If you’re an industry analyst like The SSD Guy you find that useful data can come from unexpected places. When Intel announced the sale of its NAND flash business to SK hynix this month the company also provided useful information about Optane SSD sales. You have to know a little bit about Intel’s organizational structure to understand. Let me explain.
The SSD part of Intel’s Optane business has been assigned to the company’s Non-volatile Solutions Group – NSG. The balance of NSG’s revenues come from NAND flash SSDs and a tiny bit of non-SSD NAND flash sales, both of which have been sold to SK hynix. The Optane DIMM part of the business (officially: “The Optane DC Persistent Memory Module”) belongs to DCG, the Data Center Group. Since Optane DIMMs only work with that organization’s data center processors, the “2nd-Generation Intel Xeon Scalable processor” and later, then it makes sense for the DIMMs to be assigned to DCG.
Intel did not sell the Optane part of NSG’s business to SK hynix.
Intel’s quarterly earnings statements spell out revenues and operating income for each of these groups, so we know NSG’s NAND-plus-Optane-SSDs revenues and operating income. In the press release about the acquisition Intel revealed what the NAND business alone contributed to Intel’s revenues and margins:
For the first six months ended June 27, 2020, the NAND businesses represented approximately US $2.8 billion of the revenue for Intel`s Non-volatile Memory Solutions Group (NSG) and contributed approximately US $600 million to NSG operating income.
Subtracting these numbers from the NSG numbers reported in Intel’s first and second quarter earnings statements will then provide us with Optane SSD revenues and operating income for the first half of 2020:
NSG Reported | Press Release | Difference | |
First half revenues | $2,997 million | “appx. $2.8 billion” | $200 million |
First half operating income | $256 million | “appx. $600 million” | ($344 million) |
Optane SSDs made $200 million in revenues for the company during the first half of 2020 at a $344 million loss.
If we assume that the average Optane SSD price is $100 (averaged over everything from little $45 16GB Optane Memory SSDs to big $1,000+ 375GB DC P4800X data center SSDs), then we get 2 million SSDs. If we assume a smaller average price (i.e. more Optane Memory SSDs shipped and fewer data center SSDs) then the unit shipments would be higher. If we assume that more data center SSDs shipped then unit shipments would be lower.
Back in Intel’s second quarter 2019 earnings call over a year ago management boasted that over one million Optane Memory modules had shipped, which the company said was a breakthrough. This seems to hint that the ASP we should apply would be on the low side and unit shipments might be higher. It’s natural to assume that unit shipments have increased since then.
On the other hand, a casual web search for Optane SSDs reveals that a number of Optane Memory SSDs are for sale on eBay for prices below $20. This seems to imply that end-user acceptance has not materialized and that these sellers are removing the Optane Memory SSDs from their PCs to free up a PCIe socket for other uses, and that, in turn, indicates that perhaps the ASP we use should be higher, meaning that units would be lower.
I don’t currently have a better number for the ASP, but the rough $100 figure gives us 2 million units for the first half, and I would guess that this number would be off by no more than a factor of 2 in either direction (4 million to 1 million units). That gives us an annual estimate that ranges between 8 million and 2 million units. About 280 million SSDs shipped in 2019, so the Optane share would be between 1-3% of the total.
About Those Losses
SSD Guy readers who also read the posts of my alter-ego The Memory Guy will have noted that The Memory Guy estimates Optane losses and has often explained why Intel must sell this new memory at a loss in order to make Optane succeed. In a nutshell, the company needs for unit volumes to grow high enough that the economies of scale will drive down production costs. Intel is very willing to make that investment to boost its Xeon processor sales.
This is why I didn’t make any noise about Intel’s $344 million in losses on only $200 million in revenues. This number is actually pretty low compared to the company’s Optane losses in 2017 & 2018.
Sometimes I quip that Optane is a “Proprietary product that sells at commodity prices.” That makes it the worst of both worlds!
Optane DIMMs are a key part of Intel’s Xeon strategy, though, and it makes great business sense for Intel to be underwriting these losses if you look at the broader picture. According to Intel, Xeon-plus-Optane out-performs equivalent AMD processors partly because Optane brings persistence closer to the processor.
For the first half of 2020 The Memory Guy’s rough estimate of Optane losses was $269 million in the first quarter and $31 million in Q2, which sums up to $300M. Given that he used a technique that was pretty oblique, The Memory Guy is gloating at how close his $300M came to Intel’s actual loss of $344M. In fact, it’s getting kind of hard to be in the same room as the guy lately.
The kind of analysis you see in this post is something that my company, Objective Analysis, does every day for our clients. We provide important insight into your competitors, your suppliers, your customers, and the market in general. Please contact us at www.Objective-Analysis.com to learn more about what we can do to help your business to thrive.
Don’t forget that Intel also has a revenue share deal with WD on their SAS SSDs. This licensing is unlikely to be part of the SSD Revenue and probably should be subtracted from the Optane revenue line. It also is likely very high profit, making the Optane losses worse than your calculation which is already disastrous. Spend $3 to earn $1.
Tip,
That’s good thinking. Let’s see what it gives us.
WDC’s SSD sales under the Intel arrangement, prior to the SanDisk acquisition, were around $250 million a quarter and on a trajectory to have reached only about $500 million by the end of 2019. As I understand it, this was nearly all from the company’s SAS version of the Intel SSD design. A reasonable licensing deal would have cost WDC no more than 5% of that, although it was probably much less since WDC brought a lot of valuable IP to the deal.
That $25 million per quarter would have cut out about $50 million from Intel’s Optane SSD loss the first half of this year. That implies that actual Optane losses in the first half of this year could have been $50 + $344 = $394 million.
Thanks for pointing that out!
Jim