Mostly just marketing hype without numbers to back up the claim that Moore's Law is still kicking. Advancement through die-shrink is a doomed exercise but we do have other options. Personally I hope the next decade or two will bring significant innovations in per-transistor performance, which is an area we've basically ignored since making smaller and more parallel circuits was easier.
>> significant innovations in per-transistor performance
What kind of innovation ? anything concrete - i.e. on research/dev ?
Most of the current efforts resemble Itanium, with smarter compilers and dumber processors.