Advanced Micro Devices (NYSE:AMD) just announced its low power "Kyoto" APUs, which are essentially server-oriented editions of its "Kabini" APU for laptops. These APUs are fully integrated system-on-chip solutions with thermal design power ranging from 9W to 22W. While Intel (NASDAQ:INTC) released its "Centerton" Atom S1200 late in 2012, it seems that Advanced Micro Devices actually does precisely what it is doing in the low cost/low power notebook space - leapfrogging the current Atom parts by a substantial margin.
This Is Just Kabini... But It's Still Well Suited For Microservers
When Intel released the "Centerton" for micro-servers (which was basically a "special" edition of its Atom for netbooks/nettops), it was pretty good compared to the ARM (NASDAQ:ARMH) stuff on the market because, well, none of that was competitive either. Intel touted ECC and 64-bit support, along with a rather low TDP, but unfortunately, without more/better cores, integrated Ethernet, more memory support, and a faster memory interface, "Centerton" was destined to be more of a "hey, look we got here on time" part rather than a game-changing, industry leading part.
What AMD is doing here is pretty similar - taking its notebook oriented "Kabini" APU and just stamping it "Opteron". Now, since the "Jaguar" cores in "Kyoto" mop the floor with the "Saltwell" cores present in the "Centerton" (and there are twice as many of them), this gives AMD a pretty nice advantage, albeit at higher power consumption numbers. Further AMD has integrated graphics versions that I'm sure are rather attractive to customers either looking for GPU compute, or simply looking to run graphical workloads.
Unfortunately, while the "Kyoto" is a faster/more capable part than "Centerton", it does not integrate things like Ethernet, and the core count seems to top out at 4 (which is in-line with the thesis that this is rebranded "Kabini" and not a specialized micro-server part). I expect AMD's ARM Cortex A57 based micro-server part to feature deeper levels of integration with interconnect fabric when it sees the light of day in 2014.
"Avoton" Keeps Window Of Opportunity Small
The big issue is that while "Kabini" renamed "Kyoto" is certainly a good chip for these types of workloads, as I said, it lacks a lot of the key integration/interconnect that this space will ultimately demand. Intel's "Avoton" which is an 8 core (and each core is substantially improved) micro-server part that also integrates Ethernet, and its "Rangely" iteration that features an on-board dedicated cryptography engine, should see release in 2H 2013 (and according to discussions with Intel, this should not be an "end of the year" event such as "Centerton"), and given that these parts have been sampling at partners for quite some time (since at least IDF 2013 Beijing), it is likely that this may curtail the design win/sell through momentum that AMD would have otherwise garnered had it needed to fight only "Centerton".
That being said, this part has scored a socket in HP's (NYSE:HPQ) "Moonshot" program, which could turn out to be very interesting.
AMD shares are in full-blast mode, with many positive news items having hit recently, and very likely many more to come. On any significant pullback, it would probably do you well to consider making this your speculative position for the year as the buzz from "Kabini", "Kyoto", "Temash", and perhaps even "Kaveri" should push the stock higher exiting the year, particularly if the profitability/cash flow positive goals are reached as promised. The headline for the day will be that AMD has a better micro-server offering than anybody else out there right, and I certainly would find it hard to disagree.
Additional disclosure: I am short ARMH