AI Is Now a Grid-Planning Problem

The Federal Energy Regulatory Commission just did something that sounds technical, but matters for the future of AI.

FERC ordered the six major regional grid operators under its jurisdiction to explain, or rewrite, their rules for connecting very large new electricity users — including AI data centers — to the transmission grid. In plain English: the federal grid regulator is asking whether the electric system can handle the next wave of AI-scale power demand, and how to do it without sticking ordinary ratepayers with the bill. (FERC)

That is not a small question. A major AI data center can use electricity at the scale of a town or small city. If AI is going to be used broadly — in businesses, public agencies, schools, health care, research, manufacturing, and daily life — the country will need more compute. More compute means more power.

FERC’s order is best understood as governance trying to catch up with the speed of AI.

The agency appears to be trying to accomplish several things at once.

First, it wants to speed up “time to power” for large users, so AI infrastructure and advanced manufacturing are not trapped in grid-connection processes built for slower demand growth.

Second, it wants to protect existing customers. If a data center requires new transmission upgrades, the public question is who pays. FERC is pushing grid operators to address cost shifting and make transmission costs more transparent. (FERC fact sheet)

Third, it wants rules for a changing power landscape. Data centers may bring, contract for, or locate near their own power generation. The grid needs rules for those arrangements, rather than relying on one-off negotiations.

This does not mean FERC is approving data centers everywhere. It is not replacing local land-use review, state environmental authority, or local debates over water, noise, emissions, and siting. Those fights still matter. In fact, better grid rules can support state and local decision-making by clarifying whether power is available and whether existing customers are being protected.

For The Race, the signal is not that the grid is being captured by AI. The better signal is that AI is becoming important enough that public infrastructure institutions have to respond.

That is what enabling bureaucracy looks like: technical, boring, rule-bound, and essential.

The question now is whether these new grid rules can help AI scale under public terms — with adequate power, fair cost allocation, and continued state and local authority over the places where data centers land.