If AMD screws up with GCN 2, they are dead.
That's just stating the obvious. If it turns out they can't compete, then they are in dyer straights for sure. There's nothing to support that belief though. AMD and nVidia have competed back and forth all along. GCN offered better perf/$. Hawaii added better perf/mm. Pitcairn offered better perf/W. There is no reason to assume they won't continue competing. GCN1 competed fine against nVidia and in all likelihood so will GCN2.
Also, keep in mind that GPU's are used for more than VGA. In many compute functions GCN is still ahead of Maxwell. Nobody is worried though that nVidia will die in the compute market because of it. That's something that is reserved only for AMD if they fall behind in any metric.
I'm sure you've noticed that anytime there is a metric that nVidia holds an edge in we have people acting like it's the only metric that matters and AMD no longer is capable of competing with nVidia because of it. Any metric that AMD has an advantage in is of no consequence to these same people.
AMD was first to DX11 and it didn't matter because there weren't many DX11 games. nVidia is used to demonstrate DX12 though and it means doom and gloom for AMD. It doesn't matter that DX12 was still reportedly 1.5 years away from there being even a single game to use it.
Where AMD is concerned there are people who make mountains out of molehills. Every time Freesync is mentioned you get people just jumping in all over it as vaporware etc. Did you see anyone start a single thread because Gsync was delayed to simply create negative drama surrounding the nVidia brand? No. If it was AMD I promise you we would have had negativity to no end.
You can likely do a for and against bullet point presentation for each brands advantages (real advantages not these subjective advantages people throw out there) and it would be a fairly even back and forth.