Joseph Moore of Morgan Stanley
"We were surprised to see Coffee Lake, the 4th generation 14 nm part, added to the roadmap late last year, but expected it to complement 10 nm products. But Intel basically conceded that this will be their go-to-market product for Christmas, the 8th generation Intel processor family – though they did tell us that there will be client product out this year on 10 nm, the volume will be mostly next year. 4 generations of 14 nm is certainly a surprise, and if there is minimal volume of 10 nm this year, it’s really effectively not a 2017 process […] There are two reasons to put the 2017 focus on the 4th generation 14 nm part, instead of the 10 nm part, and neither of them are positive. The most likely rationale is that 10 nm volume is simply available too late in the year to hit the back to school/holiday market without disrupting the supply chain; In the 2014 launch of 14 nm, a late launch proved to be disruptive. The other explanation, which seems more what the company is describing, is that it’s simply more economic, given mature yields on 14 nm – but to us, this suggests that the performance and cost benefits of 10 nm aren’t playing out as quickly as expected. While a mature process yield would help margins in 2017, it certainly would extend the competitive window for AMD’s Zen, which looks to be competitive with Kaby Lake and potentially also Coffee Lake (which Intel claimed was 15% better performance than Kaby Lake)."