Initially assume everyone is spread out uniformly, and that we need a broadband connection for every three people, and that 100 Mbps counts as broadband.

If all connected were in use at the same time downloading at 100 Mbps we need an aggregate bandwidth of 11000 Tbps. Actual usage would be more bursty, so let's assume we can oversell by a factor of 100. That reduced the aggregate bandwidth need to 110 Tbps.

Wikipedia tells me the satellite with the most bandwidth currently are those in the ViaSat-3 constellation of geosynchronous satellites, each with a bandwidth of 1 Tbps. Let's assume we can get LEO satellites with that bandwidth.

We would need 110 to get the required aggregate bandwidth. However, since each satellite only spends roughly 1/8 of its time over the US we need to multiply the number of satellites by 8.

That brings is to 880 satellites, but remember, we were assuming a uniformly spread out population. We need to take into account the non-uniformity of the actual population.

To do that we need to know the ground area that a satellite can serve simultaneously. That depends on the height of the satellite.

Google is telling me Starlink's around around 350 miles up, so let's assume that is what we use too. A naive distance to horizon calculation says that a satellite at that height would have an horizon around 2500 miles away.

I'm not sure if that means it could actually broadcast to people 2500 miles away from whatever its spot is over or if there would be problems due to the shallow angle the signal would have to travel to the receiver.

If each satellite can effectively broadcast to such a large circle then most of the 110 visible at a given time from the US would be visible over a wide enough area that how the population is distributed might not matter much.

This is assuming that the satellites do not interfere with each other. My guess is that there would be interference and they would have to operate more like cellular networks operate, with each satellite having a limited ground area that it broadcasts into using some kind of spot beam.

Feasibility now depends on the size of those spot beams and how accurately they can be aimed. If they can adjust the beams to cover large areas when they are serving a rural area and small areas when serving a dense area, and if they can aim that spot anywhere in that 2000+ mile radius circle where the satellite is visible then 880 satellites may be enough.

If, on the other hand, their beams are pretty much fixed in size and direction, then we would probably need a lot more satellites. Given the area of the US, for everyplace to have coverage the 110 satellites that would be over the US on average would have to each be responsible for a circle of around 200 miles in diameter, and could handle the needs of 3 million people.

If you've got more than 3 million people in a 200 mile circle you'll need more satellites over that circle.

Suppose we have an area that needs several satellites. It might be possible to arrange orbits so particular places always have a lot of satellites so that you could cover that area without having to simply make your constellation bigger.

I have no idea if that is possible. If not you may have to increase you constellation size so that you every place has the number of satellites that your highest density 200 mile diameter circle needs.

I don't think that there is any 200 mile diameter circle in the US that would need more that 20 satellites, which would mean we'd need 17600 satellites in the constellation.

Note that this is quite dependent on the circle size so could be way off from reality. Also I've not considered communications from the ground to the satellite which might also impose constraints.

Conclusion: it doesn't seem obviously impossible to provide 100 Mbps broadband via satellite to everyone in the US. A more detailed analysis is needed.