Because the foundations upon which AI shall be built aren't exotic. They're still programs. They still need an architecture to run on and they need electricity.
Even if new architectures that are specifically designed to leverage neural nets[1] or some other variety of processing are developed, the fundamental realities of computing, programs, physics and economics are not immune from being sufficiently underwhelming as far as AI is concerned.
Of course, this is precluding cloning technology where actual biological brains and therefore legitimate "life" and therefore intelligence can emerge.
I can't predict what could happen, but I would guess that once IA is more advanced the first thing it would do is to ask or take the computational power of the Internet to design systems that are able to evolve and grow. The ultimate goal of IA is to aim for better ways to decompose and construct programs to solve goal, to create a way for machine to communicate (that is to encode information and knowledge) the ultimate goal is to construct a world in which machines are rewarded for creating better programs, new ways of artificial conceptualization, unfortunately there is no place here for humans, except perhaps the analysis by machines of our brains trying to decode it looking for their god to provide sense to their existence.
