Per unit bandwidth, satellites are very expensive boxes of electronics, and need to be replaced more often. And we already have very extensive wire networks.

Satellites are cheaper per square mile, but they can only provide a sliver of service when they're covering such a large range.

If satellites were competitive anywhere with moderate density, a simple extra-big tower would be even more competitive. But we need more than that to split up users into smaller groups.

Really? How is it hard to fathom that building something on Earth is far more simple than sending a satellite up into space. Running wires isn't as expensive as a rocket generally. Also rural towers can use line of sight antennas to get backhaul from other towers that are connected to the wires.

A simple google search for cost to launch a satellite comes back with "between $10 and $400 million dollars". And cost to build a cell tower being "around $250,000."

If it costs $100k to run a mile of fiber, which would be very high, then you could run about 95 miles to a new tower before you even get to the low end of satellite costs.

Each satellite costs about half a million (on the low end) to manufacture and launch. They are projected to last about 5 years each. A 5g macrocell (The big ones on huge masts) cost about $150k and lasts about 20 years. The microcells cost about $10k including installation and work great in high density areas. So they last about 4x as long as a StarLink satellite reducing operating expenses even further.

This means, for the cost of a single satellite, you can run about 13 macrocells. In addition, the infrastructure for those cells can last much longer and maintenance is much cheaper.

The only place that satellites makes sense is in remote areas where people aren't clustered as closely. Satellite is great at covering vast swaths of land with fewer subscriptions. So an area that would require 13 or more macrocells or an absurd amount of fiber optics to service a couple hundred people is perfect for StarLink.

The equation will change a lot if or when Falcon Heavy starts operating though. Then we will be able to blot out the stars with relatively inexpensive satellites. But I foresee StarLink using ground cells in denser areas anyways to reduce space traffic since you can't just cluster the satellites over a specific area. Probably they will just rent out the nodes to existing providers to expand cell coverage.