I'm seeing a whole lot of criticism of Qualcomm (NASDAQ:QCOM) these days for choosing to use ARM's (NASDAQ:ARMH) "off-the-shelf" CPU cores for its initial 64-bit lineup. Some will claim that Qualcomm has "given up" trying to compete on performance, or that the company is failing to differentiate -- blah, blah, blah. While I understand these comments, critics seem to be missing the forest for the trees here.
A mobile chip is more than a CPU
Qualcomm sells complete wireless platforms. This includes an applications processor (which features many unique IP blocks such as CPU, graphics, image signal processor, and so on), a cellular modem (which Qualcomm in many cases integrates into its applications processor), connectivity, and all of the attendant software/drivers. These are very complex devices, and the commercial success or failure of a given part isn't driven solely (or even primarily) by CPU performance.
Qualcomm is still doing its own 64-bit CPU, but who cares?
In its initial press release for the Snapdragon 808 and 810 (which are the chips with the ARM cores), Qualcomm was quick to point out that it was still working on a custom, in-house 64-bit CPU. That's great, wonderful, and dandy, but who really cares? Qualcomm's implementation of ARM's CPU cores in its lower-end/high-volume parts has typically been very good, and I'm sure the company's physical design teams will do a great job implementing the Cortex A57.
Of course Qualcomm wants to tout its own custom CPU cores -- they're a nice marketing feature for the few technology fans who care. But at the end of the day, a user is likely to notice the impact from a more capable image signal processor (pictures and video), better connectivity/cellular (surfing the Web and downloading apps), and even graphics processor (games). If ARM's CPU cores were total junk, then there would be a pressing need to develop custom ones, but ARM's cores are quite good and differentiation can be achieved in the implementation of those cores.
News flash – the high end is slowing!
In the midrange and low-end space, Qualcomm has more or less shifted to ARM-designed cores. Qualcomm's own Krait core found its way into some of the midrange and low-end products, but the variants with ARM's Cortex A7 sold far better and became Qualcomm's workhorses. This is where most of Qualcomm's unit volume is, and -- according to ARM Holdings -- this is where most of future smartphone growth is going to come from.
Why should Qualcomm invest considerable resources in trying to differentiate its CPU core at the high end (which is fairly slow-growing), when the top "premium" player Apple (NASDAQ:AAPL) designs its apps processors and captures most of the market anyway? Furthermore, given Qualcomm's superior modem technology (as well as just general system-on-a-chip superiority), its position at the high end with the less powerful high-end vendors is pretty safe -- stock ARM CPU or custom Qualcomm CPU.
Qualcomm needs to keep its cost structure as low as possible and needs to allocate research and development resources wisely. Staying several steps ahead on modem technology, investing to capture more content share, and bolstering the IPs that can't be licensed from ARM would seem a better use of R&D dollars. Will Qualcomm keep building its own CPU cores? Probably. Does it need to? That's up for debate, but given well how the Snapdragon 808/810 are likely to sell, the answer is "probably not."