You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!

2016 & 2017 outlook: While DRAM declines NAND seems immune

Robert Maire

Moderator
As NAND pricing falls - units continue to increase. Can NAND demand grow to infinity and beyond?

Memory is the ultimate commodity in the semiconductor market and thus tends to behave like many other commodity markets. Pricing is inversely proportional to demand. Moore's law relentless price declines in NAND memory appear to absorbed by seemingly limitless demand. DRAM has more fundamental limitations on demand due to its usage and thus has "hit the wall" as lowered pricing and increased supply was not sucked up due to finite market demand.

The question at hand is exactly how limitless is NAND demand?

Gigabytes and Terabytes....
DRAM memory is measured in Gigabytes, with smartphones at 1-2 GB and laptops/desktops at 4-16 GB. Servers can have tens of GB.

NAND memory starts out on the most paltry smartphones at 8-16GB up to Apples 64-128GB. SSD (Solid State Drives) start out at 32GB for cheapo entry level laptops and more typically is around 256GB up to 512GB for most laptops and fairly typical of Apple laptops that seemingly lead the SSD revolution.

We have seen 1TB SSDs advertised as low as $200 while the cheapest rotating 1TB drives are around $50. For low end applications we have seen 100GB SSDs for about $35, which is at or below the cost point of manufacturing rotating media..

SSDs are the bottomless pit of demand for NAND...
Given the above numbers, it seems clear that while smartphones and tablets have driven NAND demand in the past, it is now SSD demand that is throwing gasoline on the bonfire of NAND demand and pricing.

Had SSD driven NAND demand not happened we would have hit the wall a long time ago, probably even before DRAM hit it. This suggests that the bottomless pit of demand we have to estimate is that of SSDs.

SSDs are still not at price parity with rotating media.....yet.....
On a pure price basis, rotating media is still cheaper than SSDs but on a price/performance trade off of battery life and access speed we likely crossed that threshold a while ago. I haven't bought a laptop with rotating media in several years now.

The only question now is just how big an SSD. Doing the math in the Apple store, its clear that Apple makes like 90% + gross margin on upgrading SSDs from baseline.

Given the current price curve of NAND memory we will hit price parity with rotating media in the not too distant future.

3D NAND accelerates the pricing curve slope
3D NAND holds out the opportunity to see an acceleration in the price declines of NAND given the increased density without a similar corresponding manufacturing cost increase. Essentially 3D NAND is a dislocation in the price curve providing for a sharp drop off in cost as compared to a more linear Moore's law driven drop.

3D NAND fabs when at full tilt will likely generate a tsunami of NAND memory at lower pricing that will get us close enough to rotating media price parity which should in turn generate another surge in NAND demand as essentially all laptops go solid state. NAND will essentially just displace rotating media on laptops.

Servers are not far behind and some may say ahead as performance matters more in the cloud as speed enables new applications.

Waves of NAND demand.....
The first waves of NAND demand Ipods, cameras, smart phones etc; with the second wave being SSDs for laptops and performance servers. The third wave is likely the death knell of rotating media all together. When NAND gets competitive for even the biggest, dumbest, slowest, archival storage in the cloud or clunky desktops, then its all over. We are likely a way off from that third wave of demand for NAND. Its going to be tough to catch up with big rotating media as disk drive pricing have followed their own cost curve not that much different from Moore's law that governs semiconductors.

Miscellaneous Demand for NAND
Those ubiquitous USB "thumb drives " and SD cards, holding anywhere from a Gigabyte on up to 128GB or 256GB. We think this demand will not likely expand as quickly as it had in the past. You could easily take enough pictures from a digital camera to fill a 4GB SD card but 128GB seems difficult to fill. Its also too much for random files to be uploaded and downloaded between computers, in a "sneakernet" application, otherwise you will be there all day waiting for the files. We think that as prices have dropped everyone has gone to larger cards which are harder to fill up and thus will be replaced less often. We recently purchased a number of 128GB USB drives for under a $30 price point and have yet to even use a fraction of the capacity. We did purchase some micro SD 128GB drives that we have permanently installed in the slots of laptops as overflow storage.

Bottom line...we see this market slowing as the average capacity increases

Disposable NAND Drives?
As the price points continue to drop, we would not be surprised to see movies & software and other large files distributed on USB drives or SD cards as fewer computers have DVD drives and downloading a full resolution 4K movie is still a lot of data to move over the internet. While we are not quite at the pennies per unit of a DVD, the ergonomics and flexibility of NAND based distribution could prove very attractive and the lack of DVD drives could make it a necessity.

Low price creates new demand for NAND storage-
Back in the 1990's we were the lead analyst on the IPO of Sandisk, at the time a spin out IPO from Seagate who wanted out of the new technology. At that time storage was measured in megabytes and use in digital cameras was still a glimmer in Sandisk's eye. Investors laughed at the mere suggestion that NAND could be used for "film storage" for digital cameras. NAND was so expensive that pictures were stored by the camera then downloaded to the computer and wiped off so the "compact flash" card could be used over and over. Today many pictures can be stored on a penny's worth of NAND......the rest is history. The investors who laughed the loudest during the Sandisk IPO were shareholders of Kodak......

