There has been a lot of hype about IBM's (NYSE:IBM) renewed efforts to attack the market for server processors with its OpenPOWER initiative. Interestingly, a while back, I read a column from Seeking Alpha's Mark Hibben that argued that in 2016 IBM might be able to "catch up" to Intel (NASDAQ:INTC) in terms of chip manufacturing technology for its server processors as it migrates to its own 14-nanometer technology.
It would appear, however, that IBM will not do this migration from 22-nanometer (on which its current POWER8 processors are built) in 2016 and will instead not do so until 2017. If anything, it would appear that IBM will fall further behind Intel in terms of chip manufacturing technology in server chips.
Intel already outgunned IBM at 22-nanometer without a "process lead"
Interestingly, both IBM and Intel have been shipping 22-nanometer products in the marketplace, so on the surface it would appear that there is no real gap between the two in terms of manufacturing technology.
Back in November, however, Hibben argued that because Intel implemented the FinFET transistor structure in its 22-nanometer manufacturing technology while IBM hadn't, Intel still had a chip manufacturing edge.
However, there is actually a problem with this analysis. IBM's 22-nanometer process used a technology known as Silicon on Insulator, or SOI, a technology not utilized by Intel in its 22-nanometer technology. Interestingly, both Fully Depleted Silicon on Insulator and FinFETs represent two different approaches to trying to solve the same problem: reducing leakage power.
I would argue at the 22-nanometer generation, both IBM and Intel were effectively at "process parity." This suggests, then, that Intel "won" that particular battle on chip architecture/design and other business factors rather than necessarily chip manufacturing technology.
Intel should gain a process advantage over IBM soon, though
According to an IBM processor road map published in The Next Platform (via Real World Tech forums), IBM plans to introduce a 14-nanometer POWER server processor in the 2017 time frame and a 10-nanometer POWER server chip in the 2020 time frame.
Intel is expected to roll out its first generation of multisocket 14-nanometer Xeon processors in the first half of 2016, and it is generally expected that its second-generation parts will come in the first half of 2017. Assuming that Intel doesn't introduce a server processor family based on its Kaby Lake architecture, we should see Intel transition to 10-nanometer server processors in the first half of 2018.
If Intel does two generations of 10-nanometer server parts, we should see a transition to 7-nanometer server parts in the 2020 time frame; if it does three generations, then I'd expect them in the 2021 time frame.
At any rate, it would appear to me that Intel should be able to gain a manufacturing edge over IBM in server processors and sustain it from here on out.
Don't fret too much about the IBM "threat"
There are many server processor vendors trying to get a piece of Intel's very lucrative data center business, including IBM. Although IBM has been increasingly vocal about its efforts, I continue to believe that Intel will be able to successfully defend itself against any potential incursions from IBM.