Enterprise HDD
We are just now at the beginning of Big Data, IoT, and AI. Steve Luczo, the former CEO of Seagate, states in his joint report with IDC that 163 zettabytes (16 trillion gigabytes) of storage will be created and replicated by 2025. That would require 16 billion 12 terabyte hard drives. If WD can execute on its MAMR technology, the public cloud storage market will be viewed through the MAMR lens.
WD’s MAMR has the potential to be a big win as it targets the Hyper-scale market and its continual growth with an elegant and cost-effective solution for expanding areal capacity. By utilizing existing PMR technology WD limits complexity and holds down costs. Pundits report that Seagate’s approach, utilizing HAMR (heat-assisted magnetic recording) requires a new bill of materials, an updated supply chain, and frequent FW changes to manage wear concerns resulting from HAMR’s approach of heating the media for write cycles. Increasing costs and complexity.
In what appears to be a winner-takes-all scenario, Seagate responded to critics that touted the advantages of MAMR over HAMR as detailed in the attached link; https://www.storagenewsletter.com/2017/10/25/seagate-hamr-next-leap-forward-now/. Seagate is targeting pilot runs in 2018 and is shipping 20GB in volume in 2019. At first blush, MAMR appears to be the cleaner and more cost-effective solution. Execution, has been Seagate’s Achilles heel with its well-publicized struggled executing around its 10TB drive. HAMR has been in the pipeline well over two years which may speak to suspected complexity issues around HAMR. The two approaches may be setting up cold storage to the possibility of sole source scenarios.
With so many exciting developments around the cloud and SSD, it is easy to forget that enterprise computing supports a variety of computing needs. Existing infrastructure, workloads and growing businesses such as edge computing, surveillance and network backup live outside out of the big cloud solution providers and require low cost storage. HDDs offer the lowest cost option in high capacity sequential read applications where cost-per-bit is of the utmost importance and will continue to play a role in such environments. Exabyte shipments will continue to grow out the CSP environment.
The decision of private-versus-public cloud is driven by TCO (total cost of ownership) which can be reduced down to the two primary drivers of cost per terabyte and use of data. If data sits idle with sparse AI and cost per terabyte is your primary goal, owning your own data center may be your best option. As evidenced by Dropbox growing to the point that it was more cost-effective to move its data in-house. If you are a dynamic user turning your data multiple times a day and finding yourself spending time remapping bad sectors to optimize performance, the public cloud is probably your best solution. If you have a limited capital budget or are looking to experiment with several types of solutions, a pay-as-you-go approach makes the best sense.
Seagate aggressively pursued Hyperscale business leading with price, thus resulting in their success getting into 3 of the bigger public cloud CSPs. The tradeoff of price-for-volume has benefited nearline results and is reflected in bottom-line results with Seagate reporting strong Q3 results. Pricing pressure from both Seagate and Toshiba has affected all enterprise caps and forced WD to lower pricing. However, let us not forget that WD owns the dominant share of the helium market with 20 million drives shipped. Seagate is targeting 1 million 10TB for the quarter. Seagate indicated that after a soft spot, the enterprise buying patterns are normalizing and we can expect to see a 3 to 5% growth in Q4. The question is if the quarter end push cannibalized Q4 sales. Although HDD revenues have increased, margins were down supporting observations that prices trended downward. WD is not as bullish going into Q4 as Seagate is and is forecasting a slight decrease in HDD revenue for the next quarter. Increasing SSD 3D yields will continue to eat into high-performance HDD, with SSD setting its sights on 10K HDD after already having surpassed 15k drives.
SSD
There continues to be a difference of opinion of when the SSD shortage will come to a close, with some indicating Q4 and others believeing that it will carry on into 2018. Samsung in their earnings call reported that they expect favorable conditions for 2018. However, we are not seeing things that way. We expect to see SSD supply catching up and possibly narrowly exceed demand in the first half of 2018. We expect 3D-based SSD pricing to continue to trend lower as yields improve and 2D-based SSD to trend upward as excess stock burns off. We have witnessed spot shortages however in talking to most end-users we are not seeing a constricted market. Many report that SSD continues to get easier to procure and prices continue to trend lower. Micron is reported to have been using pricing to gain access to gain market share in Hyperscale accounts. Silicon motion reported in its Q3 results that they saw an increase in NAND deliveries, supporting a 50% revenue increase in its SSD solutions sales. Both events indicate that product is becoming more available.
