For a while, the AI boom was discussed like a clean little stack of chips, servers, venture capital, and demos that made everyone ask whether their job was about to become a prompt. But the physical world has entered the chat. The next AI constraint is not only whether companies can buy enough GPUs. It is whether they can get enough reliable power, fast enough, without turning local utility planning into a full-contact sport.
That is why FERC’s latest action matters. The commission ordered six regional grid operators to submit plans for speeding large-load connections, including AI data centers, while making clear that big users should pay the full cost of the grid upgrades required to serve them. That second clause is the part ratepayers care about. If data centers need new substations, transmission upgrades, or generation capacity, the political question becomes: who pays, and who gets priority when everyone wants electricity at the same time?
The numbers are no longer theoretical. AP reports that more than 4,000 U.S. data centers are already operating and about 3,000 more are planned or under construction. The same report cites EPRI’s estimate that data centers now use about 5% of U.S. electricity and could triple their share by 2035. Globally, the IEA says over 2,500 GW of renewables, storage, and large-load projects are sitting in grid-connection queues, meaning the energy transition and the AI boom are now waiting in some of the same lines.
The smarter version of this story is not anti-AI. It is pro-grid realism. Data centers can become flexible energy assets if they shift nonurgent computing loads away from peak hours, procure clean power responsibly, invest in local grid upgrades, and design facilities that help rather than stress the system. The less smart version is a scramble where every company demands 24/7 power immediately and regulators discover, too late, that interconnection queues are not magic portals.
For Energy Brew readers, the key takeaway is simple: electricity access is becoming industrial advantage. The next AI winners may not just be the firms with the best model weights. They may be the firms with the best power contracts, the least congested grid territory, and the most credible answer when a regulator asks, “Nice chatbot. Who is paying for the substation?”
This also creates a new scoreboard for utilities and local governments. Regions that can connect large loads quickly without punishing households will attract capital. Regions that cannot will watch digital infrastructure migrate elsewhere. In that sense, interconnection reform is not boring paperwork. It is economic development with transformers attached.
SHARE THE BREW
If Energy Brew made you smarter before your second coffee, send it to someone who reads power markets for fun or pretends not to.
Share the brief and help build the daily energy intelligence habit.
