Update on NAND and SSD Oversupply – Trendfocus

Through Korea and Japan two weeks ago, the oversupply condition plaguing NAND and SSD suppliers shows few signs of improving near-term, according to Trendfocus, Inc.

The general opinion remains that everyone is hopeful that we will see some uptick in demand in 2CH19, however, the key word there is ‘hopeful‘. 

The market where most companies are pinning their positive attitude is with hyperscale companies, especially in the US.

With that said: pricing remains drastically low. Opinions vary on where the pricing may go throughout this year – some saying they can even see an increase in pricing in some areas.

If history teaches us anything, this is unlikely to happen except in the event of a natural disaster, which no one would ever want to occur, or multiple factors happening at once where supply all of a sudden dries up (remember when phones doubled in capacity, PC capacity mix increased, and most all vendors were delayed in ramping 3D NAND) – we do not see something like this happening in the short or mid-term. Noting this, pricing remains aggressive throughout the year.

As it relates to SSDs, even with the down quarter in 4CQ18, we can assume this decline will continue in favor of continued transitions to PCIe at various companies. SAS SSDs, on the other hand, even though suffering a slight decline in the quarter – the overall opinion has been that these offerings will continue to see healthy shipments throughout the forecast period – with some leveling off in the longer-term forecast. PCIe, although showing the largest decline in the quarter, is a segment everyone agreed is the future for growth opportunities.  When it comes to self-builds (DIY), the latest pricing environment has changed some opinions in the market. What was once a huge growth market due to potential cost savings versus finished goods SSDs, has some companies backing off their own DIY builds in the future.
 
96-layer NAND – really?
Inventory is out there. Now, does the NAND industry really push towards 96-layer as quickly as some players claim? Yes, the density does support continuous cost reductions to sustain price declines, but what to do with all of that 64-layer stuff that is just sitting there waiting for homes at customers? Wafer fab expansions have ground to a veritable halt for sure, but a real, true push towards having customers qualify 96-layer NAND may be discouraged by NAND vendors until the pile of inventory is depleted. Short-term, a pool of red ink is expected on some NAND vendors’ financial statements in the coming months …
 
Intel continues to flog Optane for client PCs – decent benchmark results mask underlying problems with write-intensive workloads.

Optane is a solution waiting for a problem in the client PC space. Obvious? Maybe. Used as an HDD accelerator, Optane has developed a niche market and using the word ‘niche’ is probably overstating its size. Now Intel wants to push Optane plus QLC, a combination that promises to enable higher capacities and better performance at costs competitive to higher performance TLC-only solutions. The problem is that while top-level testing benchmarks look good, heavily write-intensive workloads, especially long sequential writes have yielded performance that falls below that of … HDD. Whether a user will experience that type of slow performance depends heavily on how the PC is used, but since the Optane/QLC solution is meant to target commercial PC customers, it is a risk for PC companies to deploy. If it appears, it may be another niche solution.
 
QLC – it takes more than cheap NAND to make a good cheap SSD.
QLC NAND (four bits per cell) has raised the discussion of potentially replacing some high capacity HDD solutions in some select applications, especially in consumer PCs once 512GB capacities approach the price point of today’s 1TB client HDDs. Slow write speeds and low endurance compared to TLC SSDs still improve upon that of spinning disks and will likely provide the SSD industry that last push to displace HDDs in most PCs within the next few years. QLC may even penetrate low-end commercial PCs that are more price sensitive than performance-critical, but this is less likely compared to the consumer market opportunity.

The problem for PC companies right now is that QLC SSDs … well, perform somewhere in the “largely unacceptable” to “well, thanks for the test samples, but let me know when you have some real products” range. Performance of many available solutions falls below lowered specifications and essentially all QLC client SSD offerings fail to the point of falling out of consideration. It takes the name recognition of a big CPU company to push even the possibility of qualifying a sub-standard solution at a PC OEM. Later this year, though, a certain dominant SSD player expects a much-improved QLC SSD that may finally meet consumer-grade PC specs. Once again, controller design, NAND behavior and a substantial SLC cache requires extensive engineering work to combine for an adequate low-cost SSD product.