On Nov. 19, Intel (NASDAQ:INTC) hosted its annual investor meeting. At these annual events, the company usually provides substantial insight into its business, both from a high-level strategic perspective and -- particularly in the case of the presentation from the CFO -- the lower-level "nuts and bolts," so to speak.
This year, Intel's Bill Holt -- the individual responsible for the company's chip manufacturing technology development -- had a very interesting and informative presentation discussing the company's latest 14-nanometer chip manufacturing technology.
In a nutshell, the technology itself seems great, as you'll soon see, but it's clear the company is still facing yield issues with the technology. Without further ado, let's take a closer look at the content from this presentation.
A very thorough explanation of the density advantage of Intel's 14nm technology
Intel has claimed for quite a while that its 14-nanometer chip manufacturing technology offers better transistor density than competing 14-nanometer and 16-nanometer technologies from Samsung (NASDAQOTH:SSNLF) and TSMC (NYSE:TSM), respectively.
Higher transistor density means that, in theory, Intel can cram more functionality into a given area than its competitors can, or it can deliver similar products but at lower cost thanks to the smaller chip area.
This claim has actually been called into question by a number of folks in the press and the investment community. The doubt arises from the fact that Apple's (NASDAQ:AAPL) A8 and A9 processors, built on TSMC's 20-nanometer and 14/16-nanometer processes from both foundries, respectively, seem to be able to pack in more transistors into a smaller area than Intel's latest PC processors.
Holt conceded that a straight division of the number of transistors in its two lead 14-nanometer products -- Broadwell-U (2+2 configuration) and Skylake-U (2+2 configuration) -- by the chips' respective die sizes indicates that Intel's 14-nanometer technology is actually less dense than the foundry 14/16-nanometer processes (see the image below).
However, Holt pointed out that these comparisons aren't actually that informative as the average transistor density of a processor depends on the kinds of transistors/devices used to build the chip. Some devices are quite dense, while others can take up quite a lot of space.
Normalizing for the different compositions of these chips, Intel claims its 14-nanometer technology has a pretty significant edge in terms of density, consistent with the company's published gate pitch/metal pitch metrics:
In other words, if Intel's analysis is to be trusted (and I do trust it), the company's 14-nanometer technology delivers a pretty substantial advantage in terms of transistor density relative to the 14/16-nanometer technologies from its competitors.
That's great, but what about yields?
Although Intel's 14-nanometer technology looks great from a density perspective and performance and power of products in the marketplace looks quite good, Intel showed some data indicating that manufacturing yields on this technology are still behind where the company had expected them to be this time last year:
Intel had originally hoped it would be able to get 14-nanometer yields to match the yields of its lead 22-nanometer product by the second half of 2015:
Obviously, Intel hasn't improved its 14-nanometer yields at the rate it had originally hoped to.
The following slide shows how Intel expects these yield issues to manifest themselves in terms of product costs:
If you study this image closely, you'll notice that Intel is projecting significant increases in product costs for 2016 in the performance segment. Much of the volume Intel shipped in this segment during 2015 was on the 22-nanometer node, so the increase in 2016 seems to be attributable to the fact that Intel is transitioning these products will see a broad transition to products built on the 14-nanometer node.
In the mainstream segment, Intel has actually been shipping 14-nanometer products in volume for quite a while, particularly in laptops. This is why we saw a big jump in cost from 2014 to 2015, but as yields improve, costs should come down in 2016. The same phenomenon appears to be in play in the value segment (remember that Intel launched its first "value" 14-nanometer parts in 2015 in the form of Braswell and cut-down Broadwell models).
Additionally, during CFO Stacy Smith's presentation, the executive said that although the company's data center group is expected to grow sales at a "mid-teens" percentage rate but that operating profit will only grow "in the low double digits."
This, Smith indicated, is due to the company's transitioning its product line there from the 22-nanometer manufacturing technology to the 14-nanometer technology, which is expected to lead to increased product costs, negatively affecting gross profit margins and, ultimately, operating margins.
Intel's 14-nanometer is great technology, but the yield situation looks tough
It's hard to deny that Intel was able to achieve some very impressive area scaling with its 14-nanometer technology, and as somebody who owns multiple products built on Intel's 14-nanometer technology, I've seen firsthand how good products built on this technology can be. But it's also difficult to ignore the fact that the company has struggled with manufacturing yields, missing its own forecasts with respect to yield improvement.