Ground stations and EPFD. (equivalent power flux density)
In order to hit those speeds, the ground stations (both customer and uplink) need to hit multiple birds for a single location. Which means that those satellites need to be pushing sat power to those locations. You can only push so much power to a given area before you violate FCC regulations for epfd AND potentially cause interference to another party. In SpaceXs case, that other party is Dish Network who they are trying to "share" spectrum in the Ku DBS range with. Dish network has published their analysis for the interference they expect from SpaceX and show it to be illegal. I don't think SpaceX has responded to their report other than to say they'll just use 1 satellite pointing at a spot at one time, which is just untrue, The beamwidths at Ka band frequencies are much narrower and the epfd limits are also much higher. They might get away with it at Ka band.
Starlinks biggest problem will continue to be environmental. When in LEO, it takes a lot of satellites to cover an area, LEO satellites fall out of orbit after a while, and weather causes massive rain fade at Ku and Ka frequencies, Ka is especially bad in weather and SpaceX needs the Ku space.
Cox cable has 400-500 customers on my service group, a few blocks. They have a node, a few amplifiers, and lets say we all pay $100/mo. $5000 a month from infrastructure they only periodically need to maintain and it works in all weather. If the node gets congested, they can split it. How many subs do you expect to have in flyover KS? 500 in a spotbeam? 50? It doesn't really matter, it could be 10. The cost to provide Gigabit service to those 10 households will be comparatively astronomical. At 256-APSK, 8 bits per symbol, 216 Mhz transponder, forget the FEC, you get to 1 Gbps. I might get that with a 5m Dish tracking the bird in the desert and I'm the only one of the transponder. Realistically, it's snowing and I'm getting a lot less, lets try 32 APSK, 5? bits per hz, for comparison I think Dish Network uses 8-PSK and thus 3 bits per hz. That transponder isn't going nearly as far with the modulations you're most likely to get away with consumer cpe. These aren't high gain dishes they are using, rather flat panel electronically steerable arrays, they will struggle to maintain sufficient SNR for high throughput in most weather environments. When it rains, my directional relatively high gain DirecTV dish in sunny Arizona, sometimes goes out. God forbid you live somewhere where it rains regularly.
I think OneWeb, who is not targetting the consumer directly has a better plan, which used active dishes to obtain high throughput and the local areas served distribute to end customers some other way. Cable modem, 5G, etc. This improves efficiency greatly and you might get Gb/s of transit, that will never happen with Starlink customers with poor modulation levels crapping up the channels.
Then there is the economic issue, Cox cable can profitably offer me service at $20/mo, they don't, but they could. Starlinks consumer start up costs will always be multiple hundred or even a thousand dollars.
I've seen a lot of NGSO constellations fail or go bankrupt, I think I might see another.