US regulator orders PJM to set rules for AI data centers

This FERC order is one of the clearest signs yet that AI compute isn’t just a cloud or chip story; it’s a grid story. By forcing PJM to formalize how AI-heavy, behind‑the‑meter data centers interconnect with power plants, regulators are acknowledging that AI’s hunger for electricity is now large enough to threaten regional reliability and retail prices. In practice, the order creates a template other U.S. grid operators will likely copy, gradually standardizing how AI data centers are treated in capacity planning and tariff design.

For hyperscalers and model labs, this is a warning shot. The easy path of dropping gigawatt‑scale AI campuses next to power plants without fully internalizing grid impacts is closing. Future AGI-scale facilities will need more sophisticated contracting, demand-response capabilities, and perhaps on‑site generation or storage to win regulatory approval. That may modestly increase the cost and lead time of building frontier compute. However, clearer rules also reduce political risk: investors can model what it takes to secure grid access instead of betting on ad‑hoc exemptions. Over time, these rules are likely to tilt AI infrastructure toward regions with surplus low‑carbon power and responsive regulators, subtly reshaping the global map of where AGI-class clusters get built.