Well, lets see, if we've got a couple thousand of satellites, and we've got 510 million square miles of land area on Earth, that's an average of less than 250,000 square miles per satellites, or a 282 mile radius each. And of course there will be fair amount of overlap between satellites, since you're 750 miles up and you're going to have at least that ground radius with excellent line-of-sight and only moderate signal falloff. That will also mean most overseas satellites will have LOS with the coast, giving maximum effective overlap to the most densely-populated regions.
Granted, you're still talking maybe dozens of times the average coverage area as a typical cell tower, and about 10x the transmission distance, so receiving antennas would naively have to receive 10x the signals at 100x lower power levels. A challenge to be sure. But perhaps doable. You are after all almost 30x closer than a geosynchronous communications satellite, so you're dealing with almost 900x less signal attenuation than those have to deal with. Besides, they'd only be one provider, it's not like cell towers would suddenly disappear, there would just be another option with radically better coverage for those who want it.
And as long as you've got the satellites up there, there's no reason you couldn't use them to provide practically free coverage to the remotest areas of the world - after all they'd just be sitting idle in that part of each orbit otherwise. Same principle as selling movies/music/pharmaceuticals at radically reduced prices in the developing world - the sunk costs were already sunk for the primary market, and incremental costs are almost nonexistent, so you may as well make a little extra money. And if you can position that market as a humanitarian offering to boost your corporate image - so much the better.
You're not even lying. Hell, Africa is 3x larger than the US and far, far poorer. All the very real difficulties in providing good coverage and service that the cellular companies like to oversell here are far, far worse there. If you can instead completely bypass the need to build dedicated infrastructure by piggybacking on idle "first-world" infrastructure instead - well that's a game everyone can win at. Africa, India, China, the vast tracks of Russian, South-American and Canadian wilderness - lots of places in the world that it isn't worth building infrastructure - and the people who live there could benefit immensely from ultra-cheap internet - even at dial-up speeds. Why should the rich, densely-populated regions be the only ones whose populace can educate and organize without centralized control? We've already got the centralized systems in place - we're the ones who stand to benefit *least* from the decentralization so powerfully enabled by the internet.
