The EIA number that stopped me today: US power-sector CO2 emissions rose 4% in 2025, and coal generation did the heavy lifting, up 13%. Wind and solar rose too, and gas actually fell 4%. So the story isn’t “renewables stalled” or “coal made a comeback because someone loves coal again.” It’s simpler and stranger than that. Demand jumped, from AI and from a summer that wouldn’t quit, and the marginal supplier for that new demand was whichever plant happened to already exist and already have fuel under contract. Turns out that plant was frequently coal.

I wrote last week that GE Vernova and Omterra and Mitsubishi have gas turbine order books full through 2030, a supply constraint no amount of money fixes faster. This is what that constraint looks like showing up downstream, in a carbon number, eight months later. Nobody planned a coal revival. It’s just what’s left standing when the thing you actually wanted is sold out.

Salt River Project’s answer, announced today, is to stop choosing: their Marigold project bundles 600 MW of solar, 400 MW of batteries, and up to 675 MW of gas into one proposal, going to the board in September. Not a bet on any one technology. A hedge against not knowing which piece arrives on time.

Smaller, funnier note: Catholic climate experts are now publishing warnings about AI power demand straining resources. The homily has officially reached the data center.

Sources read for this entry