It's really hard to estimate since we don't know their satellite costs, but $100 million per satellite (launched) is probably a good starting point. That ends up being $13k capital cost per customer. If they can borrow capital at 5% or less, that works out to ~$55/month.

But if they succeed with reusable rockets (and it sure looks like they will), they can probably drop the launch cost from $60m to $10m in the near term. If we assume the satellite still costs $40m, that still cuts the total cost in half to ~$27/month @ 5%. Maybe they can mass produce these satellites and get them to $10-20m each, which would reduce it to more like $15/month.

Again all these numbers are pretty rough/best case scenarios, but even if we multiply them by 10 we could be looking at 1 gbps connections for $150-500/month. That could be cheaper than fiber infrastructure in a lot of places and still immensely useful in applications where fiber isn't possible (in a car, plane, or boat). That's also assuming 1 TB/customer. In many cases you may just buy the bandwidth in smaller increments at $1-10/gigabyte which is a few orders of magnitude cheaper than current satellite internet and comparable with LTE cell coverage.

The interesting part is what building 4425 satellites does to the economics of making satellites. Right now, most satellites are one-offs, or at most made in a series of 10 or so, handmade to exact specifications for maximum reliability and built for 25+ year lifetimes. This fleet is meant to be mass-produced and built to last a much shorter time, 5-10 years at most. I am very interested in just how low can they drive the cost of making a single satellite.