That infrastructure is still useful, 20 years on.
It isn't about agility or guessing right, it's about piloting attractive technologies (eg, small-scale DSL which uses existing phonelines, which was oftentimes a reliability nightmare), and keeping an eye to the future.