Kevin Gibb of Tech Insights recently claimed that Samsung (NASDAQOTH: SSNLF) has narrowed the gap between itself and Intel (INTC +12.91%) in terms of chip manufacturing technologies to just six months. This, Gibb asserts, is down from the "traditional one year or more lead on TSMC (TSM 1.75%) and Samsung" that Intel has typically enjoyed. He even lauded it as a "remarkable feat for Samsung."
Not so fast!
While it is true that Samsung began mass production of its first 14-nanometer processor just six months after Intel began manufacturing its own, there's so much more to the story than that.
For example, are the two "14-nanometer" technologies even equivalent across a number of relevant electrical/physical characteristics? Are Samsung's 14-nanometer yields anywhere close to Intel's? The answers to these questions are absolutely critical to trying to determine what sort of "lead" one company has over another.
Let's talk about the actual technologies
Putting aside the yield discussion for a moment, Intel's 14-nanometer technology and Samsung's 14-nanometer technology are not equivalent. From how I understand it, two major drivers of the density of a technology -- that is, how many transistors can be packed into a given area -- are the gate pitch and the minimum metal pitch.
Samsung has said that its 14-nanometer technology features a 64-nanometer minimum metal pitch, and Tech Insights' analysis measures the gate pitch at 77 nanometers. Intel's 14-nanometer technology, per its disclosures at the International Electron Devices Meeting last year, delivers a gate pitch of 70-nanometers and a minimum metal pitch of 52-nanometers, respectively.
This means if a chip designer were to implement the same chip in Samsung's and Intel's respective 14-nanometer technologies, the one implemented in the Intel process should take up less area.
Intel also disclosed that at the "performance critical" 80-nanometer and 160-nanometer metal layers, its 14-nanometer process includes air-gaps. This, according to Intel, provides "a 17% improvement in capacitance." To my knowledge, Samsung has not implemented air-gaps in its own 14-nanometer technology.
A subtle difference between Intel and Samsung
Neither Intel nor Samsung publish yield rates for chips implemented on their respective manufacturing technologies. However, do note that the only appearance we're seeing of a Samsung 14-nanometer chip is inside Samsung's own Galaxy S6 -- a phone that it can sell for $700 for the lowest storage-tier model.
Intel, on the other hand, sells chips. While Samsung can potentially "hide" bad yields (and ultimately high chip manufacturing costs) inside of a very expensive, relatively high-margin phone, Intel can't. It needs to maintain 60%-plus gross profit margins by selling chips themselves.
That means Samsung can "go into production" sooner than Intel can, as the latter is highly dependent on high manufacturing yields to be able to sell chips at prices the market can tolerate while keeping margins high.
Speaking of yields...
In fact, a March 2015 report from Bernstein Research's Mark Li (via Barron's), reported that Samsung "experienced new yield difficulty recently" on its 14-nanometer technology. This, Li says, "has prompted [Apple (AAPL +2.65%)] to allocate ~40% of the A9 business to TSMC" and keep the rest at Samsung.
Separately, Daiwa Securities said that TSMC "will earn most of the A9 orders thanks to its superior yield ramp." Cowen's analyst echoed that sentiment in a report.
If Samsung were building high-yielding Exynos chips now, wouldn't Apple be rushing to secure as much Samsung 14-nanometer capacity as possible?
Looking beyond the headlines
Based on the above analysis, I'm not sure it's wise to claim that Samsung is just now six months behind Intel in terms of ramping up leading-edge manufacturing technologies. I believe Intel's 14-nanometer technology is denser, and I believe Intel is getting significantly better yields on its own 14-nanometer products than Samsung is on its Exynos 7420.
It's hard to quantify Intel's "lead" in terms of years, but I think it's meaningfully larger than the six months Gibb claims.
