AMD are probably doing it with their sockets because they're behind Intel though (so they can appeal to more markets to stay in the game).
I doubt AMD would do that if they were on par or better than Intel.

Doing such a move may damage PR, but perhaps if they can somehow justify something like PCI-e x16 not offering enough bandwidth, they could transition users to a proprietary connector.  In fact, there already is somewhat a reason - GPGPU stuff like CUDA/OpenCL's main weakness is the high latency between it and the CPU/RAM - it can be argued that this new connector would reduce latencies, and justified.