To me, it seems obvious that ground based optical fibres will be cheaper than a space based system by an order of magnitude. So unless say, everyone starts sea-steading and a space based broadband is the only choice, there is no way a space broadband system can really compete with standard ones on earth.
I don't think anyone can give you a detailed answer on how they will make money, because their business plans are their own, and proprietary.
But the premise of the question seems essentially flawed -
There are huge areas of the world that do not enjoy reasonable (or any) broadband access. There are also existing markets for mobile broadband (passenger aircraft, trains, RVs, trucking, ships). There are scenarios requiring broadband access at remote locations where running fiber is not economical or practical (cell towers in rural areas, remote industrial operations like mining, scientific research stations).
Now Elon has also hinted that he believes their bandwidth and costs will allow them to compete with point to point fiber - linking datacenters for example. This is where it might be a bit harder to compete, but it's certainly plausible, especially for certain edge cases where geography and existing fiber runs means a terrestrial line would travel further than a LEO multi-satellite link would. They can, in essence, create virtual point to point links between any two points on Earth far faster than fiber could be run, and much more direct than using existing fibers usually would be.
1$\begingroup$ The idea of linking data-centers over satellite is pure moonshine. The latency issues that you talk about have largely been completely overcome IMHO. And he is not even talking about some specific application, like say, ocean mining, and getting internet connectivity for such mining operations. He is talking about launching a general infrastructure. Doesn't seem to square up at all. $\endgroup$ Oct 27, 2018 at 15:49
3$\begingroup$ @user2277550: I doubt it would be cheaper and faster to install fibers to remove areas like Antartica. There may be a lot of legal issues to solve, for example (who own the land the fibers have to run through). Maintaining a fleet of planes is indeed something that may be feasible in the not-so-distant future with solar-powered planes, but right now we don't have that. Satellites by contrast are a proven technology we know how to handle today. Also, the market for fast communication with ships alone may be interesting. $\endgroup$– DarkDustOct 27, 2018 at 16:26
2$\begingroup$ @user2277550 The latest trans-atlantic cable cost 300 million dollar - 30 StarLink are a single launch of a reusable rocket plus mass-produced satellites. That's likely about 100 million. Latency is not much of a problem - the extra of 800 km up and down don't matter much for connections around the globe. $\endgroup$– asdfexOct 27, 2018 at 16:32
2$\begingroup$ @DarkDust You can't judge StarLink on existing satellite internet latencies. StarLink satellites will live in LEO with estimated latency ~25ms. $\endgroup$ Oct 27, 2018 at 16:52
2$\begingroup$ @user2277550 I'm unclear how you feel running optical fiber to antenna spread over the ocean is cheaper and simpler than a large satellite constellation. Fiber runs on the sea floor are expensive enough (@asdfex gave an example of one that costs more than three Falcon flights), and placing antenna on the surface of the ocean with buoys and connecting to undersea cables with enough coverage to mimic what StarLink can do would be an absolutely absurd technical challenge and expense. $\endgroup$ Oct 27, 2018 at 16:54
One useful attribute of any LEO broadband system is agility, where you can deliver the service for the cost and shipping of a ground station. Which means anytime existing providers fail to keep up you have an opportunity to charge a premium until infra structure catches up with user demand. Failing to find useful numbers but the complaint level in Australia around the Broadband rollout suggests a reasonable but not huge market size here.
If the system can actually provide the claimed backhaul levels of capacity the most profitable market may actually be redundancy for existing providers for when lighting strike, lost backhoes etc destroy their connections (or they just want to do maintenance). From the business case perspective this is useful since most of the time you are being paid by a small number of tech savvy customers to not actually carry data which is a low overhead way to run a business.
$\begingroup$ Australia's "...existing providers fail to keep up..." You've phrased that in a nicer way than some have. $\endgroup$– uhohOct 28, 2018 at 2:17
1$\begingroup$ This answer was much longer before I removed the ranting. $\endgroup$ Oct 28, 2018 at 6:07
$\begingroup$ In my line of work we are often cursing unexpectedly long term outages of traditional fiber providers cutting routes between our diverse locations. Being able to ship a pair of terminals out and establish a virtual circuit at any time would be very beneficial, even if it cost a premium. $\endgroup$ Oct 28, 2018 at 16:23