It is well known throughout the industry that Samsung (NASDAQOTH: SSNLF), the world's second-most profitable smartphone vendor, designs its own system-on-chip products for use in its devices. Despite this, Samsung very often uses designs from other vendors depending on the requirements of a phone. In particular, Samsung uses Qualcomm (QCOM 1.56%) chips in many of its flagship devices, largely due to the fact that when it comes to cellular connectivity (and efficient, highly integrated smartphone apps processors) Qualcomm is king. However, while this certainly speaks volumes to the skill of Qualcomm's chip teams, there is a risk that I've alluded to in prior articles: Samsung's own in-house chip team(s).
Background on Samsung and ARM
While Qualcomm's chips feature all-Qualcomm IP from the applications processor to the graphics, Samsung's Exynos chips have largely consisted of off-the-shelf ARM (ARMH +0.00%) Cortex A-series CPUs with either ARM Mali GPUs or Imagination's PowerVR. Naturally, while off-the-shelf IP aimed at multiple different markets isn't quite going to be as optimal as custom-tailored IP, the IP that is available is, for the most part, quite good. Further, there's no reason that Samsung can't/won't develop its own custom IP if it needs something more fine-tuned for its need.
What has given Qualcomm an advantage is that its modems (baseband and RF) are simply the best in the industry, particularly in adopting new cellular standards ahead of the curve. Qualcomm has been shipping LTE for years now and embeds its third-generation LTE-Advanced modem in its latest Snapdragon 800 processor – while many of its competitors are still struggling to ship their first LTE solutions. Samsung does have its own cellular baseband efforts, but they're not quite in Qualcomm's league.
Qualcomm: A potential problem with an interesting twist
A big potential problem for Qualcomm is that if Samsung is to get its cellular baseband efforts in order (and Samsung's RF transceiver supplier Silicon Motion seems to think that this will happen sooner or later – with Samsung's intent on "sooner"), then Qualcomm could lose a good chunk of chip sales fairly quickly.
Now, that being said, the relationship between Qualcomm and Samsung is a little bit more nuanced than it would seem on the surface. Would Samsung like to cut out the middle-man if possible and use all of its own components? Sure. But Qualcomm isn't an ordinary supplier to Samsung – it's also a foundry customer. For example, the LTE baseband found in the Apple A7 is Qualcomm designed but Samsung built. Further, it's not unreasonable to expect that a portion of Snapdragon apps processors (and discrete modems if included) that Samsung uses are actually built in Samsung's chip manufacturing facilities rather than at the world's leading foundry, Taiwan Semiconductor, which, interestingly enough, lists Qualcomm as its largest customer.
So, while Samsung probably isn't too thrilled to be paying Qualcomm's design margin, it is in many cases pocketing the foundry margins, which could be enough for Samsung to not be so gung-ho on simply cutting Qualcomm off – after all, a good foundry relationship allows Samsung to profit from Qualcomm components that go into non-Samsung devices, while at the same time Samsung's phones benefit from Qualcomm's design expertise. It's really a win-win.
So, maybe it's not all bad
While there is still risk that Samsung's chip teams produce silicon that's compelling enough to give Qualcomm the boot, there are a number of nuances to this relationship that will help to keep this relationship unusually strong. That is, beyond the fact that today, Qualcomm is still the clear leader in smartphone silicon – from apps processor to RF.
