Because hardware depreciates quickly, datacenter rentals are a competitive business with much lower profit margins than the IPO prospectus requires (even if there is a temporary bottleneck now) and there's no real moat.

The original batch was probably Musks "AI will solve everything, i have a small dick, i want to buy all the hardware and be the first" which became "Ah shit Grok doesn't need all the compute, we can't sell it properly our IPO is coming soon we need better numbers.

And

"Shit why did we agree to buy so much hardware if i can't even use the current one fully?"

to

"Ah fuck it, who cares if i indirectly pivot to selling this compute. It brings money and my Fanboys probably think its some magic smartness and not just ignorance"

Demis? Democrats or something else?

Elon likes money and power.

Keep in mind Google also rents GPUs via GCP, so they could be just reselling these to GCP customers?

Thing is though, Anthropic was really against the wall with lack of compute pre xAI deal. And tbh, Gemini reliability has been abysmal which probably points to real compute shortages.

And nearly _every_ major DC project is really up against it with massive delays, etc. Stargate UAE has been badly affected by the Iran conflict.

So maybe long term this isn't a great business, but _right now_ I'm not convinced it's all financial engineering. There is a enormous shortage of compute and xAI has a load of it _available now_.

Could you please describe the other satellite launch services whose prices are competitive with SpaceX?

> Which is to say: Grok basically failed as a frontier AI and they need to pivot to a business model which makes money.

They can just run Grok as a local AI inside Tesla cars. It's actually really efficient as a compute platform because the Tesla cars are in motion at highway speeds, which gives you lots of free airflow for shedding waste heat via the car radiator. Way more efficient than trying to run AI on space satellites.

Xai seems pointless, and they've got gpus needing to be used for something.

You make it sound like they’ve given up on Grok, which I don’t think is accurate. I think it’s been mentioned the Grok 5 1.5T model is currently training on Colossus 2. And their recent deal with cursor is part of being able to eventually compete with Anthropic for agentic coding.

Strongly agree with all of this, except that charging rent for the use of an asset you own is not what economic "rent-seeking" means. I blame the dumbass economists who named it this, forever polluting the discussion to be had about regulatory capture and legalized political bribery.