Intel (INTC 4.09%) has been moving extremely aggressively following the flurry of announcements from ARM (ARMH +0.00%) licensees claiming that they will be going aggressively after the server processor market. Intel is the clear market leader in server processors, but upstarts like Cavium (CAVM +0.00%) and server veterans of yore like Advanced Micro Devices (AMD 1.31%) are getting more aggressive. Can Intel hold off the assault with market-specific products?
Atom & Xeon mount a pincer attack
Take a look at Intel's public roadmap for server-oriented system-on-chip products, and you'll see that the company has two distinct product lines: one based on the lower power/lower performance Atom and another based on the higher power/higher performance Core processors.
Source: Intel
Now, this road map tells us vague launch time frames, but the point here is that Intel is not going to leave any potential market segment exposed. If the customer wants a low-power system-on-chip with a lot of small cores, then Avoton/Denverton will be the right solution. However, if fewer, more powerful cores integrated in a system-on-chip float the customer's boat, then Broadwell SoC (known as Broadwell-DE) and its follow-ons will be the right product.
What might a Broadwell-DE look like?
Unfortunately, Intel has not given any public information about Broadwell-DE, but import data on Zauba lists a "Broadwell DE 8," which seems to hint that at least one of the variants will sport eight cores. Further, given that the current generation Atom server system-on-chip integrates Ethernet, PCIe, serial ATA, USB, and in some cases a cryptography accelerator, it stands to reason that Broadwell-DE will integrate at least that level of functionality.
But given that Intel is probably building this chip in order to counter the various efforts from the likes of Advanced Micro Devices (AMD 1.31%), it will likely need to include more dedicated offload engines. For example, the Seattle chip from AMD includes a number of dedicated offload engines that handle compression/decompression. From a general purpose CPU perspective, however, an eight core Broadwell should be best in class for the late 2014/2015 time frame.
How will Intel defend its share?
The obvious worry for Intel investors is just how the company will defend its share in the server market with the flurry of competitors coming online. Well, the intuitive answer is that Intel's R&D budget is much larger than that of most of its competitors and it has much more experience in the data center, which in itself should be a fairly large moat. But another major advantage is the company's manufacturing lead.
Note that Intel will be shipping 14-nanometer FinFET server parts in late 2014/early 2015, while competitors such as Cavium and even AMD will likely be stuck on 28-nanometer throughout 2015 and probably at least part-way into 2016. By the time these companies can ramp on 14/16 FinFET from the various foundries, Intel will have likely transitioned to the 10-nanometer node, once again giving it a density and power/performance advantage. This is not to mention that Intel's 14-nanometer process should offer a meaningful density edge over the foundries' processes.
The way to win is ultimately to offer the best performance per total cost of ownership, and Intel has been the clear leader for quite some time. In order for a competitor to take meaningful share from Intel, any competitor will need to demonstrate a fairly substantial edge over Intel's parts. Right now, it's unlikely that a competitor will match what Intel is doing in general purpose servers, let alone exceed.
Foolish bottom line
Intel's place in the data center is very hard won and based on years of very large R&D investments and smart business moves. While nothing is impossible and disruption by a smaller player could happen, it really is important to stand back and see that Intel has executed very well in this arena and continues to do so today. Just like the mobile system-on-chip market, where everybody thought that they could be the next Intel, many of these server chip upstarts are likely to see that trying to go head-to-head with an established, highly profitable market leader is far more difficult than initially believed.
