Assuming that they can achieve the success rates they've been aiming for, demonstrating lowered risk and cost through their own deployments will provide about the best advertisement you could possibly imagine. Waiting for organic outside demand to fill the launch roster to the same point would take time that Musk clearly does not want to waste. It seems to be a Keynesian approach to the slightly Chicken/Egg problem of commercial space.
Edit: There are ≈80 missions listed on the link there, 200 launches for their own missions would provide ample opportunity to derisk their customers by reusing launch vehicles beyond their currently marketable mileage and proving their safety (I can imagine that one would want some pretty steep discounts for a rocket that's been used 5 times given current success rates).
Consistent small profit launches, then as the accident %'s decrease design an ultra low risk human module and start space tourism for real
Keynes' ideas have been adapted to a range of different forms, but are more about circumstances where the economy is in a deep recession and there is some restriction preventing the economy from utilising all its resources. Here debt financed spending may have demand-side benefits by stimulating yet more demand. This only works at the aggregate economy level, so Elon would have to launch a lot of satellites to benefit from it!
To be fair he's talking about tripling the number of satellites in service with his constellation alone. Within the industry spacex would become the primary supplier and the primary consumer of rocket launches.
I hadn't heard of Arrows learning by doing but that makes sense as well for this discussion.