So far the tech press pretty much unanimously touts advances in LTE Carrier Aggregation as a way to further increase data rates for customers. With 2×20 MHz Carrier Aggregation a 2×2 MIMO network can be pushed to theoretical peak data rates of 300 MBit/s. 3CA pushes the limit to a theoretical 450 MBit/s and 4×4 MIMO and 256 QAM modulation pushes the limit even further. From my point of view, however, the main benefit of Carrier Aggregation lies elsewhere.
While it is true that mobile devices are getting faster and faster by using Carrier Aggregation the real value of the technology for me is that as networks attract more and more users and usage, they naturally fill up. The main way of compensating for the rising use is to increase the bandwidth of a network. Initial rollouts were done in most countries with either 10 MHz in the US and often with 20 MHz carriers in Europe. In the meantime, however, many network operators are using more spectrum to stun with high data rates for their glossy advertisement on the one hand but also to increase total capacity of the network to cope with rising demand.
An alternative way of using different chunks of spectrum is for the network to intelligently distribute mobile devices across the chunks of spectrum currently in use. There are several downsides to this, however. One is that usage is somewhat unpredictable and a chunk of spectrum that is little used at one point in time can suddenly be flooded by bulk traffic to or from one device. Other devices currently using that chunk of spectrum would have to be instructed to another chunk of spectrum.
Another disadvantage of distributing users is that the chunks of spectrum used by network operators are in different frequency bands. Lower bands typically offer better in-house coverage while bands in the higher part of the available spectrum typically offer more capacity. If users were distributed among the different bands it would mean that they would have to be frequently shuffled around to account for changing signal levels.
So instead of managing all of this I think it is much more clever to let mobile devices just use all of the available spectrum, if they can. When the majority of devices can use several chunks of spectrum simultaneously the network can make intelligent decisions in which chunks of the spectrum it wants to transmit data for individual devices without shuffling them around. If a user has bad signal conditions for some time his data can be transmitted on a chunk of 800 MHz spectrum. A few seconds later, perhaps, after he has turned around a corner and signal levels have improved, spectrum in the 2600 MHz band can be used to transmit his data. Obviously such a kind of scheduling is no trivial matter either but it looks much more flexible to me than to let mobiles hop between chunks of spectrum.