It's easy to forget that despite Intel's (INTC +0.42%) struggles to gain a meaningful foothold in handsets, the company operates an absolutely amazing datacenter business. It's clear that this division has been a top strategic focus, no doubt driven by its attractive growth prospects, extremely wide and deep moat, and hefty 50% operating margin.
While many large and powerful players have attempted to resist Intel's ferocious market share gains, they have all failed. Ask IBM how its POWER chips are selling these days, or Oracle how much life there is in its SPARC sales, and you will immediately appreciate the sheer strength of Intel in this space.
Intel has the more traditional enterprise server business sewn up with its Xeon processors. And it has made life very difficult for the high end niche RISC architectures such as the aforementioned POWER and SPARC chips (Intel's Xeons keep getting beefier year after year).
Yet, many of the ARM Holdings(ARMH +0.00%) offerings have been making noise about disrupting Intel in this space. It's likely that some of these players gain some traction, but Intel's product roadmap is, quite frankly, so compelling that it's difficult to see any obvious cracks in the armor for a smaller firm to try to exploit.
Intel's roadmap is extremely compelling
The big "buzz" in the datacenter today is all about power efficiency and integration. The two main arguments that the ARM players, (like privately held Calxeda), make are the following:
- The use of small, mobile-oriented processor cores (but a lot of them) is optimal for a number of workloads where a "big" core is simply overkill.
- The integration of things such as Ethernet, dedicated accelerators, and system fabrics is key to driving efficiency for these scale out workloads.
These are certainly valid claims, and it's clear that for scale-out workloads, highly integrated system-on-chip solutions are preferred. Fortunately for Intel, the company is shipping highly integrated system-on-chip products intended for micro-servers, low end switches and routers, cold storage, and wireless infrastructure. These are based on the low power "Silvermont" processor but come in a variety of core counts (up to 8) and feature-sets.
Beyond this, Intel plans to bring all the highly integrated system-on-chip goodness to parts built on its "big" core. While Intel ships "big core" based parts intended for the micro server market known as "Xeon E3," these chips don't come integrated with server/datacenter specific features such as Ethernet, cryptography acceleration, and various other things that these workloads will eventually need.
In addition, these are two-chip solutions (with things such as USB, SATA, and the like in a separate chip), which isn't as efficient as building everything onto the main system-on-chip. The first highly integrated system-on-chip built on Intel's 14 nanometer "big" core design, known as Broadwell, should be available next year.
So, where is the opening for the competition?
There are a number of key hurdles that anybody looking to go head-to-head with Intel will have to deal with. First and foremost, Intel has a sizable process technology lead. While some may dismiss this, the bottom line is that the performance and power of a given chip depends heavily on the performance and power characteristics of the transistors (that is, the building blocks of any chip). A clever design only gets you so far (and Intel is also full of these).
On top of that, Intel has pruned the server software ecosystem to be very reliant on Intel Architecture. Sure, there's the argument that software can be recompiled, but it's rarely so simple. Validating these highly complex pieces of code when it's rebuilt for a new architecture is tough, and doing so without any compelling reason to do so is even tougher.
The ARM players will need to demonstrate a significant (that is 30% to 50% minimum) advantage in performance per watt to have a chance against Intel's offerings. And, quite frankly, with Intel's product lineup so compelling, and its process advantage intact, it's tough to see these companies making meaningful inroads in the traditional enterprise and cloud server spaces.
Foolish bottom line
While it's always fun to root for the little guy, it's important not to lose sight of the bigger picture. Intel's server products have never been more competitive, and it's clear the company will continue to accelerate its efforts. Intel is likely to continue to take share from the traditional big iron RISC players, and it will vigorously defend its share in traditional enterprise as well as in the cloud/hyper-scale. The datacenter story has never looked better for Intel.
