Last week, I wrote about a rumor floating that NVIDIA (NVDA 3.65%) is set to move its chip manufacturing business over to Samsung (NASDAQOTH: SSNLF) at the 14/16-nanometer generation. Interestingly enough, Ian Ing, from MKM Partners, asked the following during the question and answer session on NVIDIA's most recent earnings call:
Could you talk about your foundry strategy a little bit? I mean, TSMC, a good supplier, but not considered the best FinFET out there. Would you consider diversification being something you'd evaluate?
Here is the relevant bit from NVIDIA CEO Jen-Hsun Huang's response:
So we're working with TSMC on FinFET now for a couple of years, and so we have quite a bit of confidence in their ability to deliver amazing FinFET transistors. I guess with respect to that, we always look to look at all foundries, and TSMC remains our most strategic, of course, and they're going to continue to be a very, very important partner for us for the foreseeable future.
Huang's comments, coupled with NVIDIA's endorsement of Taiwan Semiconductor's (TSM 2.84%) 16 FinFET+ in which Senior Vice President Jeff Fisher explicitly praised the technology for both future stand-alone graphics processors and Tegra system-on-chip designs, make me doubt NVIDIA will move significant production quantities to Samsung anytime soon.
Samsung might start production "first," but is that really relevant?
Some argue NVIDIA could move its production to Samsung based on the fact that the South Korean tech giant is scheduled to begin production on 14-nanometer earlier than Taiwan Semiconductor will go into production on 16-nanometer FinFET+.
Specifically, Samsung claims it is ramping 14-nanometer production this quarter, while Taiwan Semiconductor should start around July of this year.
However, capacity for leading-edge technology is not just put in overnight, and many semiconductor equipment vendors -- such as Ultratech (UTEK), Applied Materials (AMAT 2.00%), and KLA-Tencor (KLAC 2.30%) -- have all indicated the foundries will ramp up their FinFET technologies more in the second half of the year.
While NVIDIA is a relatively large consumer of leading-edge technology (at least in discrete GPUs), it is unlikely to be first in line at the Samsung foundry while big dogs such as Apple (AAPL 0.64%), Qualcomm (QCOM 0.73%), and Samsung's own internal teams are all reportedly queuing up for those initial wafers.
In my view, by the time NVIDIA gets ready to ramp next-generation stand-alone GPUs and the follow-on to the recently released Tegra X1, Taiwan Semiconductor should be well into production of its 16-nanometer technology. Samsung's relatively early time to market likely isn't material to NVIDIA's FinFET choice.
One more hint that NVIDIA may stay with Taiwan Semiconductor
Taiwan Semiconductor Chairman Morris Chang had the following to say in January regarding 14/16-nanometer FinFET market share in 2015-2016: "We have a smaller market share [at the 14/16-nanometer generation] in 2015, in a smaller market; next year we'll have a larger share, in fact much larger share in a much larger market./
If Chang has this kind of confidence that the market for 14/16-nanometer devices is set to rise sharply in 2016, and that his company's share within that market will increase significantly as well, this suggests a steep ramp, likely from multiple customers.
I wouldn't be surprised if a next-generation Tegra after the recently announced Tegra X1, coupled with a bunch of large graphics processors from NVIDIA in that time frame, were meaningful contributors to TSMC's confidence in its FinFET share in 2016.
