Horizon January 2018 Storage Market Brief
The storage market closed with signs of a slowing market, but opened the New Year with a bang. The market was caught unaware as the supply-demand equation was flipped on its head. As WD and Seagate scuffled for market share, the market quickly shifted gears with server manufacturers reporting lead times extending to 8 weeks on large-cap drives.
Seagate’s aggressive pricing followed by HGST’s reshuffling of its prices would lead one to believe that the manufacturers were adjusting pricing because supply was outpacing demand. It was thus unexpected to see this sudden swing in demand. However strong worldwide cloud demand and China’s growing infrastructure demand on the back of an improving global economy have kicked the market into gear. Add in a reported supply side ceramic capacitor issue compounding an already tight HDD market.
All eyes are on the HyperScale market and it is progressively becoming a stronger influencer on the storage market. Increasingly they dictate market dynamics with manufacturers, as they direct capacity to the HyperScalars. Towards the end of Q4, the HyperScalars simultaneously placed purchase orders. Combined with the growing Chinese data center infrastructure appetite for storage, to meet growing end-user demand, the HDD market quickly reversed course with reports that 4TB-through 12TB near line capacities are in a shortage, with 10TB sizes being hit particularly hard. We have seen an increase in open demand for large-cap drives, with buyers looking to pay sub $30 per TB (prices are dependent on quantity) while sellers are looking $35 plus and lead times stretch.
Last month we talked about the HDD manufacturers reducing support of 2.5” SAS offerings due to SSD offering a lower total cost option to 2.5” SAS HDD. We are seeing evidence of this as OEM’s several builders are scrambling for product as HDD manufacturers are increasingly focused on the large-cap near line. With many part number iterations and feature sets, product is proving to be increasingly scarce, driving a significant delta between contract pricing and open market prices.
The wildcard is the industry-wide passive shortage, with reports that the capacitors are short and are impacting HDD supply. To date, we have only heard of the shortage impacting WDC/HGST. We find it puzzling to not hear that Seagate and Toshiba would be feeling the pinch of the shortage if WDC/HGST is feeling the impact. Franchise distributors are saying that HGST product is being diverted to the large CSPs and that channel quotes are being handled on a case-by-case basis. We are not clear on whether this is specific to WDC/HGST or is a growing problem for storage as a whole.
Seagate reported strong numbers, reflecting favorably on their revamped market approach, reporting growth in both year-over-year revenue and profitability. Seagate’s enterprise shipments results also came in at 31% over last year’s strong-demand year. As detailed in last month’s update, WDC/HGST was sidetracked due to managing their merger. Let’s not forget they are still the largest HDD manufacturer by volume, and they will get back on track soon. However, it still unclear as to the degree and the impact that the capacitor shortage will play on them. In terms of Toshiba, it is an ideal time to introduce their 12 and 14TB caps to gain a foothold in the data center space. Despite their confidence for a 14 TB ramp sometime in 1H of this year, a year-end ramp seems more likely.
Enterprise SSD volumes are caught in the Hyperscalar demand updraft with flash array and software-defined platforms continuing to make significant inroads which are supporting SSD pricing. With a market bifurcated along MLC and TLC based SSD, demand volumes are being absorbed by improving TLC NAND yields. We expect prices to hold stable for TLC-based SSD for the near term. The Hyperscalar buying activity is not having a significant impact on MLC based SSD, but pricing is holding steady if not trending upward for MLC based SSD. We expect a gradual trend upward as manufacturers continue to shift Fab capacity to 3D based NAND.
I want to take a moment to address a comment on last month’s storage update questioning the possibility of connection speeds between data centers as a factor limiting the NVMe adoption. Typically, to realize the full benefit of the introduction of new technology, the companion components or network need to be optimized. It is pretty clear that with the growth of AI and IoT within our daily lives, this is what NVMe is ideally suited to, managing data in environments that require low latency and speed. Increasingly AI and IoT data will be managed in an edge environment where data will be processed locally before being pushed to a central location. However, that does not mean that there aren’t some hurdles to be crossed before NVMe becomes mainstream. Currently, the cost of NVMe is too high for wide-scale deployment, but that will soon correct itself. Deployment of NVMe will requires new rack and or enclosure as legacy cards and wiring will not be adequate. The next issue is around deployment. Will it be NVMeoF or switched PCIe? There still needs to be standardization around fabrics (transfer of commands between a host and target) for NVMe to be the base storage at the enterprise level. Raid controllers will need to be updated, as they typically are not configured to handle NVMe and can be a liability if a drive fails, bringing down the whole system. Approaching with switched PCIe introduces challenges with PCIe switches and storage protocols.
At the end of the day, the performance benefits of NVMe are too compelling and the growing needs of AI and IoT will require new NVMe-based architectures as tiered storage based on SAS which will prove insufficient to meet the needs of AI and IoT.
While the enterprise market tips into shortage conditions, we are not seeing the same for the Notebook and desktop storage market. Although sales continue to decline for the notebook desktop market, Q4 can be viewed as a success due to a relatively stable deceleration in sales relative to 2016 Q4’s steep decline.
Both Seagate and WDC are working to adjust costs to reflect the reality of lower volumes and encroaching SSD sales. As TLC based 3D NAND yields improve, SSD pricing will drop, marginalizing the cost advantage of HDD as evidenced by Seagate divesting from 500GB client drives. SSD consumed 50% of the client market in 2017, and that is expected to eclipse 60% in 2018. We expect to see the SSD cannibalization of HDD sales to accelerate as SSD yield improvements increase.
As the desktop market shrinks, the manufacturers are looking for new growth areas for rotating media. The surveillance market, where cost is a key variable, is tailor-made for the hard drive manufacturers. It is a growing market that is expected to exceed 10 billion in 2018. Howevbver, Iit is taking on increasing intelligence including AI and real-time analytics combined with the demand for heightened image resolutions which will only further drive the need for storage. Regulations are driving a growing need to hold data longer, which will fuel the growth of external storage such as NAS, DAS, and SAN.
In trying to understand the sudden market turnaround for hard drives a theory that I heard postulated was that Hyperaclars had been holding builds with the intent to drive component prices downward. I don’t believe that this is the sole cause, but it could very well have been a contributor. As the market becomes more and more consolidated around a handful of large data centers, it is plausible to imagine such a scenario.
Seagate investing in Ripple cryptocurrency? Ok, then let’s talk about that. What do cryptocurrency and blockchain technology mean for the storage market, where data is broken up, encrypted, and sent to hard drives all over the world rather than a centralized location such as a data center?
I would love to hear your thoughts.
Please comment on LinkedIn page or email me directly at Stephen.firstname.lastname@example.org