AMD Will Probably Continue to Lose PC Market Share During 2015

At Advanced Micro Devices' (AMD 1.30%) recent Core Innovation Update, the company talked a rather big game about how it was going back to the drawing board, crafting two new processor cores -- one that implements the X86 instruction set architecture and another that implements the ARM Holdings (ARMH) architecture. While this does represent a glimmer of hope for the company (and the competitiveness of the series of products built on this IP may well reverse AMD's fortunes), investors in AMD will probably continue to deal with continued PC share loss throughout 2015.

In 2015, Intel will bring Braswell, Broadwell, and Skylake to the table
Beginning in Q4 2014, Intel (INTC 1.39%) -- the competitor to whom AMD has been losing PC share for years -- will roll out Broadwell (most likely the Ultrabook and fanless system SKUs), which should further cement Intel's position in the low-power, high-performance space. Not only does Intel benefit from a transition to 14-nanometer (which brings performance, power, and economic benefit), but it will also bring with it a significantly revamped graphics engine, which could serve to nullify AMD's last remaining selling point in the mid- to high end.

In addition, Intel plans to roll out a platform known as Braswell, likely during the first half of 2015. This is a low-cost 14-nanometer platform based on the Airmont CPU core and a low-power variant of the new graphics engine found in Broadwell. Not only will Intel have a pretty significant die size advantage thanks to the 14-nanometer process, but -- as a result of its mobile push -- it should have a pretty significant performance and potentially platform bill-of-materials advantage.

Source: Intel. 

By the second half of 2015, Intel will be ramping its next-generation "tock" known as Skylake, likely for Q4 2015 system availability (if the current cadence is followed). This should bring both a brand new CPU core as well as an updated graphics engine. According to VR-Zone, Intel will also offer variants with fairly aggressive graphics performance even for thin and light Ultrabooks.

What does AMD have in 2015?
AMD, in contrast, will be moving its most recently announced Puma CPU core to the 20-nanometer manufacturing process. Interestingly enough, the Mullins/Beema chips based on the Puma core have yet to show up in commercial systems and in fact were just recently "previewed" by a number of technology sites. Parts have yet to be tested (for either power or performance) by independent third parties outside of controlled AMD testing.

This implies, then, that the 20-nanometer shrink of these parts (along with the ARM Cortex A57-based version of this product) should be out within a year of the Beema and Mullins launches, putting them in systems perhaps in June or July 2015. It's difficult to see these products (built on Taiwan Semiconductor's 20nm planar process) competitive on a performance, watt, and dollar basis against Intel's offerings at the time in the mass-market PC space. AMD will probably continue to lose share in PCs throughout 2015.

But 2016 looks better
By 2016, AMD will have an entirely new CPU core and new system-on-a-chip products. These should also be built on the Global Foundries/Samsung 14-nanometer process, which, while not as dense as Intel's 14-nanometer process, should bring considerable transistor-level performance improvements. Now, we don't know when during 2016 these products will roll out. If it's mid-2016, then there is a window of a couple of months that AMD's 14-nanometer parts will compete with Intel's 14-nanometer parts. If it's late 2016, then AMD faces Intel's 10-nanometer parts, putting it right back where it started -- at a full process node disadvantage.

That said, AMD's problems today go far beyond the process technology, but are instead in poor architectural choices, so if AMD can really do something good on the architectural side, it stands a shot of improving its competitive position, even if it doesn't end up setting the world on fire. The big offset to this is that at worst, AMD will be competing with Intel's second-generation 14-nanometer architectures (both on Atom and Core).

Foolish bottom line
AMD's not in a great spot, but if it can deliver significant architectural enhancements and, more importantly, continue to build out its non-PC businesses, then there is probably a good future for the company. There's less to be bullish about on the traditional PC processor space, given Intel's enormous resources and manufacturing lead, but AMD doesn't need to outgun Intel in PCs to ultimately build a viable business. It simply needs to choose its battles very carefully, and it needs to offer some differentiation, some edge that nobody else has.

Right now, AMD is proving itself quite well in the semi-custom chip design business. If AMD can broaden its customer range and really show that it can execute there, then this could be AMD's future, and it's not a bad one if the company plays its cards right.