It is interesting to compare the UK experience, the U.S. experience, and the Canadian experience. The UK has structural separation with BT as a wholesaler and fairly strong separation between BT the wholesaler and BT the retailer. Canada has structural separation but weak separation between the wholesale and retail components of its two primary infrastructure providers (Bell Canada and Rogers Cable), the U.S. abolished structural separation and does not even maintain mandatory interconnection for broadband access. However, certain elements of the backhaul infrastructure are subject to regulated pricing and mandatory interconnection.
BT has apparently determined that if they are in wholesale business, they do better by maximizing product (capacity) and aligning per-bit prices with cost to maximize production. Essentially, having determined that it is not unduly expensive to keep up with increase in demand, they intend to encourage consumption and treat bandwidth as a commodity with traditional commodity pricing.
In Canada and the U.S., vertically integrated providers are demonstrating a desire to reduce cost by avoiding network upgrades. They can get away with this because they make their money on retail, not wholesale. This creates a very different incentive structure -- especially in a concentrated market.
Consumers, we should note, consistently hate metered pricing and capacity caps with two exceptions: 1) where the capacity cap is so high that it has little practical impact, or 2) where the ability to buy metered/capped plans results in significant price savings. The later tend to appear in places with significant retail competition.
In Canada, Bell Canada and Rogers sought to enlist the regulator as a means of mandating caps on the entire industry by allowing Bell Canada and Rogers to impose caps as part of their wholesale tariffs. It was not possible for Bell Canada retail to compete with caps against uncapped competitors buying uncapped wholesale access. This triggered massive consumer backlash.
In the U.S. we are seeing the gradual phase in of capacity caps at levels purportedly designed to impact only the top tier of users. This has tended to limit consumer outrage, although there appears to be sufficient competition in the U.S. market that some providers (notably Sprint in wireless and VZ Fios for wireline) use their unlimited service as a selling point.
Need more data and more opportunity to develop this. But intuition says that the evolution of capacity caps is a function of market structure and not a function of raw materials cost.
UPDATE: This Ars Technica piece is the best thing I've seen yet trying to break out the competing claims about cost and bandwidth congestion.