For quite a while now, contract chip manufacturer Taiwan Semiconductor Mfg. Co. Ltd (TSM -0.83%) has been vocal to both investors and customers alike that it has a very robust pipeline of chip manufacturing technologies in the works.
At a glance, the cadence that TSMC has put out there seems almost too good to be true. After all, Moore's Law dictates that chip manufacturers should be able to transition to new technologies roughly once every two years.
TSMC is claiming, at least for the next couple of generations, that it will go into production on its 10-nanometer chip technology in late 2016 and its 7-nanometer technology during the first half of 2018 -- moving at a pace quicker than that dictated by Moore's Law.
And, the company recently disclosed, its 5-nanometer technology should go into production about two years after 7-nanometer, landing it sometime in the year 2020 and staying solidly on-track to Moore's Law.
In sharp contrast to this, Intel (INTC -1.13%), long known as the industry's trailblazer for new chip manufacturing technologies and geometries, is seemingly falling behind the Taiwan-based contract chip manufacturer as it faced challenges with its 14-nanometer technology and is projecting relatively long transition times to 10-nanometer and potentially even to 7-nanometer.
Although Intel beat TSMC to the punch at the 14/16-nanometer generation (and Intel's technology here appears superior), TSMC looks as though it will go into mass production on 10-nanometer a little ahead of Intel.
Intel should still have a technology advantage at 10-nanometers (TSMC's 10-nanometer technology is a bit ahead of Intel's 14-nanometer, but should be behind what Intel has planned at 10-nanometers), but TSMC will transition to the 7-nanometer node in the first half of 2018.
This rapid-fire series of node transitions should allow TSMC to reach at least parity with what Intel will be fielding at 10-nanometers with its "7-nanometer" technology, with the Taiwan-based contract chip manufacturer potentially even pulling slightly ahead of Intel in some metrics (i.e. chip density).
The question I have wanted to learn the answer to for quite some time is, "just how did this happen?"
After speaking to a source familiar with the matter, I can now bring you an answer that question.
TSMC shifts development model
According to the source, TSMC fundamentally revamped its research and development model for new chip technologies beginning with the 10-nanometer node that's expected to go into mass production at the end of this year.
For previous technologies, I am told, the same team of individuals was responsible for developing each successive node in an overlapping manner. By way of example, the folks who were working on the 40-nanometer technology were also working on the 28-nanometer technology at the same time in, as the source describes, a "staggered manner."
This, per the source, all changed with 10-nanometers and beyond. When TSMC wrapped up development of its 20-nanometer node, the team responsible for that technology shifted its efforts exclusively to the company's various flavors of its 16-nanometer chip manufacturing technology.
Interestingly, TSMC staffed up an entirely new team that would focus exclusively on the follow-on 10-nanometer technology. This team would not be burdened with thinking about either the 16-nanometer node or the follow-on 7-nanometer node -- it would be laser focused exclusively on 10-nanometer.
Once the 16-nanometer team had finished its work, it was reassigned to focus exclusively on the development of 7-nanometer, which TSMC says will begin risk production in early 2017 for mass production in the first half of 2018. And, once 10-nanometer is "in the bag" (which should be fairly soon) that team will start working in earnest (and will focus entirely on) the 5-nanometer node.
This fundamental change is what has allowed TSMC to go from a company that routinely over-promised and under-delivered to a company that can be relied on by top fabless customers to quickly move from node-to-node.
Intel, not so much
So, what does this have to do with Intel? Well, in doing due diligence on this matter, I was told by a reliable source that Intel has one, very large research-and-development team that works on new chip manufacturing technologies in essentially a "staggered" fashion.
However, according to an official statement from Intel, the company has three to four completely independent teams working on multiple generations of chip manufacturing technologies in parallel.
That being said, in performing due diligence, I went through many LinkedIn profiles of process development engineers at Intel and noticed that many of the engineers who worked on the company's 22-nanometer process also worked on the company's 14-nanometer process.
Indeed, according to my source, the team that worked hard on the 22-nanometer node (which Intel boasts is its "highest yielding technology ever") was also supposed to be working on making 14-nanometer happen.
However, as the source tells me, Intel was so focused on trying to get 22-nanometer into great shape (it is said that Intel's Mark Bohr and former then-CEO Paul Otellini wanted "showcase material"), that 14-nanometer was essentially "left on the back burner for way too long."
As a result of this, as I am told, the 14-nanometer technology was about a year behind schedule when it was put into the production factories. Then, in attempt to solve the issues "on the fly," Intel pulled R&D engineers off of the development of 10-nanometer to work side-by-side with the production process engineers working on 14-nanometer to get those yield issues worked out.
If the teams are truly parallel, then why the delays?
There are two "competing" claims here: one, from Intel, saying that the process development teams are "truly independent" and the other from a source that I believe to be very reliable saying that this is not the case.
As an investor, though, I have to ask: if the teams are truly independent, then why does it appear that the delays are related? We know that 14-nanometer got off to a bad start and that Intel is still trying to improve yields to bring its costs in-line with what its 22-nanometer products delivered (something that Intel believes it will achieve by the end of 2016).
In a situation where the process development teams are really independent, the 14-nanometer issues shouldn't have had the "knock on" effect of delaying 10-nanometer by about a year; Intel should have been able to "move past" the 14-nanometer issues as it transitioned to 10-nanometer.
Indeed, Intel had said that 10-nanometer would go into production in late 2015, but it's now looking like late 2016/early 2017, and 7-nanometer -- which was supposed to go into production in late 2017 -- is looking more like a 2020 affair.
At any rate, for investors, the bottom line is that Intel's continued manufacturing delays appear to be giving its competitors room to "catch up." Given that chip manufacturing technology leadership is the advantage that Intel seems to tout the most, I can't help but be a little uneasy as an Intel stockholder.
*Editor's note: Intel contacted the writer following the publication of this article to clarify that the company employs an independent, multi-team structure.