I know that this is an old thread, but I think that wa2ise has hit it on the head. The ampacity of a conductor is set by the combination of heat produced by the conductor, and heat which escapes through the insulation. Larger conductors have less ampacity per unit of copper cross section because for a given cross section there is less circumference through which the heat could escape.
If you want to see just how big a difference, compare the total ampacity of a single #2/0 conductor with 16 #14 conductors in parallel (separated for good heat dissipation). This is a approximately the same total amount of copper. If you could actually run the conductors at their thermal ampacity limit, you would get more than twice the current through the parallel #14s. (Note: this would not work or be safe for other reasons, and would violate a bunch of codes...just focusing on the thermal ampacity of an extreme example.)
Which leads to a thought about how to save some copper: build rope core cables where you use some sort of plastic filler material covered with copper and then covered with insulation. If you made something that had the external surface area of a #2/0 conductor, but the copper content of a #2 conductor, then my guess is that you would see a thermal ampacity perhaps 20% better than a #2 conductor, I'm guessing the equivalent of increasing the wire size by 1.5 gages.
-Jon
