE), moving from serial to cluster study, adding fast-track for smaller projects, and revising LGIA (Large Generator Interconnection Agreement) templates. Load-side interconnection (which is how most data centres connect) follows separate processes governed by the ISO + local utility, with materially different queue depths + timelines by region. Loudoun VA + Central OH + ERCOT are the most-constrained markets; adjacent markets are absorbing overflow demand.

02Key concepts

FERC
Federal Energy Regulatory Commission; US wholesale transmission regulator.
ISO/RTO
Independent System Operator / Regional Transmission Organisation; the seven US grid operators.
LGIA
Large Generator Interconnection Agreement; standard contract for grid-connecting generation.
Cluster study
Batch queue-processing methodology (Order 2023 mandated); vs prior serial approach.
Load queue
The demand-side interconnection queue for connecting large loads (data centres) to the grid.
Fast-track
Streamlined interconnection process for smaller projects under Order 2023.
BTM
Behind-the-meter; co-locates load with generation to avoid grid interconnection.
PJM / ERCOT / MISO / CAISO
The four largest US ISOs by AI-load impact.

04Deep-dive research

05Standards + regulation

07Frequently asked

Why is grid interconnection the AI infrastructure binding constraint?

Load-connection wait times in the most-demanded markets (Loudoun VA, Central OH, ERCOT) now exceed the physical construction timeline for the data centre itself. A hyperscaler can build a facility faster than the utility can extend transmission + provide firm capacity.

What did FERC Order 2023 change?

Cluster study replaces serial queue processing (faster in aggregate). Fast-track for smaller projects. LGIA reform. But it targets GENERATOR interconnection primarily; load-side interconnection is governed by separate ISO + utility processes that have not been comparably restructured.

How does BTM (behind-the-meter) power fit?

BTM co-locates load with generation to avoid the grid interconnection queue. Amazon/Talen at Susquehanna was the reference BTM attempt (blocked by FERC on ISA amendment). Post-rejection restructures exist but are more constrained. See Case Study 03.

Which ISOs have the worst load queue backlog?

PJM and ERCOT lead in absolute AI-load queue depth. CAISO has queue + local distribution constraints. MISO less constrained but still material. See Regulatory I for ISO-by-ISO comparison.

What can operators do to shorten interconnection timelines?

Site selection into less-constrained utility territories, early-stage engagement with the ISO + utility before land commitment, BTM structuring where regulator permits, and multi-year interconnection request placement (place early for future expansion capacity).

08Further reading

Definition AI data centre grid interconnection is the process by which hyperscaler campuses secure and energise medium and high voltage service from the utility. As of Aug 2026, interconnection queue times at FERC-regulated ISOs exceed multi-year windows in most territories, transformer lead times have crossed 200 weeks per industry disclosures, and substation build-out schedules are the binding constraint on gigawatt-scale AI infrastructure siting.