According to an analyst quoted in an article published in Business Korea, Samsung (NASDAQOTH: SSNLF) will be responsible for the manufacture of applications processors for NVIDIA (NVDA 4.14%) on the former's 14-nanometer FinFET manufacturing technology. This is particularly newsworthy given NVIDIA's bitter legal battle with Samsung coupled with the former's recent strong endorsement of Taiwan Semiconductor's (TSM 6.29%) 16-nanometer FinFET+ manufacturing technology.
The question, then, is how much truth is there to this rumor? To find out, I did some digging.
NVIDIA has certainly designed test-chips on Samsung's 14-nanometer process
According to the following LinkedIn profile of an NVIDIA engineer, NVIDIA has done physical design work on a "14nm" process:
Source: LinkedIn.
Now, if you read this LinkedIn profile closely, you'll see that this engineer is responsible for "test-chip implementation." As a fabless company, NVIDIA is obviously going to evaluate all of the available processes available to it. This means that it very likely evaluated both Samsung 14-nanometer and Taiwan Semiconductor 16-nanometer FinFET+.
But will Samsung begin mass producing NVIDIA chips in the second quarter of 2015?
The analyst quoted in the Business Korea article claims that Samsung will mass produce 14-nanometer chips for Apple (AAPL +0.11%) and Qualcomm (QCOM +0.15%), as well as NVIDIA in the second quarter of 2015. TSMC has stated that it takes about 1.5 quarters to go from production start to revenue recognition at the foundry level. So, let's call it two quarters before meaningful volumes of chips hit the shelves.
NVIDIA announced its first 20-nanometer part -- the Tegra X1 -- in January 2015 and parts are expected to ship in high volumes in the second quarter of 2015. If mass production on the next Tegra chip, reportedly on the Samsung 14-nanometer process begins in the second quarter of 2015, then this would imply a launch sometime in the fourth quarter of 2015. This would give the Tegra X1 a usable life of just two quarters.
Isn't Apple going to need that initial capacity?
Another thing that I think is important to note is that Samsung probably isn't going to have all of its final 14-nanometer capacity in place by the second quarter of 2015. I'd imagine that the initial 14-nanometer capacity will go to the high-priority clients such as Apple (which is rumored to be building most of its upcoming A9 chips at Samsung) and Qualcomm (also rumored to be building 14-nanometer chips at Samsung), and then companies like NVIDIA begin production in earnest later.
That being said, I do think it's rather odd that NVIDIA would build at Samsung given its glowing endorsement of the 16-nanometer FinFET+ process from Taiwan Semiconductor (and its notable absence from the press release of the Samsung/GLOBALFOUNDRIES deal).
I'm not completely sold on it
I'm not sold on the idea that NVIDIA will begin mass producing 14-nanometer Tegra chips (let alone discrete GPUs) at Samsung. Given that NVIDIA is rarely first to move to a new process, and given that longtime partner Taiwan Semiconductor is expected to go into production on 16-nanometer FinFET+ in time for product launches next year, it just doesn't seem likely that it's going to build chips at Samsung. We'll know soon enough, though.
