There has been a lot of buzz surrounding HP's (NYSE:HPQ) "Project Moonshot" microserver architecture over the last year or so, particularly as the server architecture will also support various flavors of ARM (NASDAQ:ARMH) based system-on-chip products. As many who follow the drama in the semiconductor industry these days already know, ARM and its partners - particularly Applied Micro (NASDAQ:AMCC), Calxeda, and Advanced Micro Devices (NYSE:AMD) - have been touting their entries into the dense-server market by way of custom SoCs ("system-on-chip") based on either ARM's upcoming A57 cores or custom ARMv8 (64 bit instruction set) cores. The idea is that for a number of web and storage based workloads, having multiple, lower power cores is a more effective solution on a TCO ("total cost of ownership") basis than having a bunch of higher performance, higher power draw processors.
The big fear (or hope depending on how you're financially positioned) on the Intel (NASDAQ:INTC) side of things was that ARM vendors would swarm this market and eat Intel's server lunch, so to speak. The notion that ARM has managed to drill into everyone's minds is that its chips are much more power efficient and cheaper, so Intel is going to bleed server market share and/or margins. Interestingly, though, Intel was actually the first to market with a 64-bit Atom SoC called "Centerton". Now, this chip wasn't much of a thriller as it was based on the old 2008 Atom core built on Intel's 32nm process, and didn't integrate key features such as networking controllers onto the chip.
However, Intel announced on 4/8 that it was shipping samples of its upcoming "Avoton" SoC to customers. "Avoton" is built on Intel's 22nm process (competitors are on 40nm and will be transitioning to 28nm late this year/early next year), packs the newly redesigned "Silvermont" Atom cores, and integrates key things such as networking. According to the company, systems based on these chips will be commercially available in designs during 2H 2013, giving the firm a significant head start over the first 64-bit ARM designs expected to hit the market sometime during 2014.
In my view, Intel's reputation as a reliable server chip vendor, coupled with the process technology lead (giving performance and cost benefits), should allow it to vigorously defend its position in the growing micro-server market against the ARM designs, which typically come from startups that do not have nearly the R&D or marketing budgets that Intel possesses. I believe that if the rumors of Qualcomm (NASDAQ:QCOM) and Samsung (OTC:SSNLF) getting into this game are true, then the competitive landscape will heat up significantly, although both of these vendors still lack the experience that Intel has (and will both likely be hamstrung by process technology disadvantages).
From an investment standpoint, the biggest loser here is Applied Micro Circuits, whose shares have seen a considerable run up, thanks to the hype surrounding the 64-bit ARM instruction set and its own "X-Gene" part that utilizes it. Not only will Applied Micro be stuck on a 2-generation old 40nm process against Intel's 22nm process (thanks in no small part to numerous delays and a habit of over-promising and under-delivering), but it has nowhere near the R&D budget or credibility in the micro-server market that Intel has. Further, the claims of wild levels of integration on the X-Gene, as well as a high performance core, leads me to believe that the X-Gene part will have a significant performance/watt disadvantage relative to the Intel chips, as well as even the 28nm Calxeda products.
Interesting times ahead in the chip wars, but if you're thinking that ARM vendors are going to come in and steal Intel's dominant position in servers simply by virtue of being compatible with the ARM instruction set (Wall Street's latest hip fad), then you may be sorely disappointed.
Additional disclosure: I am short ARMH, and plan to short AMCC on any real X-Gene related hype.