NAND Market Sufficiency, CPU Shortages, Nearline HDDs

Here are two Executive Briefs from Trendfocus, Inc.‘s Information Services.

Part One

NAND Market Sufficiency, CPU Shortages, and Another Tour Through Korea and Japan
 

Our last trip to Korea and Japan this year both confirmed some expectations and yielded a few new data points.
 
NAND dynamics in line with expectations: As predicted, the NAND market has recovered and bounced off the bottom in pricing. 3CQ19 output of both NAND and SSD set records, and NAND is now in a more balanced situation. Trendfocus observes that there is significant variance in opinion among the NAND makers as to how long it will remain balanced or whether a measurable shortage will occur in 2020.
 
CPU shortage more acute than previously thought
Discussions throughout the PC supply chain helped clarify information flow around Intel’s concerns satisfying demand for CPUs. Over the last several months, Trendfocus’ position was that this was a concern in the low end of the PC market. However, it is clear that even the top PC companies are now experiencing shortages that will restrict 4CQ19 production and possibly into 2020. The pressure may ease a bit into 1CQ20 as seasonal demand drops off and the Windows 7 replacement cycle is effectively completed; however, similar to earlier this year, CPU product mix may continue to challenge PC makers to optimize builds around available product. Intel’s focus on satisfying higher-end 14nm products, especially for the server market exacerbates the PC CPU shortage and the chip giant has publicly stated that it has plowed an additional $1 billion into CAPEX to boost 14nm output and to support its 10nm ramp.
 
Enterprise PCIe SSD transitions relaxed a bit …
Part of the ‘art’ of forecasting (the final touches after the quantitative processes are deployed) is factoring user transitions regarding form factors, capacities, and the like – the qualitative assumptions.

Trendfocus had assumed some large hyperscale companies would transition to higher capacity PCIe SSDs in the mid to second half of 2020. As these transitions will most likely still occur, the percentage mix to higher capacity products will not be as aggressive as originally anticipated – with certain data center deployments staying with the existing capacity products as a higher mix.

In addition, enterprise SATA SSDs should maintain significant volumes throughout 2020, with no real cliff in the immediate horizon for these SSDs. Reasons for this vary – with some being continued challenges for certain customers to transition away from SATA to PCIe, as well as pricing strategies for certain market segments maintaining with PCIe at the high end, and SATA at the low end – ensuring continued demand for these lower-priced SATA system-based solutions. In addition, with the channel/aftermarket still dominating with well over 50% of the SATA market shipments, this segment is notorious for taking a bit longer to transition to newer protocols than the other, more aggressive markets such as hyperscale, and even system OEMs.

Part Two

CPUs and Nearline HDDs
 

Client CPU shortages continue: As Intel has stated publicly in its last earnings announcement, PC demand is trending stronger than its expectations, taxing the chip giant’s ability to maintain enough supply to satisfy OEMs.

A slightly deeper dive indicates that both Intel Whiskey Lake and Comet Lake supplies are falling short, with the latter processor in severe shortage. Comet Lake is Intel’s fourth iteration of its 14nm CPU (Whiskey Lake is its third gen 14nm product), just recently shipping only in low-power configurations for notebook PCs.

Similar to the shortages earlier in the year, it appears that Comet Lake shipments are targeting higher-end, more profitable versions, which end up in higher-tier notebook PCs, meaning mobile PC price bands that are already dominated by SSDs.

Shortages targeting lower-end CPUs will tend to impact HDD configured notebook computer shipments and coincidentally or not, build plans for client 2.5″ HDDs are down sharply in 4CQ19. The final quarter of the calendar year will usually see reductions in 2.5″ HDD models going to game consoles and somewhat to PCs as a result of builds for holiday shipments in the prior quarter; however, the sharpness of 2.5″ client HDD build plan reductions in 4CQ19 seem to bolster the idea that CPU shortages are disproportionately affecting lower-end models.

Desktop CPUs, which will not see Comet Lake until 2020, continue to rely on the Whiskey Lake gen. While shortages of this older 14 nm design are also impacting supply, the magnitude of the shortage appears less severe than Comet Lake.

The ongoing commercial refresh of older Windows 7 desktop PCs has supported desktop HDD demand – a market in which the lower system ASPs tend to favor HDDs. Build plans for 3.5″ client HDDs in 4CQ19 are moving in the opposite direction of 2.5″, rising either to meet decent demand or to replenish supplies depleted as a result of earlier demand outpacing more conservative HDD builds earlier in 2019. The HDD supply chain is scratching its collective heads on the both the 2.5″ and 3.5″ component demand gyrations that have occurred in recent weeks.

Following a huge prior quarter, nearline shipments due for a (short) pause: Accelerated cloud demand for high-capacity nearline HDDs drove both unit and capacity shipments to record highs in 3CQ19 with an unprecedented volume of nearline shipping to just a handful of U.S.-based tier-1 hyperscale customers.

Between the major cloud customers, the transition to both 14TB and 16TB nearline HDDs is staggered based on a variety of factors including platform as well as established internal transition plans.

For those Trendfocus clients that subscribe to our Cloud, Hyperscale and Enterprise Storage Service, the highly volatile nearline capacity expansion cycles have been covered in excruciating detail in each quarter’s reports.

Assuming no better evidence than pure pattern recognition (by the way, more thought is given to our forecasts than this), the past decade has illustrated that exabyte growth cycles, and unit growth to a lesser extent, last on average about three consecutive quarters. The record expansion in 3CQ19 marked the third straight quarter of unit and capacity shipment increases. Combined with the staggered transitions to higher capacities from large cloud customers, the timing is ripe for a bit of a pause.

While Trendfocus forecasted this break in growth to occur in 1CQ20, word out of the Asian HDD supply chain points to two things – one, nearline shipments may decline in CQ4 after the torrid shipments in the prior quarter, and two, build plans remain well above any reasonable shipment TAM. Several supply chain indicators point to slight reductions in nearline builds for 4CQ19, potentially due to tight supply of some components, but the total build plans still outpace perceived demand. Parts suppliers remain wary of possible sharp cuts in demand sometime in 1CQ20 as HDD companies rationalize component and HDD inventories once enough product is buffered to satisfy the unpredictable cloud demand and capacity transitions.

HDD and component suppliers fundamentally support the long-term growth picture for nearline units and capacity. However, the erratic and highly unpredictable quarter-to-quarter changes have HDD supply chain personnel scrambling to explain the turbulent end demand to their executives who question why nearline expansion is not occurring with a nice even pace that long-term data growth would seem to support.

Whether or not nearline units and exabytes rise slightly, per Trendfocus’ current forecasts, or dip as a result of a pause following hyper-expansion, the fundamentals of cloud storage demand remain.

Long-term capacity shipments should still support a greater-than 30% annual CAGR, but actual annual capacity growth will fluctuate wildly, based on the timing of customer investments, as well as HDD product and server platform transitions. Those that understand the volatile dynamic of storage demand to address data growth will be better able to handle the stomach-churning short-term volatility painfully illustrated by the results of the past decade.