Intel's problems might go back all the way as early as the 22nm generation. On their presentations they say 8 quarters is the optimal time for revenue and design.
65nm - Q4 2005
45nm - Q4 2007
32nm - Q1 2010
They have kept that schedule until 32nm. Then on the next process...
22nm - Early Q2 2012
Even back with 32nm it wasn't supposed to arrive that early. 32nm was supposed to be Q2 2010. Then the 45nm mainstream parts faltered. 32nm went well and they were able to make 32nm mainstream chips in Q1. There were rumors that there might be a 3rd 32nm part at one point though.
This was also about the time when mobile wars were heating up. Intel started dedicating their resources to compete in it. 22nm was the official mobile optimized node. Maybe this loss in focus was the real problem. They basically shifted their focus from PC to mobile, which meant PC parts took the hit.
0.5 * ((Kaby Lake 2-core area) + (40/24)*(Gen 9 GT2 area)) + others
0.5 * (27.47 mm^2 + (40/24)*(54.94 mm^2)) + 16.27 mm^2
75.79 mm^2
Does this approximation of Cannon Lake 2+2 make sense?
We won't know until its released. And, we have to find a die shot too.