In light of Intel's (INTC +4.36%) recent delay of products built on its upcoming 10-nanometer manufacturing technology, SemiWiki's Daniel Nenni speculates that the chip giant might ax its 10-nanometer product portfolio altogether and refocus those efforts on bringing out the 7-nanometer technology node while at the same time cutting costs.
This course of action seems highly unlikely. Here's why.
Intel probably won't lose its lead even with this 10-nanometer launch schedule
According to Intel, its first 10-nanometer products, code-named Cannonlake, should arrive in high volumes during the second half of 2017. This is around the time that one would reasonably expect 10-nanometer product from Intel's foundry competitors to hit the market.
Remember that Intel beat Samsung to shipping 14-nanometer product into the market by six months (and I would argue that Intel's "14-nanometer" is ahead of Samsung's "14-nanometer" technology) and it beat semiconductor foundry giant TSMC by about a year.
Although TSMC and Samsung claim that they will go into volume production on their respective 10-nanometer technologies by the end of 2016, I'm not convinced that if Intel is having such difficulties with its 10-nanometer technology that the foundries will have such an easy time getting there.
Skipping 10-nanometer would require a significant product pipeline flush
Intel engineers have been designing processors on the upcoming 10-nanometer technology for quite some time. These designs span everything from ultra-mobile applications processors to server processors . If Intel were to simply "skip" the 10-nanometer node, then it would have to flush years of product development down the proverbial drain.
Nenni suggests that in the interim period between its current 14-nanometer process and the 7-nanometer process, it could roll an improved 14-nanometer process technology. However, Intel would still need to bring its new architectures -- which the company has been designing on the 10-nanometer node -- to the 14-nanometer node.
Node-skipping would have significant negative economic implications
It's worth pointing out that one of the reasons that Intel focuses so much on trying to reduce the effective cost-per-transistor of its manufacturing technologies is so that it can add more functionality (which ultimately require more transistors) to its chips while keeping its cost structure on the chips intact.
If Intel were to put out an "enhanced" 14-nanometer technology, it might be able to deliver better performance through an improved transistor architecture, but it likely wouldn't be able to deliver a material chip area reduction.
Now, Intel is obviously bringing out a third 14-nanometer product family in 2016, code-named Kaby Lake, which will likely include a greater number of transistors than the 2015 Skylake parts. However, Intel will likely be limited in the incremental chip area it can dedicate to increased functionality in Kaby Lake as a result of the use of the same 14-nanometer manufacturing technology.
At any rate, if Intel can get its yields at 10-nanometers under control (which it should by the time Cannonlake launches), then going ahead with 10-nanometer product launches beginning in the second half of 2017 time frame makes sense.
What about the cost cutting that was mentioned?
Nenni also notes that skipping 10-nanometer and accelerating 7-nanometer would allow Intel reduce its operating expenses. Even if doing so would allow Intel to reduce expenses, Intel has not signaled that it plans to cut costs related to its chip manufacturing technology development.
In fact, on the company's Q4 of 2014 conference call, Intel CFO Stacy Smith said that the company was actually increasing its investments in manufacturing technology. This is a core capability for the company and as such I wouldn't expect the company to make material cuts here; if anything, I would expect increases here and reductions in less essential areas.
