On the 16th of March 2021, Micron announced their intention to immediately end development of 3D XPoint products and the sale of their Lehi, Utah fab currently dedicated to 3D XPoint production.
TECHunplugged was not on the investors conference call, but we have been following Intel and Micron’s 3D XPoint involvement since 2018, and covered that in our 2019 Flash Memory Industry report (free download, registration required), and we also reviewed the Micron Update Call remarks document on Micron’s website.
This is Micron’s final move with regard to 3D XPoint in their long journey with Intel. Both companies started the adventure in 2006 in a joint-venture called IMFT (IM Flash Technologies), which ended up being 100% controlled by Micron in 2019 after the 2018 joint-announcement from Micron and Intel on ceasing joint development.
In the announcement, Micron stated that based on their analysis, there is insufficient market validation to continue sustaining investments in 3D XPoint development, manufacturing and commercialization, and are thus ceasing activities in that market with immediate effect.
Micron had developed some 3D XPoint products such as the Micron X100 3D XPoint NVMe SSD (Gen-2 3D XPoint, announced in October 2019) but never made them commercially available.
Alongside with shutting down operations, Micron will be selling their fab in Lehi, Utah. The question remains whether Intel will be looking at acquiring this facility, or if it will be repurposed for other manufacturing activities, as stated by our colleagues at AnandTech.
Outcomes for Micron
From a financial perspective, this is a positive outcome for Micron. The company has been bleeding USD $ 400M a year due to underutilization of the Lehi fab, and beyond this, R&D activities and costs focused on 3D XPoint are also to take in consideration (but were not detailed by Micron).
Looking at the technology, Micron is convinced that Compute Express Link (CXL), an open standard tied to PCI Express 5.0 is the way forward to connect compute and memory subsystems. All efforts previously undertaken under the 3D XPoint division will be redirected towards developing products adhering to CXL standards.
This is however just the beginning of another journey, and it will be a few years before we eventually see CXL products hit the market. While Micron claims that CXL may delay the adoption of 3D XPoint, the cold reality is that 3D XPoint already exists in the market and is used not just for storage use cases, but also for Persistent Memory use cases (in the so-called App Direct Mode). This is the case with Oracle databases, SAP HANA and several other commercial applications used in the Enterprise.
If you want to learn more about CXL, there is a great article covering the topic on AnandTech’s website.
Impact for Intel
From a production perspective, Intel has secured capacity with Micron to maintain production levels intact. This is in line with info we had when we wrote our 2019 report, which stated that Intel had reserved manufacturing capacity at the Lehi fab.
The report further stated that Intel would be shifting their production to 100% controlled facilities by 2020-21. It could well be the case: well-known semiconductor industry analyst Jim Handy from Objective Analysis reports in his blog post on the news that Intel could ramp up the production of 3D XPoint at their Rio Rancho, New Mexico fab.
Depending on how fast Intel will be to ramp up capabilities might determine if they will be even looking at purchasing the Lehi fab or not, and it is also not known under which provisions the “reserving manufacturing capacity” will be transcribed in its practicalities: a potential Lehi buyer might be contractually bound to continue providing capacity to Intel for several quarters.
The Future of 3D XPoint
Perhaps the biggest question raised is what with the future of 3D XPoint. Intel committed to 3D XPoint recently (in December 2020) when it was announced that the 3D NAND business would be divested to SK Hynix. At this point in time, we still consider that the continued development of 3D XPoint is strategic for Intel, especially at is allows the company to provide more value to customers looking at differentiators with Intel processors vs. AMD.
On the other hand, the change at the helm of Intel, with the return of Pat Gelsinger may (un)expectedly shake up things in the months to come, with a potential massive refocus on processor development – an eventuality which is seriously considered by most industry analysts and that cannot be left out of the bigger picture.
To learn more, we invite you to read the excellent analysis post from our industry friend Chris Evans. Chris has gone into great detail to document all of the potential technological and strategical implications of this announcement not only for Micron, but also for Intel, with a realistic approach on future options.
Finally, our friends at Gestalt IT also recorded one of their Gestalt IT Rundown episodes covering the announcement. Our industry friend Stephen Foskett has a great slot covering the news in detail, covers options for Intel and gives his thoughts on the matter.
Two years after parting ways with Intel, and two years without a commercially available 3D XPoint product (if we exclude the limited release of the X100 to select customers), Micron’s decision doesn’t comes as a surprise, if only for the financial reasons stated earlier.
And while some in the IT press see that announcement as a major blow for Intel, we are of the opinion that this move doesn’t impacts Intel. The divorce between Intel and Micron was already an established fact in July 2018, when both companies announced the dissolution of their partnership, and Intel is well off on its own, manufacturing, selling and showcasing the positive impact of 3D XPoint on some of their customers / ecosystem partners (we especially recommend the use case sessions from Storage Field Day 21).
Subsequently, commercial 3D XPoint (under the Intel Optane brand) is very tightly coupled with Intel x86 hardware (notably its processors) and especially on the Persistent Memory part, where it is de facto a proprietary standard.
That is an area where Micron would find it hard to compete, not just because of the current lack of traction in the market (though use cases are emerging), but because they do not have the adjacent ecosystem that Intel has to provide value to the customer, other than compete on pricing (and eventually performance, although at fractional levels).
Regardless, what matters here is Micron strategic decision to exit 3D XPoint and focus on CXL, a decision that puts the final point on a 15 year long journey. This refocus allows Micron to stop burning cash exit a niche market where they were already cornered, and refocus their efforts and R&D towards a potentially promising technology. This time without a partner in the journey, or so it seems.