Upon export all frames of animation in the sprite will combine into files that are the size of power of 2 (8,16,32,64,128,256,512,1024) in size.
It efficiently saves them based on how much space your animation frames on a single sprite takes up. It's more efficient to load 2 64x64 sheets and 2 smaller sheets than it is to load 1 128x128, especially if majority of the space in the 128x128 would be transparent. Of course once the majority of 1 128x128 would be filled by sprites, upon export Construct 2 will use that size to store all the frames until it needs more space, then it'll make another file with the size of power of 2 and repeat the process until the next power of 2 size.
Downloading you download less since you are not downloading empty pixels, and memory wise you are not loading empty pixels as well. Graphically wise your card works more efficiently with power of 2 sized images as well.
Basically, C2 already efficiently handles sprite images upon export.