The “memory” market generally gets lumped together, but more recently, there has been quite a divergence in fortunes between the DRAM (Dynamic Random Access Memory) and NAND Flash (non-volatile memory) players – at least from a business perspective (all their shares seem t move in tandem!). Look no further than the recent earnings reports of SK Hynix (OTC:OTC:HXSCL) and Western Digital (Nasdaq:WDC) for proof of that.
Demand for both NAND and DRAM was supposed to remain robust over the long term due to “computing everywhere” trends in cloud and artificial intelligence, as both storage (NAND) and persistent memory (DRAM) will be in high demand from new technological applications.
Despite NAND Flash likely being in more demand going forward (40% growth, per long-term estimates) than DRAM (~20% annual demand growth), DRAM is currently the much better business due to its supply side situation. This is both from an industry makeup standpoint as well as a technological standpoint.
The warning from Western Digital
Western Digital recently reported a down quarter, with its non-GAAP EPS of $3.04, down from the previous quarter’s $3.61. It has been a bruising year for NAND Flash prices, but it’s about to get worse, as the company guided for the next quarter’s non-GAAP EPS to be cut in half to $1.45-$1.65. The stock fell another 20% and is now trading in the $40s after hitting $100-plus in the spring.
It was a fairly bad quarter and quarter-out guide. Western Digital said that it would reduce its wafer capacity by 10%-15% by next year, even though bit growth will increase 34-36% next year (in line with the company’s expectations) as it completes node transitions to 64 and 96-layer NAND Flash, which will increase efficiency.
SK Hynix, on the other hand…
SK Hynix, however, has a similar mix of DRAM and NAND Flash as Micron (Nasdaq:MU), which I have written on a bit recently. It’s recent earnings and guidance were anything but disastrous, and after a bad October, the stock actually rose after its earnings release.
SK Hynix managed to grow revenue 10% QoQ and 41% YOY, and expand operating income 16% QoQ and 61% YoY. This is in stark contrast to Western Digital, which saw its revenue decline (2%) QoQ and down (3%) YoY, while operating income declined (15%) QoQ and (21%) YoY.
That’s a pretty huge difference in outlook, and the difference lies in SK Hynix’s 80% exposure (by revenue) to DRAM.
All About Supply
Over the recent past (since about 2014, with the exception of 2016 the DRAM industry has been much steadier and more profitable, despite NAND experiencing twice the bit growth rate over that time.
Source: Micron Annual Reports. Year end is June 30. Chart by author.
The difference is two fold. One, the DRAM industry, which used to be the problem segment and most volatile, has consolidated to just three players. On the other hand, there are five (but really sort of seven) players in the NAND Flash industry – Samsung (OTC:OTC:SSNLF), SK Hynix, Western Digital/Toshiba Memory Corporation (owned by a Bain Capital consortium, with Seagate (Nasdaq:STX) as a minority investor), Micron, and Intel (Nasdaq:INTC). Obviously when there are twice the number of suppliers, you can get twice the instability (or more).
However, there's also a technological component to the imbalances as well even beyond industry concentration. DRAM is actually beginning to bump up against the limits of physics and Moore's Law, making technology transitions more timely and expensive. This has had the effect of constraining growth, which is why even at a lower 20% demand growth rate, the DRAM players have had trouble keeping up with supply over the last two years, leading to rising prices.
Ironically, NAND already had bumped up against its physical limits earlier than DRAM, but the innovation of 3D NAND – in which memory modules are “stacked” on top of one another, became technically feasible and has led to another breakthrough in increasing bit density. Over the past year, the five-seven industry players have all achieved bit crossover on their 64-layer NAND techniques. When 64-layer supplants 32 layer NAND, that’s an immediate double of capacity per chip.
Of course, the process of 3D NAND is very difficult and more time consuming than planar NAND because of high flatness requirements and exact precision needed to get yields up. However, once the process is perfected, a flood of new bits come on the market.
The unevenness of output, along with huge expected demand, has caused an oversupply in the market just as demand is hitting an air pocket.
That “air-pocket” has come due to two reasons: One, many of the cloud hyper-scalers spent a huge amount over 2016 and 2017 in building out their data center footprints, and are now going through a digestion period as cloud customers fill up all those data centers with data and applications. At the same time, rising interest rates and the worsening trade war with China has caused some customers to be extra cautious in their capex spending and production until the smoke clears (maybe after the mid-term elections?).
At any rate, if we are coming off of a year when exabytes shipped grew 55% (as it did in 2018, according to Western Digital) and you expect demand of 40%-45% next year, but instead it comes in at 36%, that can yield a huge gap in supply/demand dynamics, as the NAND Flash guys are all learning now.
