It has been a rough several years for Advanced Micro Devices (AMD, Financial) . After beating Intel (INTC, Financial) senseless during Intel's ill-fated Pentium 4 generation, Intel came back swinging and reversed every last point of PC and server market share that it had gained during that time. While Intel's execution was pretty breathtaking at the time, Intel's landslide victory was also driven by AMD essentially shooting itself in the foot with poor strategic and technical decisions. AMD hopes to wipe the slate clean with a brand-new, clean-sheet design.
AMD's K12 is a fresh start
The biggest problem AMD has faced over the last several years has been an extremely poor micro-architecture codenamed "Bulldozer." It ran hot, took up a lot of die area, and was just plain uncompetitive along any axis that mattered -- performance, power, and manufacturing cost. Over the last few years, AMD has refined Bulldozer, bringing out Piledriver and more recently Steamroller, which have fixed some -- but not all -- of the competitive problems vis-a-vis Intel.
It seems, however, AMD is finally tossing away the Bulldozer legacy and rebuilding a new processor core from scratch codenamed K12. In fact, AMD is doing two flavors of the chip: one that implements the X86 instruction set and one that implements the ARM (AMRH) instruction set. From what AMD has indicated, these will be roughly comparable in terms of performance, power, and area.
Tossing away Jaguar, too
It seems AMD is consolidating its development efforts to the X86/ARM CPU pair and will likely no longer have two separate cores for different segments -- one converged core to "rule them all." This is very similar to what Intel does with its "big" Core product family, which is used to span everything from tablets to high performance servers. Intel also has a lower-cost, lower-power core targeted principally at smartphones and tablets, but also scales to low-cost PCs. Since AMD is not gunning for the smartphone market, this isn't a problem.
The K12 design point
Interestingly enough, while AMD's new K12 (both X86 and ARM variants) seem to be of the "bigger" variety. This makes sense since AMD is aiming for its core business to shift from the traditional PC market to things like:
- Networking infrastructure
- Dense servers
- Semi-custom
- Embedded
Having fairly powerful ARM and X86 cores that exceed the performance (and power consumption) design points of stock ARM cores (ARM Cortex A, to be precise) actually makes a lot of sense. If you need a really low power core, go for a Cortex A53, and if you need a solid low-power client (tablets, phones, low cost clamshells), the ARM Cortex A57 and successors will make sense. If you need more, design your own.
Android now open to AMD?
Thanks to an apparent shift to ARM across the stack (client to server), AMD can now more easily access client markets within which ARM is dominant and X86 still struggles. While AMD likely has no business going after the smartphone market, the Android tablet market is open to the company. However, given the robustness of the off-the-shelf ARM IP and the crowded nature of the landscape, it's tough to see AMD offering any meaningful differentiation, here. It's very likely AMD will participate on Android, but it will probably not be a large player there.
The TAM is bigger, but the competition is broader
AMD has always been able to tell a beautiful story about how it's changing things up, and that this time it's different. If AMD can execute, win back a good chunk of PC market share from Intel, and deliver on its goals in these new, adjacent markets, that would be absolutely fantastic for AMD's stock -- the recipe for well more than a doubling of today's share price.
However, it's important to keep in mind that just building an ARM core isn't enough to deliver the riches. In PCs, AMD still competes with Intel -- the goliath that won't take its foot off the pedal. In networking, AMD competes with Intel, Avago/LSI, Marvell, Cavium, Freescale, and Broadcom. In dense servers, AMD will be competing with Intel, Applied Micro, Qualcomm, Marvell, Cavium, Samsung, and potentially others.
Conclusion
By hopping onto the ARM bandwagon, AMD "escapes" an "unhealthy duopoly" as one of its executives put it, but instead of facing just Intel, it is now facingIntel and a slew of other, much stronger semiconductor players. The TAM got bigger, but the competitive landscape got even broader.
But, with a brand-new set of processor cores, AMD is finally ready to improve its competitive positioning starting in 2016. This won't be easy, but if AMD can deliver what it claims it will be able to, then there could be a long-term growth story here. But that's a big "if."
