David A. Chapa, Head of Competitive Intelligence, Qumulo.

Are you buying an Edsel or investing in the continuity of your business?

‘The Car of the Future’

I must be on a Ford kick for some reason, maybe it is the F-150 I have in my driveway; nonetheless, when I heard the news of Micron selling its fab to Texas Instruments and TI planning to cease all production of the 3D Xpoint wafer, I immediately thought of the Ford Edsel. For those of you who aren’t ‘into cars’, the Edsel is most famous for its failure.  More on this later.

Back story

Back in March of this year I published a blog to my LinkedIn profile, which was subsequently reposted to our corporate blog page at Qumulo called, ‘Betamax v. VHS’, where I cover the news about Micron exiting the storage class memory (SCM) business leaving Intel as the only company with any commercially vested interest. You can read Micron’s own press release where the company states, “Micron has now determined that there is insufficient market validation to justify the ongoing high levels of investments required to successfully commercialize 3D XPoint at scale, to address the evolving memory and storage needs of its customers” (boldface added for emphasis).

Remember, ‘at scale’, because this is an important part of the labyrinth many may find themselves in due to selecting a solution like VAST Data, that seemingly is tied programmatically to this technology. Based on my understanding, without Intel Optane SSDs their solution may simply not work as advertised.

So, that’s the backstory…what’s happening now?

The Micron fab in Lehi, Utah, that currently manufactures the 3D Xpoint wafers (mentioned in my previous blog), was sold to Texas Instruments for $900M.

Problem solved, right?  Wrong.

TI will scrap the 3D Xpoint manufacturing and leverage the fab for its own technologies. Micron said, in its press release, that it will be selling off some of the tooling it used for the 3D Xpoint manufacturing to help offset the loss it is taking on the fab itself, which they say has a book value of about $1.5B. This leaves Intel without a place to manufacture, at scale, the 3D Xpoint wafers for its Optane line.

Many have reported that Intel, being the sole company to actually use 3D Xpoint memory, may only have the option to manufacture the wafers. This would be at its Rio Rancho, New Mexico fab where all of the 3D Xpoint R&D is done, but this does not appear to be a fab that can do this at scale.

As Editor in Chief for AnandTech, Ryan Smith, writes in his article, Micron Sells Lehi 3D Xpoint Fab to Texas Instruments for $900M, “…for the sole company to actually use 3D XPoint memory, the sale of the Lehi fab calls into question Intel’s own 3D XPoint production plans. Having sold their share of the fab to Micron, Intel transitioned to being a customer of the world’s only 3D XPoint fab in late 2019 – an arrangement that left the fab operating in the red for Micron, as Intel’s 3D XPoint orders weren’t enough to fully utilize the fab. IM Flash jointly developed Intel’s current (second-generation) 3D XPoint memory technology as well, and it’s believed that the Lehi fab has been producing all of that memory for Intel.

So it remains to be seen just how Intel will be impacted, as the sale puts a running clock on how much longer they can buy 3D Xpoint memory from the third-party fab. Eventually Intel will need to set up their own fab – likely in Rio Rancho, New Mexico, where their 3D Xpoint R&D takes place – but so far the company hasn’t announced any such plans.”
Why didn’t Intel just buy the fab and be done with it?  Intel was in the process of bringing on a new CEO in February of 2021, Pat Gelsinger; who I have a great deal of respect for, by the way.  Shortly after his arrival, Pat announced in March that Intel would be building two new fabs in Arizona with a price tag of $20B. It was also noted in several articles that it would take at least five years for these fabs to run at scale. Based on the press release by Intel, it doesn’t appear that they will be equipped to manufacture the 3D Xpoint wafer required for the Optane line.

DANGER WILL ROBINSON!

It is five and a-half months from Micron meeting its contractual obligations for the year to Intel for the wafers and yet, there hasn’t been a definitive announcement as to where the wafers will be manufactured next.  This means the supply chain goes away, unless it is moved elsewhere, which according to many of the tech news outlets, Intel has not publicly announced.  However, if the new fabs in Arizona are to be used for this production at scale, it could take up to 60+ months for it to be at full capacity. And, based on the numbers Micron was alluding to, the demand for these wafers is relatively low, which is why they chose to exit this business to begin with.  So, it may not be a priority for Intel over the other more lucrative business lines they sell.

