FERC pushes power grids to rewrite rules for AI data centers

What this is about

The Federal Energy Regulatory Commission issued orders on June 18, 2026 requiring six regional grid operators to justify or reform the rules for very large electricity customers. The target is not only data centers, but also big manufacturing facilities and other loads that no longer fit neatly into ordinary commercial connection queues.

This is not a routine AI product story. It is an infrastructure story with direct AI consequences: as model training and inference move into larger data centers, the bottleneck is not only GPUs. It is also grid connection, reserve generation, and who pays for new transmission.

What the FERC order actually does

FERC is using Section 206 of the Federal Power Act and issuing show-cause orders to PJM, MISO, SPP, CAISO, ISO-NE, and NYISO. Those operators have 60 days to explain why their current tariffs are sufficient for large loads, or to file concrete changes.

The five reform areas are specific: faster application and study processes, transparency around transmission costs, rules for co-location with on-site generation, new services for flexible large loads, and procedures for generation facilities that serve nearby or co-located loads. The operators must also file a report within 30 days explaining how enough generation will be available for existing and new large loads.

Why it matters

AI data centers compete not only for chips, land, and water, but for grid capacity. FERC describes the orders openly as part of the global AI race. At the same time, it says consumers must be protected from cost shifting.

That is the central tension. Faster interconnection can make new data centers and factories possible. But if costs are allocated badly, transmission upgrades, reserve generation, or congestion costs can end up on household and small-business power bills. The American Action Forum notes that the 30- and 60-day deadlines will show how operators try to balance AI demand with customer protection.

In plain language

Imagine a town where several huge bakeries suddenly want to connect to the water system. Each promises jobs and bread, but all need strong pressure in the pipes. The town has to decide who connects when, who pays for new pipes, and how ordinary homes still get water in the morning.

That is what the FERC order does for the power grid: it is not anti-data-center. It is aimed at unclear connection rules for customers so large that they can change the local system.

A practical example

An AI company plans a 600-megawatt data center in a region with long grid-connection queues. Today, the study process might take three years, and no one may be able to explain cleanly whether a new transmission upgrade benefits only the data center or other customers too. Under the new process, the grid operator would need to assess faster and more transparently: what line is needed, which costs belong to the project, which load can be reduced flexibly, and whether an on-site generator behind the meter actually helps.

For the operator, that can save months. For nearby customers, the key question is whether the bill stays with the project or spreads into general rates.

Scope and limits

First, this is not yet a finished national standard. The operators still have to respond, and the details may differ sharply by region.

Second, the order does not automatically solve the physical problem. Lines, transformers, power plants, and permits still take time.

Third, the boundary between federal and state authority remains sensitive. FERC says it is not intruding on state siting or retail-rate authority, but that is where many real cost and permitting conflicts appear.

SEO & GEO keywords

FERC, AI data centers, large load integration, power grid, PJM, MISO, CAISO, electricity tariffs, AI infrastructure, grid interconnection, data center energy, United States energy policy