Makes sense- if 10nm is pushed out, 10nm equipment buys get pushed out too.
So here's the deal.
Intel has this neat 2 year (24-26 months) cadence, which painfully broke dowm recently because III-V is very hard, and even more importantly, because double, triple and at 10nm for Intel maybe even quadruple patterning is very hard. So they're now at something like 2.5 years for 14nm and presumably 10nm as well, even though they learned from 14.
So, obviously 10nm didn't happen 4 years after 22nm. If it had, then Intel would have started 10nm spending in Q4'14 (a year ago). Amazingly, they said at the investor meeting that you would see the front-end of the 10nm startup. Even more amazingly, it actually showed in the quarterly reports of Q4 and I think also Q1 (but not as much as what it had been for 14nm).
But after Q2, they announced the delay. So instead of 4 years after 22nm, 10nm will arrive 5 years after it, averaging 2.5 years for 14 and 10 (actually a bit more than that since IB was released in May).
And this is what we're seeing now. The start-up costs will start in Q4 (minus 1.0 point gross margin reconcillation), and will increase in H1, for (hopefully) production (qualification) in Q4'16 and HVM in H1 for high volume H2 release of 2017.