Have you seen the article that says that AMD won't release 20nm parts this year at all, and most(if not all) of Nvidia's cards coming in Q4 2014/Q1 2015 are still 28nm? Qualcomm is saying their first 20nm chip is Q1 2015 as well.
Q2 2012 - 28nm
Q4 2014/Q1 2015 - 20nm
27-30 months
The 14nm is effectively a half node. Do you believe that? Then as a half node should, it would arrive at half the time of a full node. Making it 13-15 months. That makes it...
Q4 2014 + 13 months = Q4 2016 to, Q1 2015 + 15 months = Q2 2016 for 14nm.
Do some people believe its much easier? Sure they can execute well, but its counterbalanced by the fact that its a complicated FinFET process.
Intel's first 14nm part is coming in Q4 2014. Even if Broxton is Q4 2014/Q1 2015, that is a lead of 3-6 months at least. Not to mention Intel's 14nm is better. Did I tell you the most important reason? Broxton is a second 14nm part.
Ok, that doesn't make it a "hey lets get rich in 5 years stock". But company-wise, if they are doing as they claim(40 million + sales + no contra revenue next year + PC stabilizing), I bet they'll stroll along fine.
IMO, the half-node-every-year approach is a dangerous game. There are overhead to everything, and even if ideally half-node has half the overhead of a full node, it usually isn't the case. The reason is simple. Because it takes engineers some time to learn something(overhead). 2 year cycle is alright because the process guys have some time to learn the process(first year) to perfect it(second year). 1 year means they'll have to learn it every year, and even assuming they won't have catastrophic issues(like terrible yields), they'll lose bit of advantage every year doing that. I know for a fact that stability of a "platform" is important looking at how consoles fare against state of the art PCs. Optimization gives developers opportunity to extract much out of the platform. Of course there are balances to be made...
Conclusion? Half-node-per-year will lose little by little against full-node-per-2 years, the cumulative effect which could be devastating in the long term. I have speculated for some time that its part of the reason AMD lost the "process war". They kept changing the process in the middle of a supposedly product lifecycle(of course Intel does it too but not as much, because they reserve it for the process bump).
As long as Intel executes of course. :) But if I know this industry well, its that executing original plans are enough to kill competitors.
News 
Market Data 
Discover 
