For many years, microprocessor giant Intel (INTC 1.14%) has told investors that transitioning to newer chip manufacturing technologies provides significant product and economic benefits.
Chips are made up of transistors, and newer manufacturing technologies are usually expected to make these transistors smaller, and thus cheaper to manufacture. Equally important, chip companies can cram more of those smaller, cheaper transistors into a single chip without increasing total chip costs. This means faster, more feature-packed chips for the same prices as before.
However, smaller transistors are only cheaper when they can be manufactured at similar yield rates to the technologies that came before them. If yield rates from newer manufacturing technologies are substantially worse than prior-generation technologies, then the cost benefits from reducing transistor size are wiped out and potentially even reversed.
Various leaks coming out of Intel suggest that the chipmaker will keep producing a large portion of its processors during 2018 on its older 14-nanometer manufacturing technology rather than moving them to its newer 10-nanometer technology. I suspect that this is due to manufacturing yield issues with the 10-nanometer tech.
Let's dig a little bit deeper to get a sense of just how small a portion of Intel's overall product shipments 10-nanometer chips could account for in 2018.
Intel primarily manufactures the following types of processors:
- Desktop personal computer processors and related chipsets.
- Notebook personal computer processors and related chipsets.
- Data-center processors and related chipsets.
Market research company IDC says that in 2017, desktop personal computer shipments are expected to hit 99.2 million units. Notebook personal computer shipments are expected to hit 152.2 million units. In 2018, IDC says that the personal computer market will see "stable volume," so for the purposes of this analysis, let us assume that the 2018 desktop/notebook numbers are the same as the 2017 numbers.
Per Statista, server unit shipments are expected to hit 11.7 million in 2017 and 12.1 million in 2018. It's important to keep in mind, though, that this likely doesn't reflect how many data-centers chips Intel will sell, for the following reasons:
- Many servers contain more than one processor, and often even more than four.
- Top cloud service providers build their own servers, which might be difficult for third parties to track in terms of units.
So I expect Intel to ship significantly more than 12.1 million server processors in 2018.
Intel mainly sticks with 14-nanometer for 2018
Intel's 2018 processor release plans have largely been leaked at this point. Intel is planning to release a family of processors known as Coffee Lake-S for desktop personal computers. Coffee Lake is expected to be manufactured using Intel's older 14-nanometer manufacturing technology.
If we assume that Intel's notebook-to-desktop personal computer shipment ratio largely reflects that of the overall personal computer market, then that's already 40% of Intel's personal computer processor shipments that will stay on the 14-nanometer technology.
Intel breaks down its notebook processors into four major categories: N, Y, U, and H.
The N chips target very low-cost systems and are usually based on Intel's lower-performance Atom architecture.
The Y and U chips target thin-and-light notebooks and convertible machines, with the former targeting fan-less systems (think the 12-inch MacBook) while the latter can usually be found in higher-performance computers that require a fan to keep the chip cool (such as the 13-inch MacBook Pro).
Finally, there is the H-series, which is aimed at very high-performance notebook computers, such as the 15-inch MacBook Pro as well as dedicated gaming laptops.
In 2018, per a recent leak, Intel is planning to manufacture the N and H processors on its 14-nanometer technology. The Y processors will be built on the 10-nanometer technology exclusively, while the U processors (Intel's highest-volume tier) will be split between the 10-nanometer technology and 14-nanometer technology.
It is also likely that Intel's server processors stay mostly, if not entirely, on 14-nanometer technology throughout 2018 (though I would love to be wrong about that).
Coming up with an estimate
Intel doesn't break down its notebook processor shipments based on type. However, I estimate that about half of the notebook processors that Intel ships in 2018 will be manufactured in Intel's 10-nanometer technology; the other half should be built in Intel's 14-nanometer technology.
If that estimate reflects reality, then I'd expect around 70% of Intel's personal computer processor shipments during 2018 to be based on its 14-nanometer technology, with just 30% coming from its 10-nanometer technology.
I don't expect Intel to transition its server processors to the 10-nanometer technology until the second half of 2018 at the earliest (with early 2019 being a real possibility). In that case, even more than 70% of Intel's total processor shipment volumes in 2018 could be manufactured using its 14-nanometer technology.
I suspect that the 10-nanometer processors will dilute Intel's gross profit margins due to manufacturing yield challenges. Thus, I believe that with Intel's plan to keep the bulk of its 2018 shipment volume on its 14-nanometer technology (which should be very mature by then), Intel is trying to protect its profitability.
Intel seems to be betting that its 10-nanometer manufacturing will be in good technological and financial shape come early 2019, which is when I expect the chip maker to transition all major product lines to that technology.
If Intel doesn't have it together on its 10-nanometer technology by early 2019 (i.e. Intel guides to significant margin degradation from the transition), it would be a sign of serious execution issues on the part of its manufacturing group. In that case, I would expect that Intel would "clean house" within the manufacturing group, potentially installing all-new leadership in that division.