SpaceX is an AI company without a frontier model. Until Jan 2026 SpaceX was an aerospace company. Then xAI was merged into SpaceX on January 30, 2026, so SpaceX became an AI company less than 6 months ago.
Mainly it’s a REIT that bought a bunch of hardware which is now depreciating like fresh lettuce. They don’t even have a way to use it productively for AI purposes themselves and are instead renting it to competitors, who are using it to lap them.
Musk for Prime Minister is what I am hearing here.
Unless they manage to build such a model, then their plan looks like they intend to be an AI _infrastructure_ company. Which is probably viable and healthy business! One with lots of competition and low margins that is completely misaligned with their current valuation.
- offering to pay some nominal subsidy for all households electricity bills in the region.
- committing to only closed loop cooling
- if the location is amenable, then providing free heating to neighboring buildings.
- only building in blighted locations with no current economic value
Other random ideas: I wonder if anyone is considering building server racks directly into offshore windfarms. Replacing failed hardware is annoying, but surely more economical than in space where IIUC, there _is_ no plan to ever maintain hardware at all, just deorbit and replace. All the water is right there for cooling (though I'm sure saltwater has its own problems with corrosion). While the current US administration is clearly anti windmill, that might work nicely in Europe.