Another tidbit from their earnings, from CFO commentary.
Apparently in Q3 10nm startup costs were lower than they had expected (a 0.5pt decrease in gross margin versus 1.0pt expected). Usually when Intel lowers its startup costs forecasts, that's a bad sign. We've seen it with 10nm last year, which was lower than expected in Q1 and then again in Q2, when they decided to and announced the delay by a year.
So now it seems the big gross margin impact by 10nm will come in Q4. So to me it seems that the real volume of 10nm will come in 2018.
BTW, it's astonishing that Intel beat expectations by 200M and 900M versus original forecast, yet stock is down 6% because of flat guidance for Q4. They will have revenue of almost $59B, a decent increase, although this of course also includes about $2B from Altera.