A widely accepted position in the HDD industry is that demand for capacity enterprise (nearline) HDDs is driven by large hyperscale companies, typified by cloud service providers (CSPs) and “as-a-service”, or XaaS datacenter companies. What is difficult to read is the exact buying cycle that these companies follow, so demand spikes are typically followed by a period of weak purchases. Based on historical tracking of nearline HDDs and exabyte shipments, the clearest pattern that emerges is that unit growth usually lasts from two to four quarters, followed by a two to four quarter lull. Beyond these fairly wide windows of time, only backward-looking analysis can explain certain events. One example is the four quarter period of weak nearline HDD demand from CQ3 ’13 through CQ2 ’14 – a period where quarterly unit shipments hovered around the 8 million mark. In hindsight, this period was one in which many companies employed erasure coding to reduce the number of replicas required ensure against data loss due to a failed drive. Erasure coding improved storage efficiency to such an extent that companies spent several quarters utilizing freed-up capacity before significant purchases of HDDs were required.
Interestingly, throughout these lull periods, the drives that were purchased continued to mix up in capacity, so even on lower unit shipments, exabytes remained fairly constant, until the next buying cycle accelerated both units and capacity shipped. In CQ3 ’14 and CQ4 ’14, nearline HDD units spiked to the highest ever unit shipments and even though units declined in CQ1 ’15, the volume was still the second-highest ever. By CQ2 ’15, nearline HDDs dipped significantly, dropping nearly 9% sequentially – demand from hyperscale customers was off sharply and despite a continued increase in average capacity – one that is approaching 3.4 TB – the total capacity shipped dipped in excess of 5% sequentially.
CQ3 ’15 is not looking like a strong recovery quarter for nearline HDDs with hyperscale demand remaining soft for many companies in the space. In addition to utilizing the record number of HDDs shipped over the previous quarters, some public CSPs have signaled CAPEX reductions over the near term. Will this mean that demand for nearline HDDs will not see the expected pop in CQ4 ’15? Perhaps not, but mapping large hyperscale company CAPEX to HDD shipments on a quarterly basis has only occasionally shown the two curves to move together. Another somewhat obvious point to consider is that the timing of datacenter scale outs between different companies does not occur on the same cycles. Given the volume of HDDs some of these companies purchase per quarter, a large increase in demand by just one or two hyperscale customers can offset the weakness of others.
With all of the uncertainties in play, TRENDFOCUS remains cautiously optimistic for a decent quarter for nearline HDDs in CQ4 ’15. Maybe not a record number of units shipped, but certainly strong enough to help boost an HDD industry looking for any positive business drivers.