This should raise considerable red flags for business continuity if you’re considering a solution that is dependent on this technology.  While technically I really like the concept and capabilities of the 3D Xpoint technology in general, it is the persistent memory I find most interesting and not the SSD line.  But this is neither here nor there if the adoption is taking too long and now the supply chain may dry up. Point being, purchasing solutions based on technology that has yet to be validated in the market presents considerable risk when that technology will be on for business continuity. This is how the market works, and has worked for several decades.

Now, don’t get me wrong, I’m all for R&D and testing the waters on new technologies to see if it will meet market demands. As we all know, some of these technologies simply don’t make it either because they are too early, don’t fill a need gap, are  too expensive, or the commercial demand just isn’t panning out. There are several companies I thought had great technology that should have made it, but didn’t.  Some you probably never heard of such as Ikadega, Banyan Vines, and Nebula just to name a few. Cool technology that, for one reason or another, didn’t meet one or several of the criteria I shared earlier.  This is why I made the point of saying in my previous blog that betting the farm on a technology or a system that has yet to gain widespread adoption or validation is a poor business decision.  Sadly that’s what I believe some have done or are in the process of considering. I’m hoping this blog can shed some light on the subject for you and provide some guidance.

Where does this leave the market?

There remains a great deal of uncertainty in my opinion for companies like VAST and for the customers looking at purchasing these systems, a great deal of risk for business continuity.

If the decision is made to simply focus on the Optane persistent memory product, and scrap the SSD product, then consumers of those solutions tied to Optane SSDs will essentially be stuck in a zero-growth pattern and potentially left with no support should those companies cease to exist. Again, it is a painful place for anyone to be, but it is why science experiments should be left in a lab and not released to production.

There is good news ahead!

First, for the hundreds if not thousands of employees of the Micron fab in Lehi, Utah, it has been reported that Texas Instruments will work to retain those employees after the acquisition completes at the end of the year.  That’s great news and I am sure the families of those employees are also quite happy to hear it as well!

Second, stop what you are doing and take a long, serious look at Qumulo for your hardware independent storage solution that has been supporting NVMe for a very long time, a standard that has stood the test of time and is not going away.  One of the main design principles of the Qumulo filesystem was, and still is, that it remain independent of hardware. In other words we will not programmatically tie our solution to a particular set of hardware technology for the sake of “performance” or “features.”

On the contrary, there is no “easy-button” here at Qumulo.  We work hard to achieve the high performance and advanced features we have with our solution through software innovation that enables hardware flexibility.  For example, our performance is based on a number of factors in our filesystem design, one of which is by eliminating tree walks and increasing the overall write efficiency to NVMe.

We also don’t use NVRAM in our solution for a very simple reason.  There is no NVRAM in commodity hardware, and we don’t typically find NVRAM in the cloud.  This is another reason why we are a truly portable solution between on premises and cloud.  This is not something many legacy and new entrants to the storage market can deliver, as they rely on NVRAM. Software defined and hardware independent serves our customers’ business continuity well.

In Conclusion

Why choose Edsel as the example?  Just for its failure? Yes, failure.  Many people believe the failure of that car was due to its appearance, some would even call it ugly.  Ford spent a year marketing the vehicle, named after his son, Edsel Ford, prior to it being released as the “car of the future” in 1957.  However, it wasn’t the fact that the car was ugly, or not appealing to most that caused its failure, but rather its inability to live up to expectations created by the aggressive marketing campaign the company launched.

After about three years and losing around $350M (that’s about $3.4B in today’s money), Ford scrapped the car and finally admitted defeat.  Some say the shoddy workmanship was core to its failure, others believe it was the “advanced transmission” system that wasn’t gaining adoption that caused its demise. Either case, the car failed and perhaps we can all learn something from this mishap.

Adoption is key for any new product, solution, or technology and market adoption helps assure business continuity.  

Don’t get stuck with an Edsel. While the marketing may woo you and give you a vision of what the “future” looks like, it is incumbent upon you to take a closer look at what the reality holds for the vendor(s) you select and for the business continuity you are supporting. If the risks outweigh the benefits, as it truly seems to be in this case, then that must be the beacon that guides your decision. Where the rubber meets the road, is a solution that has been designed for business continuity now, and in the future.

Qumulo engineering purposefully designed the Qumulo File Data Platform with your future needs in mind. Our software solution offers you, the customer, freedom of choice whether you decide to deploy in your data center on hardware you select from our HCL, or in the cloud. You get the same experience, either way, and you are getting the performance, reliability, and scalability with technology that has stood the test of time.