Image source: Intel.
For several years now, Intel (INTC 2.11%) has offered low-power server processors based on its Atom architecture. These chips don't offer anywhere near the performance that its higher-end Xeon chips do, but in some cases they offer higher levels of integration and lower power consumption.
Back in the 2012-2013 timeframe, Intel was very aggressive about rolling out these Atom-based server chips. In 2012, the company released Centerton, a dual core product built on its 32-nanometer Atom architecture and in late 2013, Intel released Avoton, an 8-core product based on a newer generation 22-nanometer Atom.
These moves were undoubtedly in order to try to fend off potential threats from chipmakers building chips based on the ARM (ARMH +0.00%) architecture to try to take share from Intel.
However, after Intel released Avoton and its networking-focused counterpart, known as Rangeley, innovation around the company's Atom-based server lineup seemed to come to a screeching halt. Avoton came out in late 2013, but as of mid-2016 Denverton still hasn't been formally announced.
Given what appears to have been multiple delays around Denverton, I had actually thought that the part might have just been cancelled outright. However, according to an article from AnandTech, Denverton lives and will, in fact, be released.
The rationale for releasing this product
I don't believe that the market for micro-servers -- that is, servers powered by many low-power, relatively slow processors -- has taken off as some had hoped. Indeed, Intel senior systems engineer Dave Hill told ZDNet in a late 2014 interview that the micro-server market is "just really not interesting."
That said, Intel wouldn't bother investing in these low-power Atom-based server chips unless there were some market for them. I believe that market is the networking chip market.
Recall that last quarter, Intel saw average selling prices for its data center-bound chips drop 7% sequentially and 3% year-over-year. CFO Stacy Smith attributed this decline to the fact that its networking chip segment grew 60% year-over-year, driving unit growth but diluting average selling prices. Smith noted that had product mix by segment remained constant year-over-year, Intel would have seen average selling price growth here.
Management previously blamed the weaker average selling prices for its networking segment on customers in that segment buying more Atom-based server chips than buyers in other segments do.
Atom may be critical to attacking the networking market
Intel has made it clear that it intends to gain significant share in the market for networking chips. In fact, at a recent investor conference, CEO Brian Krzanich said that the company recently "crossed over 10% market share" in that segment.
Intel pegs the total addressable market for networking silicon at around $18 billion, so if the company thinks that Atom-based chips are critical to gaining share in this market, it's not surprising to see it continue to invest in them.
It is worth noting though that Intel also sells higher performance Xeon E5 and Xeon D processors into the networking market (and I have little doubt that the company would rather sell those chips than the Atoms to customers here), so this segment isn't all about Atom.
