Industry Expert Blogs

Much has been written and said about the possible ARM invasion of the datacenter. The hype has been building lately as the first 64-bit implementations are being demonstrated at AMD and AppliedMicro, while Cavium should become more public sometime this summer when they get their very interesting chip ready for early sampling. Meanwhile, big players like Broadcom have development teams working on SOCs, with products likely in late 2015. Intel has responded to the potential threat with aggressively priced Atom SKUs to meet the need for “wimpy cores”…all while insisting that nobody really wants them. Let’s look at how large this market may become. Intel commands over 95% of the units volume that makes up the ~ $10 billion server processor market. the battle lines are reshaping as ARM-based processors reach server-class performance. You can expect 64-bit ARM servers to ship in the second half of 2014 and gradually ramp to significant volumes by 2016. Early adopters likely will be web, storage, and media servers primarily installed in internet-scale datacenters, as well as private cloud infrastructure. The enterprise ecosystem is not an immediate target, but this could change dramatically if Microsoft decides to bring out an ARM port for Windows Server. Like any disruptive technology, the ARM server chip opportunity is hard to predict. We know the market could be quite large, although probably not large enough to sustain the 16 SOC partners that ARM Holdings counts as server SOC licensees. Various third-parties, primarily sell-side analyst firms, have forecast robust growth in the micro-server SOC market through 2017, ranging from a low of $1B to over $3B in SOC sales (double that for server sales). How much of that could ARM take? That depends on how badly people want an alternative to emerge and the extent to which Intel is willing to go to prevent ARM from gaining traction.

Click here to read more ...