With so many variables potentially affecting the supply-demand equation (including a slowing market, inconsistent TLC yields, demand from competing entities, and end-users double booking) it is difficult to get a clear view of who is feeling the shortage and who is not. Particularly disruptive to the supply-demand equation is the practice of double booking which makes it very difficult to gauge what is out there for stock and what one should be paying. Having weathered several past DRAM shortages, this shortage was atypical with remarkably few OEMs buying from the open market which would support the view that there was a quantifiable level of double booking. Regardless, with so many part number SKUs, interfaces and performance criteria pricing never moves in unison and there is always bound to be product mix issues resulting in a problem part. Unfortunately, the manufacturers cannot treat all customers the same, with manufacturers having limited allocation and the need to endorse customers that they see as strategic to their business. A good example is the rumors that Pure Storage (all flash array mfrg) benefited from price breaks that others did not see. The wildcard for Q4 is handset sales which can quickly gobble up NAND volumes, upsetting SSD supply.
Does WD’s MAMR announcement reset the HDD/SSD equation? As detailed earlier, MAMR’s advantage is cost, but it is nonetheless limited by its 6GB transfer speed. Economically, QLC, the next generation of SSD, is logical due to the 33% capacity improvement per cell, in turn lowering flash costs. However, there are significant technical challenges to be met. To increase density, manufacturers increase layers of silicon (32 layers for 2D, 64/72 for 3D and 96/128 for QLC) which in itself is a more complex and costly manufacturing process. Second, when writing data to a cell it induces cell degradation resulting in increased cell leakage. This results in the need for increased management at the controller level to offset bad cells. The more writes performed the further damage to the cell, moving the SSD closer to failure. This is why manufacturers over provision flash where they spread write cycles across many cells requiring increased memory size to ensure a reasonable life span. Will MAMR slow in investment in QLC based NAND?
Notebook/Desktop Storage Sales
PC shipments declined in the low single digits Y/Y According to both Gartner and IDC which is better than expected. With our being a month into Q4, we see no events that will significantly alter the seasonality downward trends of client builds, and market feedback also indicates a slowdown. Both Seagate and Toshiba also were aggressive with pricing in Q3 and we expect that to carry on into Q4 as improved 3 SSD yields hasten the transition to SSD.
Although we are in a declining market that is quickly transitioning to SSD, both Toshiba and Seagate continue to show commitment to the client space by releasing new products such as Toshiba’s MQ04 singles disk 1TB model and Seagate’s 3.5” 2TB platter drives supporting 2 through 8TB capacity points.
In terms of SSD we are seeing increased availability and prices that are trending lower. With the combination of improved 3D NAND yields and seasonality, we expect to see prices trend downward as we enter Q4 and speculators look to offload stock in anticipation of the seasonal slowdown. We expect to see a continued increased adoption of M.2 interconnect utilizing PCIe interface with an increasing number of notebook designs utilizing PCIe to leverage improved performance.
Final Thoughts
In today’s environment of heightened media competing for readership, it is easy to get caught up in the hype. Although SSD is taking computing and storage to new levels, HDD still makes up the majority of exabytes shipped with 80% of enterprise market and 65% of the PC market’s storage sales. SSD will continue to make inroads and when supply catches demand, SSD usage will accelerate. But let us not forget that HDD has a cost advantage. With MAMR projecting to support 40TB drives at a 10x cost advantage over SSD, we expect to see HDD play a role in storage for the foreseeable future.
With Toshiba just getting to Helium-based drives, is this an opportunity for a quid pro quo opportunity between WD and Toshiba via licensing agreements to help bring their nasty SSD dispute to a close?
