I've been having a pretty forthright debate over on the Forum Oxford discussion site, which ties in with a piece of analysis that Morgan Stanley did on mobile network capacity and broadband usage.
I haven't had a chance to go through the whole MS report with a fine-toothed comb, although I've had a glance at the section on mobile data growth.
My initial observation is that MS seems to be under-estimating short-term growth in mobile data consumption, and over-estimating network capacity. I'm also not convinced that the problem is always necessarily backhaul - the radio network is also creaking.
One thread of the ForOx debate was about "what's causing the congestion?" with lots of theories about video of different types, web browsing, downloads, uploads etc. Also asking about whether it's the uplink or downlink or signalling that's the problem.
I'm actually doing an awful lot of research into mechanisms for mitigating the impact of traffic growth on mobile networks. What are the solutions, and more interestingly, what are the business models around them. I've written a Use Case for the guys at Telco 2.0 on "Managed Offload", and there are various other things I'm pursuing here as well (contact me privately if you are interested in advisory work in this field).
But observation I'd make about this question of "what's soaking up our precious mobile capacity?" is that in many instances it's not apps that are causing the problems. It's concrete and brickwork. Moray Rumney from Agilent gave a great presentation last year which points out where the real mobile capacity increments will come from - WiFi and femtocells, primarily.
I'd also add in extra cost-savings from offload in the transport and core networks. I'm less convinced that traffic-management at the application layer, or things like compression of images/video have such important roles to play, although there will be exceptions.