Supply channel planning has been as challenging as it ever has been as not only do manufacturers and data centers have to manage the cost of HDD against performance, but it must be done against a moving backdrop of rapid technological change and a component shortage. Hyperscaler builds came on quickly driving a very steep demand curve that caught many by surprise, resulting in extended lead times for high cap drives. Weak smartphone sales are causing manufacturers to divert discrete NAND to SSD builds, in turn lowering SSD prices. Consequently, this necessitates planners to reassess the back end of 2018. And by now you have received your notification from WDC citing increasing component costs that HDD prices will be increasing July 1st. We have a lot of moving pieces in what is already a dynamic market.
HDD
As indicated in the introduction, hyperscaler demand and weak smartphone sales set the tone for the first half of the 2018 storage market. Nearline HDD sales, driven by hyperscaler demand, have been a bonanza for the hard drive manufacturers, allowing them to off offset weak personal computer hard drive sales. Meanwhile NAND, as a result of weak smartphone sales, is being diverted for SSD assembly driving down SSD pricing.
Despite conversations last month with planners within WDC around our hearing of component shortages, they were not willing to call them shortages, instead referring to component mix issues around customers’ bills of materials. Hyperscaler demand is often tied to a specific bill of materials requiring specific part numbers SKUs, adding complexity to an already complex forecasting model.
But now it is official, WDC is citing component shortages and has announced a 5% price increase across all product lines starting June 1st. Shortages with DRAM and MLCC capacitors have been well documented, and with the steep demand increase in capex expenditures by the hyperscalers, it is easy to reconcile the price increase for nearline drives even if it is tough to do for client drives. We have seen weakness on the client side with prices eroding and increased volumes of excess being offered, particularly a large lot of 2.5” 1TB. In the enterprise market, most action has been around high cap nearline drives. There are rumors that Seagate stuffed white box builders with mid-cap enterprise drives to meet quarter-end numbers. It will be interesting to see how WDC’s across-the-board price increase plays out over the upcoming months.
Regarding nearline demand, there is concern that build cycles may be winding down. Even if demand subsides, we expect nearline to carry strength into Q3 as manufacturers fill their backlog and current demand. It is difficult to ascertain future demand with several variables impacting the demand-supply equation. When a handful of customers can raise markets they can also quickly lower them. I have a sneaky suspicion that Hyperscalars may be influencing market dynamics to lower their buy prices. If you recall in September/October timeframe of 2017 Seagate and HGST were competing for market share, and prices were trending downward. Suddenly Hyperscalars were gobbling up the product. How can forecasts be so far off?
We expect to see continued downward pressure on client drives because of weakness in the personal computing market, and the cannibalization of HDD NB drives by SSD. At Horizon, we have seen a shift in our notebook customer base towards SSD indicting that cost performance ratio has become attractive enough to move customers to SSD over HDD. With more consumers storing content in the cloud and leveraging SaaS applications there is less a need for local storage capacity than there is speed in accessing data to run content-rich applications.
SSD
2018 is looking to be a pivotal year for SSD with our already having seen manufactures divert NAND from slowing handset sales to the SSD market, resulting in lower SSD contract prices for the first half of 2018. Maturing 3D NAND yield and the pending ramp to 96 layer NAND technology promise to continue to drive down the cost per bit. Artificial intelligence (AI) and machine learning (ML) applications are delivering richer and more relevant interactions driving demand for SSD based storage and are expected to propel demand for Purley PCIe based systems.
Manufacturers are leveraging their additional NAND supply to drive SSD adoption within the notebook space which had been hampered by tight supply, limiting the adoption rate to 45% in 2017. With additional NAND from lagging smartphone sales and increased yields from 3D NAND, we have had reports of second-quarter contract prices drops between 5 and 10% for client SSD contracts. We expect to see further declines as manufacturers look to capture market share in anticipation of additional supply. Most likely, Q2 results will reflect an uptake of SSD past 50% for notebooks. Not surprisingly we have been hearing some manufacturers are offering backend monies to distribution to help move overstocked client SSD.
The client market’s transition to SSD is a foregone conclusion, and it looks like we will see a measurable uptake of SSD within the enterprise market in 2018. AI and ML are increasingly becoming part of our daily lives, and much of the computing will take place at the edge with much of that data being unstructured, which PCIe is much better suited to handle. PCIe allows tens of thousands of parallel queues rather than having to utilize single serial command queues of SATA and SAS interfaces limited to 250 commands, resulting in a measurable drop in latency. NVMe over fabric (NVMe-oF) utilizes NVM Express 1.3, which addresses the limitations of PCIe as PCIe SSD only work in the physical server they are in. NVMe-oF pushes NVMe and PCIe further into the mainstream by supporting NVMe over fabric and allowing for the use of PCIe SSD in a virtualized environment.
As we look to see what is in store for the second half of the year for SSD, there are mixed opinions. I read a report within Digitimes that smartphone sales will improve the second half of the year, with a couple of analysts believing this might push NAND into an undersupply situation on the back end of the year. I will not rule that out but it would require a significant reversal of smartphone sale trends, which have seen significant declines. In the enterprise market, we will see continued penetration of all-flash arrays to address the changing storage topography of the web.
FINAL NOTES
I must say I have questions around the WDC price increase. Market dynamics do not justify an across-the-board hike. The timing seems odd, especially with a reasonable window to buy at current prices before the increase.
There has always been a love-hate relationship between buyer and seller in the technology market, at times seeming like the ultimate participants in game theory, with both sides willing to seize an opportunity when they can.
