One of the most significant recent developments I’ve observed regarding WDC is the long-term HDD demand signal conveyed by company management.
Speaking at the Citi Global TMT Conference on September 9, 2026, Western Digital CFO Kris Sennesael stated that customers are seeking long-term supply agreements (LTAs) extending through 2030 and 2031.
There is a crucial detail here:
WDC has not yet signed contracts for 2030–2031.
However, customers already want to secure HDD supplies for those years.
To understand why this matters, I believe we first need to look at the kind of company WDC has evolved into today.
WDC is no longer a "PC/HDD cycle" company in the traditional sense.
The company generates approximately 90% of its revenue from cloud customers. Management anticipates a Compound Annual Growth Rate (CAGR) of over 25% in exabyte demand over the next five years.
More importantly, the factor limiting WDC’s revenue today is supply, not demand.
In other words, the question is no longer, "Will there be sufficient HDD demand in the AI and cloud sectors?"
The real question has become, "How much of this demand can WDC meet?"
There is an important distinction to be made here to understand the HDD aspect of the AI narrative.
When a hyperscaler installs new GPU capacity, that compute capacity can be utilized repeatedly using the same resources.
However, new data is generated every time a training session, inference task, or other AI workload runs.
This data continues to accumulate in the form of model outputs, user data, logs, media, datasets, and other content.
Even if compute capacity remains static at a certain point, previously generated data does not simply disappear. On the contrary, the volume of data is growing over time.
In other words, the growth of compute in AI doesn't just mean more processing power.
It also creates a need for storage that expands over the years.
According to WDC, approximately 80% of the data generated today is stored on HDDs.
That is why tracking AI investments solely by asking "how many GPUs are being installed?" is not enough.
One must also ask the following question:
Where will the massive amounts of data generated by these GPUs be stored?
Another key point made by the WDC CFO is that hyperscalers plan their data center investments more than five years in advance.
Land, permits, and capacity for XPUs, memory, and storage must all be secured years ahead of time.
WDC operates on a build-to-order HDD production model, and the production lead time for a single HDD is approximately 52 weeks.
Therefore, discussing storage capacity for 2030–2031 today is actually more significant than one might think.
Moreover, the majority of 2027 is already covered by Long-Term Agreements (LTAs).
Parts of 2028 and a smaller portion of 2029 are also covered by such long-term agreements.
And on top of that, customers are seeking further assurances for 2030–2031.
Alongside the demand side, the technology roadmap is also critically important. WDC:
• Completed the transition to 32TB ePMR
• Is ramping up production of 40TB ePMR products
• Aims to ship first-generation 44TB HAMR drives in the first half of 2027
• Plans to move to 50TB+ products thereafter
• Has a long-term roadmap for 70TB and 100TB+ HDDs
Regarding HAMR, qualification processes are currently underway with four major hyperscalers, and the company reports positive customer feedback.
WDC’s capacity expansion is not driven solely by building more factories.
It aims to extract more exabytes from existing infrastructure through higher areal density, increased platter counts, advancements in head/media technology, and factory automation.
The financial transformation is also noteworthy.
The price per terabyte rose by 18–19% year-over-year in the last quarter.
Gross margins climbed to the mid-50% range, while operating margins reached the mid-40% range.
Incremental margins exceed 70%.
The company anticipates that the cost per terabyte could decline by approximately 10% annually in the long term.
And the long-term expectation for CapEx as a percentage of revenue is in the 4–6% range.
In my view, interpreting the WDC story merely as "HDD prices are rising" is incomplete.
The real story is this:
AI compute → more data → increased storage demand → hyperscalers planning capacity years in advance → WDC gaining long-term supply visibility.
Furthermore, as higher-capacity HDDs are introduced, WDC is working to generate more exabytes from the same physical infrastructure. Therefore, for me, WDC receiving Long-Term Agreement (LTA) requests from customers extending to 2030–2031 is not merely news of a "long-term contract."
It is a significant signal indicating that demand for AI infrastructure—specifically on the storage side—extends years into the future.
The infrastructure that stores the data generated by AI is also a crucial part of this story.
Not investment advice.
