Jahanara Nissar

The surge in investor interest this year in the HDD sector has been huge. Its lateral implications to semis and the AI space in general have been sizable enough for us to step in, suggest a few ideas and address a few myths.

Lead time for the high-capacity HDDs, especially for SKUs preferred by hyperscale CSPs are north of one year. Across the industry, there is no expectation of lead times coming down anytime soon. Large-capacity HDDs are on ~100% allocation to hyperscale CSPs. New drives are simply not available to enterprise customers. Enterprise customers have no choice but to go for drives in the secondary market. There is pent-up demand at enterprise customers for new HDDs

We expect the HDD names to continue to perform well, despite the more than doubling in STX and WDC since earlier this year. The recent sideways movement due to understandable fears of capacity increase, offers an opportunity to participate in further upside. There is a certain amount of demand ‘leakage’ to the SSD side of the storage business, in our view. Deliberate increase in HDD manufacturing capacity likely captures revenue currently lost to SSDs instead of putting downward pressure on HDD pricing power. We think on a relative basis, the HDD names may offer superior risk/reward to SSD names. We would be long into prints of STX and WDC.

Laterals: 1) MRVL should see positive demand pull as and when HDD vendors increase manufacturing capacity, raise their demand for drive heads. 2) MSFT and AMZN/AWS could be witnessing a powerful storage trend as they monetize the ongoing HDD upgrade cycle by selling high-value storage services to enterprise customers.

–xx–

Myth: The surge in storage demand, HDD and SSD, is due to insatiable demand from AI training and inference workloads from the present generation and future generations of AI data centers.

Reality: We think the current HDD cycle is driven by an upgrade cycle in the existing installed base of traditional storage at major hyperscale CPS. As such the demand could be much larger than if it were merely limited to the AI space.

SSD: While SSDs are linked to AI-related storage, SSDs too could be seeing demand pull from the traditional storage market. Due to extreme shortage of high capacity HDDs some CSPs may be willing to pay up for SSDs to satisfy storage needs. There may well be a level of ‘leakage’ of demand from HDD sector into SSDs despite the higher pricing per bit of the latter.

HDD – A real product cycle: The demand pull from an upgrade cycle of the massive installed capacity in the traditional storage space is at the core of this HDD cycle. Tier 1 storage providers are upgrading their installed base of storage infrastructure from tape drives and SMR-based HDD to large capacity drives based on the latest HAMR technology, which came out of extensive field testing just last year. There is an honest-to-goodness product cycle going on, which does not come too often in the HDD space.

32TB is the sweet spot: Major storage providers such as MSFT and AMZN/AWS we believe are transitioning their installed base of HDDs (28TB and older) to 32TB models, the first major SKU based on HAMR drive technology. HAMR drives went through extensive qualification at major CSPs last year. 2025 is the first year of mass adoption of HAMR-based drives. While the HDD companies offer HAMR drives at other densities too, the sweet spot appears to be 32TB. And for the 32TB drives the lead time is over a year. According to our checks the limited supply is allocated exclusively to the Tier 1 hyperscale CSPs.

Pent-up demand: Enterprises do not have allocation from the HDD suppliers. Enterprises too prefer the 32TB SKU. And they are willing to procure pre-owned 32TB HAMR drives in the secondary market. The pent-up demand at enterprises we expect will maintain elevated lead times for an extended period even after the demand pull from Tier 1 CSP slacken.

Manufacturing capacity increase – should investors be concerned? Under usual circumstances, investors should be indeed be concerned about capacity increase, but perhaps not in this case, in our view for three reasons:

• Nascent technology: The HAMR technology is still nascent enough and not fully understood. HDD suppliers are likely wary of expanding capacity too quickly less they be stuck with unoptimized yields. In the HDD industry, technology transitions are few and far between. Unlike the silicon industry, the drive process technology takes years to fully mature. HDD vendors are typically loathe to jump in and expand capacity before they fully optimize yields. And that could take years.
• Loss of business to SSDs: The HDD industry may be witnessing unmet demand for large capacity drives (>30TB) overflowing into large capacity SSD flash drives, thus leaving money on the table. Expanding manufacturing capacity, within limits, captures lost revenue without necessarily impacting pricing power
• Slow test time: Due to the slow speed of the read channel, the testing/partitioning phase can take up to one day for each drive. As a comparison we note that test time for a large capacity SSD is negligible due to the order of magnitude higher data bandwidth of the flash controller vs. the HDD read channel. The bandwidth of an SSD flash controller is typically ~13Gb/sec vs. a HDDs read channel ~0.5Gb/sec. Net/net: Even with expanded manufacturing capacity, the testing phase of HDD is agonizing slow compared to SSD testing, thus keeping HDDs volumes from surging in a short span of time.

HDD margins may have upside due to special circumstances: As nearly 100% of the supply of large capacity HAMR drives are on allocation of a handful of hyperscale CSPs, the HDD vendors do not need to use the typical distribution channels of OEMs and VARs. As such, the HDDs being supplied to the hyperscale CSPs may not be completely commoditized. We think to a certain extent the HDD suppliers could be making slight modifications to suit the specific needs of individual hyperscale customer, thereby extracting better value vs commoditized HDDs.

Key ideas: 1) Lead time for high capacity HAMR drives extend out to a year, 2) new supply on allocation to hyperscale CSPs, 3) pent-up demand at enterprises, 4) slow but deliberate increase in HDD supply captures revenue that is currently lost to SSD industry; downward pressure on pricing may be negligible, 5) capacity addition at HDD suppliers, as and when it occur, is likely to be slow compared to the pace typical of the semis industry.

Net/net: With lead times of key SKUs stretching out to a year and with no sign of lead times coming down, earnings multiples are likely to head upward, driving STX/WDC stocks higher. We will roll out our model and price targets after the earnings calls. We would be long into print.