The problem is that you're trying to relate manufacturing to deployment/integration, and they are really separate beasts.
How much do you think they needed to modify the production lines to manufacture different routers every year? They're still paying people roughly the same wages (with meager annual increases), they still have the same packaging costs and raw material costs, the only thing that makes a new device more expensive if that you are covering the R&D costs. It's not like 1.5TB drives cost more in terms of labor and materials than 250GB drives -- the price difference is to pay for the folks that figured out how to arrange the platter material and write methods to hit those densities.
Networks are a little more complicated in that you can't consider simple point solutions. The design and integration needs to look at the whole system, or you build yourself into problems like having two 4-lane highways try to merge into a single 2-lane highway.
The digital broadcast transition is another example of this. From a manufacturing standpoint we had this thing solved years ago -- the broadcasters have been pumping digital signals for the better part of this decade, and we've had hardware out on the market capable of receiving that signal for just as long. The logistics of deployment are the sticking point - you need everyone on the same page with compatible hardware or the system doesn't work the way it should.
In your AP example, you also have to be careful about what technologies you pick. You could have gone from 802.11b to 802.11a to get to 54mbps, but if you had made that move you would have stranded yourself from any backwards compatibility with 802.11b hardware due to the different frequency space. Not every upgrade path is a clear winner.