Intel (NASDAQ:INTC) is the leading vendor of server processors, with the overwhelming majority of market share. This position is richly deserved as the company, year-after-year, releases best-in-class products.
It's worth noting that 2015 was something of a "lost" year for Intel's mainstream server products. Had the company stuck to the ideal one-year refresh cycle, we would have seen the company launch its Broadwell-EP family of products in the September/October 2015 timeframe.
Instead, though, we saw those chips launch on March 31, 2016.
I don't believe that this delay was due to the company's product development being off track. Rather, it would seem that it was done in order to protect the gross/operating margins of the company's Data Center Group. Allow me to explain.
14-nanometer chips still more expensive to make than 22-nanometer chips
In an ideal world, Intel would have gone into mass production on its 14-nanometer node in late 2013/early 2014 at strong manufacturing yield rates. Then, after a year-and-a-half or so of manufacturing chips for notebooks and desktop PCs, it would begin manufacturing the harder-to-build server chips on a process that is very mature.
Unfortunately, the reality is that Intel didn't go into production on 14-nanometer chips until mid-2014, and even to this day yields are still not quite up to the levels that the prior generation 22-nanometer technology were at.
In fact, according to comments made at the company's 2015 investor meeting, 14-nanometer costs aren't expected to match/come below 22-nanometer costs until the second half of 2016.
Indeed, Intel CFO Stacy Smith said that the reason that the company's Data Center Group saw operating income drop faster than revenue sequentially was the ramp of 14-nanometer server products.
What if Intel had gone into production for a September/October launch?
To put it simply, if Intel had launched its 14-nanometer server chips last year, this would have had a materially negative impact on gross profit margins in its server chip business. Operating profits for the segment would have come down, negatively impacting the company's total operating profits for the year.
During 2015, Intel's Client Computing Group, its largest business by revenue, was already suffering from the increased costs associated with moving the its laptop product line from high-yielding 22-nanometer silicon to relatively poorly yielding 14-nanometer chips. Intel management probably figured that it didn't make sense to take the hit in both businesses.
Implications for 2017
The interesting thing here is that in 2017 the company should still be selling server chips built on its 14-nanometer technology (though based on an updated architecture). This means that Intel's Data Center Group should enjoy a tailwind associated with much improved 14-nanometer yields throughout 2017. With better gross profit margins, operating margin should move up here.
On the Client Computing Group side of things, Intel will probably spend much of the year selling 14-nanometer products. The improved 14-nanometer costs should, too, help gross profit margins (and ultimately operating margins) for at least the first half of the year.
Intel is expected to start shipping 10-nanometer product in the third quarter of 2017. Depending on how yields on those products look and the proportion of the company's overall product mix those chips are, this could serve to deduct from gross and ultimately operating margin.