You can build out landlines far more easily than you can launch satellites; you just need the cable/fiber and some guys to dig a trench. The reason it hasn't happened is because it's not worth the financial investment without government subsidies (and this is why the federal government provides some subsidies for rural broadband buildouts).

Iridium tried to be Starlink a decade or two ago, and they ran into the same problem: the cost of servicing rural and remote areas far exceeds the money you would receive from rural customers without massive government subsidies.

Each satellite costs SpaceX about $250K to build and $500K to launch and services thousands of customers.

That same $750K might buy you a mile or so of trenching and cable. Many rural customers need multiple miles each.

If you want to make this about the economics...

All studies on this topic report that it costs less than $250k/mile in metro areas, which cost more due to the need to get permits, plan around utility lines and existing infrastructure, dig into asphalt/sidewalk, and then repair the damage.

In a rural area where you can just dig into dirt and not repair anything, $750k should get you all of a town and then some, and I'm assuming we're talking about one scattered out across tens of miles. You're primarily paying for labor, and it's possible for a small team to dig dozens of miles of utility trenches in a day with common construction equipment. Once you lay the cable, your expenses are essentially fully booked.

A single Starlink satellite, assuming maximum satellite bandwidth and minimal user bandwidth would service at most 2500 customers. At $750k/launch, assuming a useful life of 4 years (per Musk's Techcrunch interview from April 2020) and maximum customers, Spacelink would need to charge at least $6.25/month to each customer not accounting for operational or marketing costs just to break even. You would need to at least double the price to cover ongoing operational costs. Plus, Spacelink would need to pay other ISPs for interconnection agreements, which is another 1x cost there. Add in the costs of paying off R&D and other capitalized costs, and you're looking at a minimum of 5x the putative minimum price of $6.25/month, or roughly $30/month.

Or in other words, Starlink would be charging no less than its current satellite and landline competitors already charge today.

Shotwell has said that Starlink will be competitive with $80/month internet.

It's a critical point in fact, because you're never going to capture an entire market, you're going to capture at best a modest fraction of the maximum market (for numerous obvious reasons).

Those 41 million potential customers are actually more like maybe 10-12 million stable or semi-stable households. The 41m is an incorrect number to go on, that represents the total number of people rather than the household count.

You're only going to get a fraction of those 10-12 million households no matter what you do. So now you're talking about probably at least 1m, up to maybe 4-6m households / accounts. Starlink can make a large dent in that.