In the early days of the mobile revolution, one of the big selling points of the ARM (NASDAQ:ARMH) architecture was that -- in theory -- a company with enough cash to take out an ARM processor license (or, if it wanted to get really fancy, an architectural license to design its own cores) could simply build its own chip. Why go through a middleman if you can just roll your own and potentially keep the design margin all to yourself, particularly if you're an OEM looking to cut component costs?
Well, it's not so simple, and many players (most recently Broadcom (UNKNOWN:BRCM.DL)) that thought they could cash in on this "revolution" are finding that this space is exceptionally research and development-intensive and that there's only room for a couple competitors.
Samsung and Exynos
Samsung (NASDAQOTH:SSNLF) designs its own Exynos line of processors. Historically, these chips have integrated ARM core designs, either ARM or Imagination Technologies' graphics blocks, with the rest of the system-on-chip (image signal processor, memory controller, etc.) designed by Samsung.
However, Samsung often opts to use chips designed by Qualcomm (NASDAQ:QCOM) for many key flagship phones in higher-volume markets. And it often taps other vendors such as Marvell (NASDAQ:MRVL) and even Broadcom for some lower-cost models internationally. Exynos usually finds itself in some of Samsung's Wi-Fi-only tablets and some variants of its flagship phones abroad. It's actually shocking how little Samsung uses Exynos in its own designs.
Will a custom processor core help? Probably not
There seems to be a misconception that Samsung's Exynos problems will magically disappear should the company introduce a variant with a custom CPU core. Samsung has indicated that this is planned for future Exynos products, but the fact of the matter is that ARM has done a pretty solid job with its own custom cores. Qualcomm's Krait processor and even Apple's Cylone aren't all that much better than a reasonable implementation of the ARM Cortex A15.
What makes a system-on-chip "great" is not necessarily any one IP block (ARM and Imagination IP blocks are already fantastic), but the implementation of those blocks and integration with in-house IPs that aren't so commoditized. The modem, particularly phones, is hugely important, and everything from the quality of system-on-chip fabric (to make sure all of the IP blocks efficiently communicate) to the image signal processor (which is responsible for a big part of the picture- and video-taking experience on a device) make a huge impact on the user experience.
A custom CPU core these days is mostly for bragging rights if you're an ARM player, or a necessity if implementing a proprietary architecture such as Intel's X86.
Exynos also gets in the way of Samsung's foundry ambitions
Samsung is also getting much more aggressive in pursuing the foundry business, and the vast majority of the leading-edge process node volumes that matter are -- you guessed it -- from the mobile-chip vendors. As long as Samsung develops Exynos, the various fabless companies such as Qualcomm (which, for example, sells a lot of apps processors to Samsung) will probably be somewhat wary that in the long run, Samsung just wants to cut them out when or if Exynos becomes competitive.
So instead of pouring ever-increasing amounts of money down that mobile chip-development money pit, why not simply save that R&D cash and instead profit from the foundry business? Samsung's core business is selling mobile devices and DRAM, so it doesn't depend on designing chips for a living. With the Exynos program simply shut down, Samsung becomes a much more approachable foundry (as it won't compete with its customers) and could profit quite nicely from the existing process technology R&D infrastructure already in place.
Given that Samsung uses Exynos in so few products, and that mobile chip R&D is likely to remain more expensive as time goes on, Exynos makes little sense. Furthermore, "competing" with potential foundry customers will make it hard to win their business.
But if Exynos dies, not only does Samsung likely save a good deal of annual R&D expense, but it could establish itself as a powerful foundry player, taking meaningful share from competitor Taiwan Semiconductor and becoming an even more powerful force than it already is in the semiconductor industry.
Ashraf Eassa owns shares of ARM Holdings and Intel. The Motley Fool recommends Apple and Intel. The Motley Fool owns shares of Apple, Imagination Technologies, Intel, and Qualcomm. Try any of our Foolish newsletter services free for 30 days. We Fools may not all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.