Image credit: Intel.
A Seeking Alpha contributor, who goes by the handle Motek Moyen, recently argued that microprocessor giant Intel (INTC 0.61%) should cancel its upcoming processor family codenamed Kaby Lake, and "skip straight to CannonLake [sic]."
By way of background, Kaby Lake is Intel's third (and final) microprocessor family built on the company's 14-nanometer chip-manufacturing technology. The product line was added to Intel's product road map as a result of manufacturing challenges that the company is facing in ramping up its next-generation 10-nanometer manufacturing technology, which delayed Intel's first 10-nanometer product known as Cannon Lake.
The crux of Motek Moyen's argument is that, because PC sales are weak, Intel will -- for the good of its PC partners -- "withhold new processors to persuade people to buy whatever is available." Motek Moyen adds that Intel should "cancel the 14-nanometer Kaby Lake and Apollo Lake and wait 1-2 years before releasing the 10-nm Cannonlake processors," adding that this "grace period" would allow PC vendors to "dispose [of] all of their existing PC inventories."
Here's why that doesn't make sense.
Newer is better for everyone
The first thing to note is that Kaby Lake is what is known as "socket compatible" with Skylake. This means that, following some validation work, Kaby Lake chips will be able to pop into the same systems that the prior-generation Skylake chips were able to.
This means that PC vendors should be able to sell largely the same systems -- if they don't want to put in the effort to design new systems -- but would be able to market them as packing seventh-generation Core (Kaby Lake) processors rather than sixth-generation Core (Skylake) processors.
From a marketing/psychological standpoint alone, a user with, say, a third- or fourth-generation Core processor is much-more likely to be enticed by a system with a seventh-generation Core processor than a sixth-generation one -- no matter the magnitude of the technical improvements generation over generation.
Further, in contrast to what Motek Moyen says, Kaby Lake is expected to bring a number of enhancements to the table, particularly in media capability, bringing support for HEVC Main10 decode/encode support, as well as hardware VP9 decode/encode support. I also expect that, as a result of fine tuning/tweaking of the underlying 14-nanometer manufacturing technology, Intel will be able to squeeze out greater frequencies -- and thus performance -- from Kaby Lake's CPU and 3D graphics engine relative to Skylake.
Kaby Lake should enable better user experiences than Skylake, so Intel would be remiss to simply cancel the product line and force PC vendors to sell the same Skylake systems for another year or so.
What about the inventory situation?
Motek Moyen cites a DIGITIMES report that claims that Intel is postponing the launch of its Kaby Lake processors as a result of weak PC demand driving "serious inventory issues" as a reason to cancel Kaby Lake. This, too, doesn't make sense.
Intel has already made it clear that it expects to generate around $57 billion in revenue, the majority of which will come from sales of PC processors. If PC makers have so much inventory of fifth- and sixth-generation Core processors (as Motek Moyen claims), then Intel's PC customers should simply stop ordering processors from Intel altogether.
The fact that Intel isn't projecting that its PC processor sales will plummet to zero during the second half of 2016 means that it still plans to ship processors to its customers. In that case, it would make sense for Intel to want to ship its customers its freshest processors rather than older chips with fewer features, so that the PCs that its customers build are ultimately more attractive to said customers.
Circling back to the whole reason Intel introduced Kaby Lake
The impression that I got when Intel first announced the existence of Kaby Lake is that Intel was responding to customer -- that is, PC-maker -- demand for annual product refreshes. CEO Brian Krzanich indicated as such, noting on a conference call last year that customers were telling Intel that "we really want you to be predictable."
Intel had tried to provide customers with such a release cadence under its "tick-tock" methodology, but as migrating to new chip-manufacturing technologies has gotten more difficult, Intel was no longer able to sustain a yearly product introduction cadence. By doing three generations of product on a given manufacturing technology, Intel can make sure that PC vendors have a new generation of processors to power a new generation of PCs each year.
