22nm: Haswell June 2013
14nm: Broadwell: End of 2013/Start of 2014
14nm: Skylake: 2015
10nm: Skymont: 2016
28nm: Production has been increasing but is still insufficient to meet market demand. Substantially all of the production is coming from TSMC, which is a significant bottleneck.
20nm: Volume production in 2014/2015. TSMC has promised to ramp 20nm faster than 28nm but this appears to be nothing more than rhetoric as 20nm will be yet another level of complexity when trends have show each successive ARM die shrink to ramp significantly slower than the previous.
14/16nm: ARM foundries will pursue a half-step to FinFET by making it half FinFET and half 20nm. This clearly illustrates their lack of confidence in bringing a full step online in a reasonable time frame. There are so many unknowns with the half-step process, that any prediction of volume production cannot be reliably made.
1.) Intel will continue to have significant opportunities for high margin foundry work, as illustrated by the Altera deal.
2.) Intel's fabrication economics will continue to improve just as ARM's are degrading. ARMs 20nm planar and the half-step to FinFET both have inferior economics to the Intel nodes.
3.) Intel will reach 450mm wafer implementation well ahead of ARM, giving them significantly improved economic benefits.
Intel's lead in fabrication and manufacturing is increasing in virtually every area