Earlier quoted context omitted.
Here's the problem: That $2000 is per house. But its not about the people living in the house, it's about the physical premises being connected to the wider network, just divided up on a "per house" level. That's a piece of infrastructure which will exist 50 years from now. In fact, if technology is any guide, even with upgrades it'll likely exist 200 years from now even if the light going through it is used for very…
(edits: clarity) I divided $43 billion [1] by a population of 22 million, although now I see now that $1,500 would have been better (I read the wrong cost :[). Anyway, it was meant to be a per person, not per house, figure. The NBN would be a good piece of long-term infrastructure, no argument there. It is better than a cash handout. I don't think it is sensible to divide up-front cost by years of service. $100/year…
It's saving $20 billion, but still spending $20 billion and possibly more. Spending 2/3rd's the cost, with no upgrade path (FTTN does not upgrade to FTTP easily) is a terrible deal. We're keeping power and maintenance costs of the old copper network, you're splitting the type of plant you need to keep (since you have fiber, copper and coax all over the place), and you still haven't escaped the need to trench cable (since a huge amount of that copper needs to come out of the ducts to ever support 25mbit ADSL2, or needs to be completely rerouted to get the distance to the exchange to under 1.5km).
One huge benefit of fiber is that we don't need to worry about distances from exchange to premise - as long as it's under about 60km it'll work.