Moore's Law caused NAND to steamroller over many other applications....
Could Intel & Micron's "XPoint" cause a disruption or bump in the road of NAND? Its been a really, really long time since we have seen a new semiconductor memory technology. We are very positive on XPoint and think it could be huge as it has speed closer to DRAM while the permanence on NAND for a price point between the two. Who wouldn't want an Iphone that could back up or restore its contents in something faster than 15 minutes (or what seems like...). Servers that had near DRAM like main memory access speeds.

This is all kinds of wonderful stuff that in the long run could do to NAND what NAND is doing to rotating media.....The problem is that the ramp has been very, very slow and it could easily be 5-10 years for XPoint to dent the monster of NAND.

The short answer is it has huge potential but will not disrupt NAND in the next few years (our investment horizon)

3D NAND drivers

We are in the midst of a huge 3D NAND build out party, with Samsung as the lead reveler. The party has just started and we are in the early innings. but the industry is in a huge rush as 2D NAND technology is hitting the limits of current lithographic patterning with the lack of EUV being a limiting factor in cost and technology. 3D NAND has greatly relaxed design rules which allows for less intense lithography.

In 2D NAND you get more bits by making each cell smaller thus getting more cells crammed in per square inch. With 3D, the size of the cells remains more or less the same and just more are stacked one on top of the other much like high rises versus ranch homes.

This essentially frees 3D NAND from the time constraints of Moore's law and density can be ramped as fast as yields can be maintained.

2016 & 2017 outlook

Its clear that the balance of 2016 will see continued build out of 3D NAND capacity. We think price declines will slow as we get into Q3 as the industry starts to see NAND capacity absorbed by the Apple Iphone 7 build up.

Chip production has already started in order to be able to build enough inventory for Apples normal September launch of a new Iphone. Given that this is a major upgrade cycle and that Iphone sales have fallen in front of the Iphone 7 debut, we would assume that Apple will likely build up significant inventory for its fall launch.

Rumor has it that the new Iphone will sport up to 256GB of NAND memory (which makes sense). Given where we are with Apples custom processor, combined with perhaps some extra DRAM and more NAND the Iphone 7 will be a laptop class machine in a handheld size.

While Apple only works with certain NAND makers on a contract basis, it still take significant volume out of overall NAND supply which should help balance supply and demand and thus pricing of NAND.

However, once we get past the Iphone 7 launch and beyond the holiday season late in 2016 we will likely see a combination of a lot of 3D NAND capacity finally coming on line combined with a post partum slowdown after the initial euphoria of the Iphone 7 launch.

We would predict that we will see a re-acceleration of NAND pricing declines towards the end of 2016 as these factors kick in fully.

We should then see the normal seasonally weak memory pricing in the first part of 2017 likely be even worse than normal Q1 seasonality.

Follow the leader

The NAND industry has no choice but to follow Samsungs strong push into 3D. Anyone left behind in 2D NAND will essentially be out of business as the price disadvantage will be too big.

It is this "arms race" to get to 3D NAND that is sustaining almost the entire semiconductor equipment industry right now, and for the balance of 2016, as DRAM spending is near zero, logic spending is low and likely getting lower and foundry spending has not been as good as hoped in the transition to 10NM. Without the 3D NAND build out the semi equipment industry would be sucking wind......

This begs the question about 2017...what will drive spending as we come down off the peak spend levels in the 3D NAND conversion wave? Will DRAM make a comeback? (we doubt it) Could foundry increase spend for 7NM? (maybe..a better chance of this) We don't see logic in a come back mode in 2017.

Its still a bit early to tell but our guess is that the first half of 2017 will likely be down from the second half of 2016 overall.

The Stocks
In the near term, Micron and Sandisk should be in reasonably good shape as we think NAND pricing should hold up better going into the second half of the year. DRAM likely won't get worse and could see some light if average installed DRAM ticks up.

Intel is not moving as fast as hoped with XPoint and we now seem resigned to a slower cadence for processor progress.

Applied Materials saw blow out orders related to OLED tools (not semi equipment tools) likely related also to Iphone 7 build out will likely continue to see good numbers as 2016 plays out so there is still legs in the stock. Lam and KLA will also benefit nicely from the 3D NAND build up perhaps with some tempering related to the second request from the DOJ which at the very least is slowing things down in the merger and could soften the benefit.Samsung remains Lam's biggest customer and the industry's big spender.

ASML will continue to lag as the EUV roll out just drags on and on forever. 3D NAND is essentially a work around for the lack of litho progress. EUV benefitting the bottom line of ASML is still years away and not in the investment horizon.

Even though the stocks have been bid up, the news flow will likely continue to be positive enough to support them overall for the next few months. We would however look to lighten up when the iPone 7 party hits a crescendo some time in the fall as the hang over afterwards may not be generous to stocks in the supply chain.

So far, 3D NAND has been very, very good all around for the industry but nothing lasts forever because forever is a very long time in a fast moving industry and sooner or later gravity will take hold......so ride it out for now then watch out.....
 
Last edited:

benb

Member
There may be an emerging theme in the equipment world: For 40 years, photolithography enabled semiconductor progress, and it matured to the point there was one supplier on the leading edge. Now with 3D NAND, photo is a constant, while etch becomes enabling--how deep the etch goes determines how may layers (32 initially, now 48, with up to 128 potentially possible) and thus how dense the 3D NAND can become. Maybe the new etch-dominant world isn't fully reflected in LRCX stock price?
 
Top