Horizon September 2017 Storage Market Brief
In researching for my monthly storage update, I talked to a variety of industry insiders with varying opinions and perspectives on the storage market. There is no overriding metric I can reference to discern the overall health and/or direction of the market. The closest that I come on a quantitative level is when OEMs report on their results or trending of market pricing, such as for NAND, using a resource such as DramExhcange. Taken at face value, this information can be misleading as this never tells the whole story because there are so many nuances that can mean different things for different market participants. Trying to size up Q4 is challenging as there are a variety of market drivers in play, including a NAND shortage, a transition to TLC-based NAND and an overall movement of data into the Cloud. When talking to an analyst that I very much respect, who I asked about the overall health of the storage market, he replied that “I don’t think it’s a bad server or storage market, rather I think it’s a bad HDD market”. A comment that can be parsed in a number of ways.
Last month we talked about a slowing Q3 market. It seems that this is now official, with one industry executive telling me that it is the slowest quarter he has seen in 6 years. Overall, business was flat from an exabyte perspective, driven by a lack of significant increase in builds within the big 8 cloud providers and consolidation within the market, as evidenced by the acquisitions of Tegile by WD and Nimble by HPE. This has impacted on the supply demand equation for both SSD and HDD sales.
Broadcom, in reporting its Q3 2017 results, gave a dreary forecast for Q4 HDD business. “However, as noted in our last earnings call, we do not believe this strength to be sustainable in hard-disk drive and, sure enough, we expect a sharp decline in demand for our hard-disk drive products in the fourth quarter, driven by the start of an anticipated correction in a hard-disk drive market. On the other hand, we expect our server and storage connectivity business to start to benefit from the launch as it starts to ramp during the quarter. So, despite these anticipated sharp corrections in hard-disk drive year-on-year, we expect this enterprise storage segment to show double-digit growth.”
Is the sky falling on our heads? Not really, this is has been expected as the storage market increases adoption of SSD. There are two key variables at play here. These are the well-worn topics of the movement to the Cloud and the transition to SSD. We all know that both are inevitable but what is hard to understand is how quickly this is happening and to what degree.
It is difficult to precisely quantify SSD adoption within the backdrop of an SSD shortage, but it appears that we are at a tipping point and adoption is accelerating. Reading Broadcom’s statement above, they are quite clear that they expect SSD demand to grow and HDD demand to fall off. Although HDD petabyte sales continue to outstrip SSD by a long shot, SSD is closing the gap.
Albeit declining. Last month we talked about the reasons why the legacy storage market was going to continue to be in play. Although this is a viable market, it is one with a limited lifespan. Let’s not kid ourselves, the action is within the data centers. But what will that look like? Read all the hype and one might think that all the world’s data will be stored with a handful of data centers. IDC reports that the data shows that public Cloud investments are growing, but it also predicts that, even by 2020, more than 50 percent of spending on IT infrastructure will remain on traditional data centers, private Clouds and/or hybrid models. There are a number of reasons for this, including concerns around organizations not being in control of their data (security), the transition to the Cloud being difficult and questions around the governance of data, to name a few of the reasons holding end users back from moving all of their storage onto the Cloud. Organizations such as Microsoft recognize that companies still have questions around the Cloud and are providing hybrid-ready on-premises platforms, like SharePoint 2016, to allow them to maintain a level of on-premise computing. Which leads to the next question: what percentage of data will be stored on SSD and what will be stored on HDD? In order to maintain a seat at the storage table, the HDD manufacturers are going to have to push the cost advantages of HDD over SSD as end users look to maximize performance at the lowest price point possible. Look to see HDD manufacturers battle for market share in their sweet spot of bulk storage in data centers and tiered storage systems.
In August we talked about Seagate aggressively going after the market and leading with cost. They recognize the changing market conditions and have been leading the charge with very aggressive pricing, which has been eating into HGST sales with rumors that it has earned them 50% of Microsoft cloud business. As a result, they are now actively shipping into Amazon Web Services.
Seagate has a lot on its plate, with less than stellar results for its last quarter as they have been navigating quality issues, changing within management, experiencing layoffs, a high cost basis and a market quickly transitioning to SSD. With one industry insider also indicating that enterprise SSD channel sales have outpaced HDD enterprise sales. Without an answer for SSD, Seagate feels the weight of the market transition to SSD particularly hard. Possibly their involvement in the Bain Capital bid for Toshiba’s NAND assets may be a reversal of fortune, but we won’t know until we learn more about the structure of the transaction. In addition, there are reports that they are struggling with notebook sales, which only further contributes to the pressure that they feel.
With channel business somewhere from flat to down in the Cloud and in enterprise space, Q3 was very competitive. Evidence of this includes reports that Seagate moved 300,000 units of sub 8GB cap enterprise drives into a northern California white box server builder, at a discount to market pricing.
WD/HGST enjoys a competitive advantage, with lower manufacturing price points and better quality drives, however, they have been distracted with their being embroiled in their dispute with Toshiba. This has allowed Seagate to make inroads on market share. Not only is Seagate making a hard push into data centers but Toshiba, surprisingly, is also making a claim by leapfrogging 10TB and 12TB capacities with their sampling of a 14TB 9 platter helium SMR drive in December. It will be interesting to see how well they execute based on their past record. The manufacturers recognize the direction of enterprise-rotating media and are positioning for a fight for market share.
The long and short of it is that we see buyers benefiting as the manufacturers look to establish themselves in this shifting storage market space by leveraging their cost advantage over SSD. There will be a focus on large cap drives and niche markets, such as surveillance and legacy, where cost is the primary driver of demand.
