I think it is naive to assume that just because there is an
over-supply of bandwidth today that it will be that way
forever.
Some assume, some do something to make sure it's going
to be like that 
Who knows when this will happen. With every dial-up customer
moving to DSL, cable and wireless, and stuff like voice and
video moving to IP-based networks, seems like it'll happen
sooner rather than later.
I know at least one company which is going to be in
field trials this year with the technology which can
actually handle traffic in case if every household in
US gets a T-1.
At 1000% annual growth rates
(according to Lord Sidgemore) in bandwidth usage,
1000%/yr sounds like BS. 200-400% is more like it.
Reality check - see how backbones grow, not what
marketing people say.
it'll be only a matter of a few years, even with
yet-to-be-deployed and yet-to-be-developed DWDM
enhancements.
Never underestimate bandwidth of 1 sq ft. bundle of
fibers. Even if no advances are made in WDM or
femtosecond optical valve technology the petabit-per-second
networks can still be built for cost comparable to
the cost of POTS.
--vadim
PS Trend-wise speaking, backbones are catching
  up with host performance. Before long, hosts
  will become the bottleneck, not the network.
  After that, network growth will follow the
  same trend line as hosts: i.e. Moore's law
  performance increases with pretty much fixed
  cost.