I don't believe "too little, too late". I meant AMD was nearly going bankrupt just a few years ago.There is some good news, the second generation discrete graphics cards are getting rave reviews. But it's probably still true that this is too little, too late.
The US has a preference for capital-lite businesses. That's the route AMD took and it was successful. But a second AMD would just be a second AMD. So I think that approach is dead, as a standalone. Could Intel X86 become part of AMD X86? Would that pass through antitrust scrutiny?
As far as the fabs go, I think this is the saddest part of the story. There are basic incompatibilities among the facilities that would prevent Intel fabs running TSMC process successfully. The failure of Samsung 14nm at GF is the illustration of this. The facility is where a lot of the secret sauce is. So the fabs would just be junked and that makes the economics of merging with a larger, healthier player impossible.
It is similar to playing a state-based game. Given the current state, which action should Intel choose?
If there is any activity between TSMC and Intel, I don't think it is about economics but about economics with external factors considered.