Before I dive too deep into the point I want to make here about the challenges Intel faces, and it is worth calling out that my initial theory about Intel’s long-term position turned out to not play out. In 2016, I wrote this article titled Intel and the last foundry standing theory. My intro paragraph laid out what this theory was in a nutshell.
Coming out of the Intel Developer Forum (IDF), I thought I would share more on what I believe is Intel’s long-term thesis about themselves. Quite simply, I believe they feel they will be the last foundry standing with their leading-edge process technology. If you follow this as closely I do, you will note Intel and other foundries like TSMC and Samsung are in a race to get to the next process node. Today, these foundries manufacture at many different nodes, but 14 nanometers and 16 nanometers are the leading edge processes of today. With each jump to new process nodes (10nm being next and 7nm after that), designers can pack more processing power onto a single chip while still keeping a low power profile. The industry calls this “performance per watt” and the amount of performance increases with each new process while still maintaining a lower power voltage.
At the time of my writing of the article, Intel was not yet behind in process technology. And although they struggled to get to 14nm on the timetable they proposed, they were still the leading foundry. But moving from 14 to 10nm proved to be the challenge that sent Intel from leader to underdog. In the time it took Intel to get from 14nm to 10nm TSMC has successfully offered 16nm, 7nm, and 5nm process technologies. In the same time frame, Samsung has transitioned from 14nm, to 10nm, and currently 7nm EUV. Essentially, since 2016, each competing foundry has moved two full process generations ahead of Intel.
Intel should be fully moved to 10nm in 2021 but their process at that density has around 100 million transistors per millimeter square. Which is roughly the same amount of transistors per millimeter square as both Samsung and TSMC 7nm process.
This chart, which UBS created in a recent note on Intel, creates a compelling visual both in terms of process generation timelines and transistor density at each process.
This all ties into my initial thesis that Intel had always believed they would lead in process technology, and that is clearly no longer the case. That being said, the above chart highlights both the reason why Intel struggled for so long to get from 14nm to 10nm. The main reason always stated by Intel for the delay was because of their aggressive targets of transistor density for 10nm. As you can see, Intel’s 14nm had ~40m transistors per mm square, and both TSMC and Samsung’s 10nm had ~55 transistors per mm square. Intel’s target for 10nm was 100m transistors per mm square, which backs up the overly aggressive density target Intel was shooting for. Had they not been so aggressive, it is likely they would not be that far behind TSMC and Samsung.
Intel’s management and manufacturing group has been clear they will not be as aggressive on future notes, which bodes well for Intel executing on a more predictable timeline. But I do not believe they will catch up or surpass TSMC or Samsung.
One of the other elements of my theory was an acknowledgment of how difficult it would be and how much money in research, development, and patent innovation would be required to keep advancing manufacturing to new process technology. With that in mind, another chart I found in the aforementioned UBS report is a bit of a surprise.
The below chart maps the CapEx spend of Intel and TSMC over time.
One of the indicators most of us watched for was Intel’s CapEx spending as it related to buying manufacturing equipment in order to scale up for the next process development and manufacturing. As you can see, Intel had been spending the most in Capex and now is trailing TSMC, although both have been spending less than their peaks. While both are slowing slightly their annual CapEx spend, it is noticeable Intel is behind TSMC, which is telling in my opinion.
While the lower spend in CapEx could signal the less aggressive density targets or the long delays in process technology transition, at this point, Intel committing less and less to CapEx each year further convinces me that Intel will not catch up to other foundries, and further worries me that Intel can sustain their foundry business in the long-term.
Intel’s management, and the company as a whole, find themselves in a situation they have never been in before. They went from undisputed industry leader to now chasing the field as the underdog. In many ways, Intel was a dynasty that is facing a cross roads where decisions will need to be made that could dramatically alter its trajectory and some scenarios would leave us with an Intel so different from its roots for better or worse.
But the great thing about an underdog story is the chance of the upset. Intel does have a rolodex of great technology. They remain competitive still despite the shortcomings they have had the last five years. Process technology is not everything, but it still matters more than some companies want to admit. AMD has been gaining ground on Intel every quarter and in every category. Apple just released silicon for PCs that will likely set a new bar of expectations for notebooks and desktops in the years to come. In some ways, there has never been more excitement, or more ambition, or more validation for a computing architecture other than Intel’s standard of x86.
A well known venture capitalist Josh Wolf coined one of my favorite sayings “chips on shoulders put chips in pockets.” We can only hope that Intel has a massive chip on its shoulder with everyone counting them out. Because while we may not see Intel put chips in pockets, a competitive Intel is good for the industry and honestly good for the US as a country. It would be a terrible shame for no leading semiconductor company to be based in the US. Intel is down but they are still in the game and I do hope the adjustments they make at halftime will keep them competitive in the second half of the game.