In a piece published in EETimes, analyst Mehdi Hosseini is quoted as saying that Samsung's (NASDAQOTH: SSNLF) 14-nanometer chip manufacturing yields are currently "exceeding 70%," and declaring doom and gloom for TSMC (NYSE:TSM), which plans to start its 16-nanometer manufacturing in roughly the July 2015 time frame.
I'm going to cut right to the chase and say that there is significant evidence out there to suggest that things may not be going as swimmingly as Hosseini claims for Samsung's 14-nanometer technology.
Checking in with Ultratech
According to semiconductor equipment vendor Ultratech (NASDAQ:UTEK), there is still significant uncertainty among the semiconductor equipment vendors as to whether they plan to install additional 28-nanometer chip capacity or whether they're going to build out more 14/16-nanometer capacity.
"We continue to see the industry grapple with issues for FinFETs to achieve acceptable yields," said Ultratech CFO Bruce Wright on the company's latest earnings call.
The message I got from the Ultratech call is this: If the semiconductor foundries can improve yields sufficiently to make 14/16-nanometer viable for foundry customers, then they'll increase 14/16-nanometer capacity to satisfy chip demand. If they can't, then the foundries may build out more 28-nanometer capacity.
If true, the decision would already be made
The three relevant non-Intel (NASDAQ:INTC) manufacturers of FinFET devices for the foreseeable are TSMC (NYSE:TSM), Samsung (NASDAQOTH: SSNLF), and GlobalFoundries.
Now, according to Ultratech's most recent form 10-K filing, Intel, Samsung, and GlobalFoundries were its three largest customers last year at 18%, 14%, and 13% of revenues in the year that ended on Dec. 31, 2014. Intel has been successfully ramping its 14-nanometer technology for quite a while and is even taking older generation 22-nanometer technology offline and reusing it at 14-nanometer.
Since there's no "indecision" on Intel's part whether to build out older technologies or 14-nanometer technology, I am led to believe that Samsung and GlobalFoundries are probably among those struggling to achieve acceptable yields.
If Samsung's 14-nanometer Exynos chips were yielding in excess of 70%, this would be a no brainer. 70% isn't optimal, but it's far from bad -- and for a foundry hungry to try to grab share from foundry leader TSMC, it would seem like a no brainer to just build out a bunch of capacity to capture that demand.
But if Ultratech's view is correct, then Samsung and GlobalFoundries – which are both using the same Samsung-developed 14-nanometer processes -- don't have great yields on the lead products built on the Samsung 14-nanometer technology.
Always follow the money
The fact that Ultratech is saying that its customers are unsure as to whether they want to build out more 28-nanometer capacity or add 14/16-nanometer capacity due to yields leads me to believe that the claims that Samsung's 14nm Exynos yields are above 70% are inaccurate.
I'm also not ready to accept Hosseini's claim that Samsung's 14-nanometer technology is just as good as TSMC's, but comes at a significantly lower cost.
TSMC is on record stating that in 2016, it will have "much larger" market share than Samsung will at the 14/16-nanometer node; that's not likely if it has the more expensive technology. If Hosseini is right, then I'd expect Samsung to continue to lead TSMC in terms of 14/16-nanometer FinFET share in 2016.
We'll see who was ultimately right next year. I, for one, can't wait to see how it plays out.