I have very little faith in the success of these devices, I have explained my position at length some 3 or four years ago on the SI bu$$ thread. In essence, the transition is a first order transition (unlike MRAM where it is a second order transition), namely, there is an enthalpy of transition (fusion/recrystalization) associated with the transition, that will limit the spead at which the transition can occur to the rate at which the enthalpy eluted can be withdrawn (it is easy to very rapidly "inject" the heat, tough to take it out). MRAM will have other problems associated with minimal size of a "collective" of spins to create a magnetic domain. I think that both system will not compete in the main stream high density very fast NAND, but they will find niche applications that they can fill better than Si based NAND.
Still short and hurting from last night on RIMM bop ($98.5 and just under $103), hopefully some of that enthusiasm will evaporate and let me extricate myself.