You dont need to worry about land, power, red tape or NIMBY s in space which are big problems. The sun is always shining in a sun synchronous orbit, and space is plentiful. People think these are going to be massive megawatt things as well, but they arent, they're going to be small satellites, so each one doesnt have as much heat to deal with as people think, and they know how to radiate heat away, they already do it on their satellites.

SpaceX also knows how to make relatively cheap satellites as demonstrated with Starlink where its a mass manufacturing line for them which brings costs down.

With starlink though people actually need the service and its worth paying for if you're in an area that needs it. Its hard to say the same about all this AI stuff, so even if they can do it, we still haven't seen that its profitable, that enough people want it, and that people wont eventually turn to a local AI if they do want it. So its possible, but its not solving a need as big as starlink is, so it could be an utter failure even if technically possible.

Personally I dont think its going to work because they wont get enough use to be profitable, not because the idea wont work or wouldn't work if enough people wanted it.

Edit: also SpaceX has a proven track record of being able to launch satellites at a high cadence, so its just a matter of will starship be rapidly reusable. Starship has to be rapidly reusable for this work. If it turns out its only reusable after heavy refurbishment, that really messes with the economics of everything.