Ironically, if demand is growing at a lower rate (such as the 20% for DRAM), but is more consistent and with smaller standard deviations, it makes for a much easier time keeping supply-demand in balance, even beyond the industry concentration factor.
That’s why, while NAND prices seem to be cratering (even beyond the figure given in the chart above), SK Hynix’s management said that DRAM prices should only remain flat or slightly decline during this “downturn.” While management could be wildly off, that’s a strikingly small decline in price, especially since all of these companies are admitting that the NAND situation is a problem. Per the SK's management:
“So regarding the DRAM price, as was explained earlier, there has been a slowdown in the DRAM price increase throughout the year this year. And we believe that this trend will continue into the fourth quarter and the first quarter of next year. So we do not expect any sharp decline. But then, it's likely to remain flat or have a small decline in the price. But it is difficult to - for us to give you a range of the decline… And then for the second half of next year, again although it is difficult to predict at this point, the price is likely to remain flat or turn around to increase… And the basis for this assumption is that for the data center big players, now their demand bit growth in the first half of next year will be in the single digit percentage. But then in the second half of next year, it will be up to 2-digit percentage…”
Source: SK Hynix Q3 2018 Transcript
Can NAND Turn Around and Become a Better Industry?
While the DRAM players are in much better position today, and should be for the foreseeable future, when one looks a couple years out, it’s not unthinkable that the NAND industry could actually improve significantly and become even quasi-inflationary like DRAM has been the last two years.
How could this happen? Especially when there are seemingly huge amounts of supply that can be brought on to overwhelm even the high demand of the cloud era (as we’ve seen this year?).
A few ways this can happen. First, remember that exabyte demand is growing 40% annually (though not evenly) for the foreseeable future. That's a lot of demand! One, each layer of 3D NAND becomes increasingly difficult and also gives you less “bang for the stack” than the layer before it. So, 64 layer doubled the bits of 32 layers, but 96 layer will only increase the bit density by 50%. 128 layer will only increase the density by 33% over 96 layers. So while planar NAND has Moore’s Law, it’s not unthinkable that 3D NAND could hit Moore’s Law Limits again the more vertical it gets.
At the same time, the cratering price of NAND should stimulate demand, with more storage present in smartphones, and with Flash more rapidly displacing hard disk drives in PCs. While it was a minority of PC storage a few years ago, NAND Flash is now up to around 60% penetration in PCs, and will likely reach 100% in due time as long as the price per bit keeps falling and performance requirements go up.
So, as the industry moves from 64-layer to 96-layer, supply won't grow as fast. Next year, as the industry cuts back on near-term capacity expansions (Samsung, Western Digital, SK Hynix, and Micron have all said they were dialing back NAND spend in Q4 and/or next year), and as NAND’s demand elasticity kicks in, returning demand to its long-term 40% growth rate, and you have the makings of a possibility for the NAND cycle could turn very positively in about a year.
Add in the possibility for even more consolidation in NAND – Bain Capital is a PE firm after all, and the company hinted as much in an interview last summer regarding the Toshiba acquisition – and you could have the makings of a future when the NAND industry settles down into steadier profitability, not unlike DRAM today.
As for now, the DRAM players – which themselves have NAND exposure as well - are in a much better position. But one shouldn’t give up totally on the NAND players – the people at Bain Capital, who just spent $18 billion for Toshiba’s Flash Memory unit in June, aren’t stupid, after all.
Disclosure: I am/we are long MU, WDC.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: All information contained herein is provided “as is” and Billy Duberstein expressly disclaims making any express or implied warranties with respect to the fitness of the information contained herein for any particular usage, application or purpose. Any information, opinions, research or thoughts presented are not specific advice as I do not have full knowledge of your circumstances. Prior to making any investment decision you should consult with professional financial, legal and tax advisors to determine the appropriateness of the risks associated with such an investment. No assurance can be given that the objectives of a particular investment will be achieved or that an investor will receive a return of all or part of his or her investment. In no event shall Mr. Duberstein be responsible or liable for the correctness of any material used herein or for any damage or lost opportunities resulting from the use of such material. The information contained herein may not be copied, reproduced, published or distributed in any way without the prior written consent of Mr. Duberstein. Mr. Duberstein and the terms, logos and marks included herein that identify Mr. Duberstein 's services and products are proprietary materials. The use of such terms, logos and marks without the express written consent of Mr. Duberstein is strictly prohibited.
Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.