For nearly three years, I’ve been predicting that 2017 was the year for flash/solid state disks (SSD) to leave hard disk drives (HDDs) in the dust. Now that it has finally come to pass, we are hit by a perfect storm on the flash die supply side. With a transition to 3D NAND running late, and high demand in all sectors from smartphones to servers, flash die supply is very tight today. Prices are rising by 20% or more and shortages in segments like PCs are impacting sales.
It’s taken a while for IT staff to accept the obvious. SSDs are a much faster solution for computer storage than spinning rust. We are talking huge factors, around 1000x or even more in performance, so you might wonder what took them so long. I guess price was an issue, though only the wise realized that a given server could do twice as much work with SSDs, saving more than any SSD premiums might cost.
Still, it’s generally recognized that “enterprise” HDDs are no longer the right choice for primary storage. Further, we are on the edge of accepting that the rest of the HDD market, mainly SATA drives, are also challenged by SSDs as 3D NAND allows capacities per drive in the 30 to 100TB range.
There’s the first hard spot! 3D NAND requires a retooling of the wafer fab lines and this has taken longer than expected. Most vendors plan to come on-stream with 3D NAND in the second half of 2017, followed by some months of qualification testing. Since they stopped making 2D NAND die to retool, this has created a whopping supply gap.
This occurred just as SSD demand for servers surged tremendously, even as RAID storage and data appliances such as All-Flash arrays added large demand increases. If this isn’t enough, the smartphone business also surged and added to this is the drama behind Samsung’s Note 7, with substantial callbacks and an attempted restart adding unplanned demand.
To add to the problem, the weak PC desktop, laptop, and tablet segment also discovered the joys of SSD. An attempt to at least stabilize double-digit declines in year-year sales hit the market with SSD-based units the primary focus. A major sales campaign is in danger of sputtering out as prices for SSDs increase and supply shortens.
PC-class products could be the first serious casualty of the tight supply market. This was probably the last hurrah in a segment that is way down from its peak of five years ago and sinking fast. Supply problems will mean higher list prices and much less in the way of bargain pricing and we may yet see a limited supply of notebooks, desktops, and tablets bringing segment revenue down a great deal.
The server market is in a full-speed transition to SSD and flash. Most of the big system vendors get preference on drive availability, simply because they are huge customers for storage devices. Still, Hewlett Packard Enterprise (HPE) is now putting many drives on limited availability, with resulting server delivery windows sliding out by as much as 10 weeks.
So what is happening to the drive vendors, whose delivery schedule underpins all of this? There are two vendor classes for SSDs, characterized by having foundry capacity or not. In the foundry owning class we find Western Digital, Intel, Micron and Samsung, the largest vendor by far. In the foundry-less class is Seagate, who were boasting about their agility in the spot market a year ago, together with companies such as Kingston and PNY. These companies are struggling to find die and the shortage may well impact their quarterly results. Toshiba has substantial foundry capacity, but the company is warning of serious financial problems related to its reactor business and could be a risky supplier.
In the online marketplace, the supply issues are very noticeable. The line cards are limited to a few drives from each vendor, which are typically small drives (256 GB or less). This contrasts with 3Q 2016, when 500GB, 960GB and 1TB drives were common. The same picture is true in distribution.
Drive prices have risen between 10% and 20% and suppliers are extending lead-times to distributors out into the 20+ week range, so further increases are possible. This isn’t likely to mess up the long-term dominance of SSD, since the industry confidence in 3D NAND delivery in 1Q 2017 is well-placed. Some parts are already sampling for quality testing.
The shortage issue should be far less serious as we enter 2018 and on-streaming of further foundry capacity next year will continue to improve the picture. This will bring a swift end to the “enterprise hard drive” class of products and lead to the huge bulk storage SSDs (likely using quad-bit cells, QLC) that have been announced. As the transition to SSDs continues, the rate of decline in RAID/hybrid arrays will accelerate as we move to compact storage appliances and hyper-converged infrastructure.
There will, however, be other casualties. PCs and desktops will miss a revival and I predict accelerated decline by the end of the year. Tablets are less sensitive to flash shortages. There will be some increases in price and some availability problems, but the tablet fits the new compute-in-the-cloud paradigm and so will rebound in 2018 to absorb the losses in the PC/desktop segments.
There may also be some shake-out in the SSD supplier base, with the foundry-based surging in revenue and profit, while the foundry-less flounder.