sw, hw encoder
For Encoder:
Main complexity is still in Motion Estimation part, but Intra prediction and mode decision is not simple at all. It includes chosing between Intra4x4 and Intra16x16 and 4 Chroma modes, and between all submodes (9 modes for each 4x4 block). Implemenation from reference SW is very bad starting point (it does cross searching between different Luma and Chroma modes, without any interdependences between them (same thing as calculat separatelly best Luma and the best Chroma mode).
Also, if design choise is high-complexity mode decision from reference SW (with some simplification, probably) between all diferent modes and submodes, complexity increases exponentially.
For decoder:
Inverse Intra prediction is very simple. There are some aditional complexity due to frame/field stuff, but still it is simple process.