Image source: Intel.
For quite a while, chip giant Intel (INTC 5.68%) has been talking up a next-generation non-volatile memory technology known as 3DXPoint. The technology is billed as an entirely new class of memory -- faster and more durable than traditional NAND flash memory while much cheaper than DRAM.
At Intel's 2015 investor meeting, the company claimed that the total addressable market, or TAM, for 3DXPoint memory modules (that is, 3DXPoint being used as a substitute for DRAM rather than as a substitute for NAND flash as part of a storage drive) could be worth approximately $34 billion by 2020.
Intel had previously said that these 3DXPoint memory modules would be supported on a "future Intel Xeon processor." Based on the leaks around Intel's upcoming Purley server platform, it was expected that the future processor that Intel was talking about was Skylake-EP, a product that's slated to launch in the first half of 2017.
However, on Intel's most recent earnings call, CEO Brian Krzanich indicated that this wouldn't be the case.
3DXPoint memory modules only for second-gen Purley
Whenever Intel releases a server platform, it usually supports that same basic platform for two processor generations. It usually takes anywhere from a year to a year and a half for Intel to come out with new processor generations, so in practice one can expect an Intel server platform to remain current for about three years before being replaced.
With that in mind, it's worth looking at what Krzanich had to say about Purley and 3DXPoint memory module support (transcript from Seeking Alpha):
On the Purley platform we're actually starting to sample those products already to some of the leading-edge customers. And they're seeing not only just an overall TCO performance advantage that we typically see with each one of these but this also continues our integration of things like the omni-path fabric, it has more integration of the silicon photonics so it's still the adjacency functions that are quite strong, and they will get more and better as we go through each one of these. There will be a second generation of Purley that includes 3DXPoint.
According to this, the "future Xeon processor" that will support 3DXPoint memory modules will not be the upcoming Skylake-EP, but instead its successor, known as Cannonlake-EP.
When will Intel start selling these 3DXPoint memory modules?
If Intel's upcoming 3DXPoint memory modules require Cannonlake-EP to work, then investors should realistically expect that Intel won't be selling those modules until either late 2018 or in the first half of 2019. I'd be pleasantly surprised if it were the former, as I'm expecting the latter.
This suggests that any revenue contribution from these memory modules on Intel's part probably won't be seen until that time.
What does this mean for investors?
At the end of the day, Intel's data-center group brings in most of its revenue from the sale of processors and related chipset components, which Intel refers to as "platform" products. Last quarter, just over 8% of its data-center revenue came from non-platform products.
The best way to think of technologies like 3DXPoint memory modules, silicon photonics, and these other non-platform products that Intel talks about is as "nice to have" technologies. Obviously, investors should welcome every last bit of profitable revenue growth with open arms, but the core of the business has been and will continue to be sales of platform components.
If the platforms continue to do well, Intel's overall data-center business should do well -- robust non-platform sales in this case would serve as a cherry on top. If platform sales disappoint, then it's unlikely that any growth from the non-platform components, like 3DXPoint memory modules, will generate enough revenue to ultimately matter.