SSD headlines continue to be dominated by the ongoing battle between WD and Toshiba over control of their joint venture. With Toshiba announcing that it is selling its memory business to the Bain consortium, what was looking to be a knock down drag out fight between two tech giants now feels like a backyard brawl with a number of competing interests. If the Bain bid goes through, it will be interesting to see how this consortium of organizations of differing needs work together.
Of all the consortium members, Apple’s involvement was the most interesting and offers a view into the future of NAND supply. With reports that both Toshiba and Micron 3D NAND yields are falling short of expectations, Apple does not want to have to rely on Samsung, its primary competitor in the handset market, for critical NAND supply for its phones. The fact that Apple is involved seems to inject an uneasiness about the future of NAND supply.
WD is clear on the strategic importance of being vertically integrated and is going to do everything in its power to maintain a grip on its investment. Expect WD to counter with injunctions that will drag this fight well into 2018, if not beyond. Good for Micron, Samsung and Intel who will continue to enjoy the high margins of a tight market and get to focus on future technology developments. Not so good for WD and Toshiba who will lose valuable time battling over FAB rights in a competitive market space.
How important is owning FAB capacity? Storage News recently published a list of 170 SSD manufacturers. Drawing a simplistic comparison between SSD manufacturing and the manufacturing of memory modules, in that both involve assembling components into a completed assembly, we can draw conclusions on the importance of owning FAB capacity. Let’s take some of the lessons learned from the memory module market. Those that control the manufacturing of the discrete components enjoy the advantage of secure supply, lower pricing and access to developing technology. Similar to the consolidation we witnessed in the storage business, we have seen the same within the dram and module business. With venerated names such as Infineon, Fujitsu, Renesys, Toshiba and Elpida disappearing and leaving Micron, Samsung and SK Hynix as the primary players with FAB capacity within the memory module space. This has left a whole host of third-party players having to rely on their competitors for NAND supply to build SSDs. Clearly the stakes are high for WD.
The long and short of it is that a protracted legal battle will extend the NAND (SSD) shortage into 2018, by delaying Toshiba and WDC’s building of FAB 6 and sidelining SKY Hynix expansion, as it holds capital to jump into the Bain bid.
How does a NAND shortage intersect with a soft quarter? When markets go short the market takes on a level of exuberance, with distributors and speculators doubling up on inventory and end users double-booking to ensure adequate supply. No one knows exactly when the shortage will end and demand and supply never operate in a linear fashion. However, at any sign of a slowing market, many of those that have been hoarding stock move it into the market to capitalize on their investment before the bottom falls out.
When talking to storage buyers at OEMs and storage appliance builders, they indicate that the manufacturers tell them that the market is tight, yet they say that allocation is easier to come by. One buyer stated: “Intel paints a really tight Q4 for S3520 SATA drives, but they keep coming up with upside over the last couple of weeks. Even on the EOL’d S3510 series – all of a sudden they are offering to their channel partners for sale.” We have also heard from within the distribution channel that their clients’ NAND allocation doubled. However, they quickly sold it off and without making price concessions.
We have seen prices drop over the past month from between 15% to 20% for Intel’s recently obsoleted enterprise SSD 3710 series with the product suddenly appearing after limited availability. We also witnessed a marked increase in Samsung PM863a and SM863a enterprise SSD series availabilities, with reports that franchised distributors are scrambling to move product off of their shelves in the face of a potential drop in pricing.
We believe that this is a temporary correction that will reverse course. We are hearing that Toshiba and SK Hynix yields are falling short of forecast roughly by 30% and that prices on EMMC and EMCP NAND (used in mobile applications) are increasing. This was anticipated with the launch of Apple and Samsung’s flagship phones, as well as with several Chinese manufacturers launching their phones as well. As product is diverted to the handset market, away from SSD, excess capacity will burn off. This will lead to increasing lead times.
DramExhcange forecast supply to outrun demand by 5% in 2018, closing the demand supply gap, however this is predicated on 3D NAND yields improving. Which is still unclear with reports that several manufacturers are falling short of forecasts. However, supply will close the gap with demand in the Q1/Q2 time frame as mobile demand subsides after the initial rush. However it is unclear if demand and supply will remain in balance as we move into the latter part of the year and demand accelerates.
Notebook/Desktop Storage Sales
Despite worldwide shipments growing in August when compared to July, year-over-year sales continue to trend downward. Historically Q3/Q4 are the busier quarters of the year, tied to back-to-school and game console sales. However, there is little in terms of product innovations to get one hopeful that desktop sales will give a significant boost to the storage market come the fourth quarter. One possible bright spot is gaming, with Intel introducing its Coffee Lake 8th gen processor, geared toward gaming, but we don’t expect this to significantly impact sector sales as a whole. Indicative of the overall mood within the PC space is Wistron; a mainstay in N/B manufacturing who are cutting back their commitment to the N/B space to look to more profitable areas of business. Another rumor revealing market weakness is Seagate is struggling in the N/B sector, with their having lost $2 per drive on 500GB adb 1TB capacities and they may face further price erosion in Q4.
Despite a declining market, the sector continues to be a mainstay in manufacturers’ portfolios, with Toshiba introducing a 1TB single PMR drive for notebooks and game consoles. This is a different approach from WD and Seagate’s SMR approach, which has led to some issues with OEMs and may provide an advantage in this hotly contested market.
Ultimately, markets are driven by demand, and the Cloud is shaping what the future of storage will look like. However, this is a large market featuring customers with a variety of storage needs. As described above, on-premise computing will still have a place in the market. And within on-premise, you will further have a variety of needs. As user weigh their differing needs of cost, performance and storage durability; there are always going to be supply imbalances in one form or another due to the high number of variability across the supply chain.
Lastly, can I ask, just how ambiguous is the language in the contract between Toshiba and Western Digital to allow this to drag on for this long?