Here is an interesting article in the Financial Times on the economic limits of Moore's law. The article gives an interesting account of which type of fabs the various chip manufacturer have. It also states Intel's (my brand new employer) statement that only companies with about $9bn in annual revenue (give or take a dime) 'can afford to be in the business of building new fabs', that reduces the playing field quite a bit.
The question on when Moore's law is going to end has been ongoing for a while and multicore has given it a bit of a boost with 'simply' putting more cores in a package (as opposed to more transistors per core).
However, this can only go on as long as you can make things smaller and the lithographic process in making chips is one of the limiting factors as to how low you can go in size and hence how much content can go into a package, combined, of course, with how fast you can make the packages. Current lithographic processes use light in some form or another, but Wikipedia provides an overview as to what is in store next in Next-Generation Lithography.
I am anxiously awaiting to see what is next, disruptive change in the technology field often leads to cool things. And these cool things will eventually trickle down to the embedded field. The question of course is always how long it takes to trickle down. It further amplifies my feeling that live in a very interesting time in the embedded industry.
P.S. You can follow me on Twitter as markhermeling