The Federal Energy Regulatory Commission voted 5-0 on June 18 to issue tailored show-cause orders to the six largest U.S. regional grid operators — PJM, MISO, SPP, CAISO, ISO-NE, and NYISO. The orders, issued under Section 206 of the Federal Power Act, give each operator 30 days to report spare generating capacity and 60 days to defend or revise the rules governing how data centers and other large loads connect to the transmission system. Texas's ERCOT, which operates outside federal jurisdiction, was not included.
FERC deliberately skipped the conventional Notice of Proposed Rulemaking path — a process that typically runs for years — in favor of customized orders aimed at each operator's specific queue and tariff. Chair Laura Swett framed the action as a national priority and tied it to the rapid growth of AI training and inference loads, which have begun to dominate new interconnection requests in most U.S. markets. The stated goal is to clear the backlog without shifting costs onto existing ratepayers or eroding reliability.
The action follows an October 2025 directive from Energy Secretary Chris Wright asking FERC to consider reforms for large-load interconnection. It also lands in a year in which Microsoft, Google, Meta, and Amazon are collectively committing roughly $725 billion to AI infrastructure, much of it gated on grid access. Hyperscalers have spent the last twelve months signing nuclear power-purchase agreements and exploring behind-the-meter generation specifically because public interconnection queues in regions like PJM stretch past 2028.
Takeaway for learners: the bottleneck for the next phase of AI is no longer just chips or capital — it is electricity and the regulatory paperwork that controls who gets to plug in. If you are studying AI, it is worth understanding that the systems you build sit on top of a physical stack: substations, transformers, transmission lines, and tariff filings. The same FERC orders that decide whether a Virginia data center comes online in 2027 or 2030 will quietly shape what models get trained and where.