Musk's real practical talent seems to be as a "scaling enthusiast". While I don't want to come across as a fanboy, I've been re-thinking my assumptions recently based on his perspective of what the real costs of things should be when manufacturing, testing and QC is mostly automated and I can't find fault with a lot of his calculations.

If they can put up 20 sats per launch, that's a cost of $75m per launch. With first stage re-usability each launch might cost ~$35m, leaving $2m to pay for each of those 20 satellites. This isn't taking into account the cost of ground stations and operational costs such as personnel, but it doesn't seem completely outlandish.

Presumably they've thought this through before announcing it.

Elon Musk != Donald Trump. When he makes statements/estimations, even if they prove significantly conservative, they are truly grounded in verifiable facts.

I certainly would not call Musk's statements conservative. He does ultimately deliver, but seldom on schedule.

People have been trying this for decades without success. McCaw and Gates tried it in the 90s with Teledesic, for example. Iridium also tried it without success. IIRC, Teledesic's early estimates 20 years ago were ~$10B.

But hasn't satellite technology changed significantly in the interim since 90's with much smaller payloads i.e CubeSats and FemtoSats. These should theoretically result in lower deployment costs:

small sattelites: http://news.mit.edu/2016/better-views-smaller-satellites-071... http://arstechnica.com/science/2016/04/the-next-big-thing-in...

Sure, but smaller payloads == smaller capabilities. Telecommunications satellites need massive switching capabilities and also massive power. They also need large antennas to service particular areas and they need propellant to maintain their orbits. How many CubeSats would be required to stream HD video to many thousands of people, for example? Not saying it can't be done, just saying that it isn't an easy problem to solve and there has been a lot of money already spent trying to solve it prior to Musk's announcement.