BONUS: Western Digital CEO Irving Tan Talks AI Storage Demand

**SPEAKER_1** (0:02)
Bloomberg Audio Studios, podcasts, radio, news.

**Ed Ludlow** (0:08)
We're getting back to the AI hard disk drive story, Irving. It's great to have you back on the show. You know, like, investors still believe in that story, AI hard disk drive.
Your stock is under some pressure. Just reflect on like the moment that you tried to outline in this earnings print.

**Irving Tan** (0:27)
Yeah, well, first of all, thank you very much for having me back. If you look at the results that we delivered in Q4, fiscal 26, they were very strong. We had over 40% revenue growth. Gross margin was above 54%.
Very strong cash flow generation as well. If anything, probably the expectations were very high in terms of the increased amount of exabytes that we could supply into the market. But if you take a step back, really the fundamentals for ongoing revenue growth, margin expansion, cash flow generation remain very strong. On the demand side, demand from hyperscalers, neo-clouds, physical AI remains very robust, and we continue to see over 25% CAGR demand from a demand perspective over the next five years. Similarly, on the supply side, we have a very robust roadmap that we have put in place that will be able to meet this demand. And so it's pretty much a function of just the quarterly transitions that we're going through.

**Ed Ludlow** (1:27)
Irving, what you're touching on there is storage is distinct in many ways from what's happening in memory. In some ways, highly analogous, right? There is a pricing story there. There is a demand outpacing supply story. And your industry facing the balancing act of expanding output, manufacturing capacity, on a permanent basis or not, how are you managing all of that?

**Irving Tan** (1:54)
Yeah. So first of all, thank you for highlighting that. There's a clear distinction between memory, which is like NAND and DRAM versus storage, right? Because memory and like compute is being recycled all the time for different workloads. But the data that's being generated continues to be stored. It isn't getting deleted. So the amount of data storage requirements is just compounding over time, which puts a lot of more demand on our products on an ongoing basis.
And how we're trying to solve for that is really through technology. We're not investing in increasing the number of drive units we are producing. We continue to invest in technology to increase the amount of hits and media that we produce, because that's the big driver of capacity in a drive. So we are currently shipping a 32 terabyte hard drive as our top-of-the-line drive. But starting this quarter, we'll be shipping a 40 terabyte drive. So that's an ability to drive over 30% capacity to customers without adding any unit capacity.

**Ed Ludlow** (2:52)
Could you go into how that works? You know, bring a 40 terabyte drive to market, the engineering complexity of it, and how much do the customers have to sort of re-architect their infrastructure to accommodate for that new generation of HDD?

**Irving Tan** (3:10)
Well, hard drives are actually, as an engineer, I always view hard drives as a marvel of engineering. We use two-thirds of the periodic table in our products. As we move from capacity point to capacity point, there's a lot of re-engineering that goes on into the recipe of our media designs, the recipe of our head design. So it's quite an engineering feat. Obviously, we have great teams that have years of experience doing that. At the same time, we work very closely with our customers to make sure we go through very robust qualification processes, because customers are very sensitive to the quality and reliability of their data storage capability.

**Ed Ludlow** (3:47)
The other sort of highly analogous factor that is in front of you is long-term agreements. With the memory market, we see long-term agreements coming in as well. So you have these in place like 2029, 2030, 2031
How do you know how to price an agreement today, going that far into the future?

**Irving Tan** (4:07)
It's a really good question. We do have LTAs in place locked in till 2029
We have customer demand. Customers are coming us to put LTAs in place to support demand for 2029, 30 and 31 They have really good visibility in terms of the storage requirements that they need because they're looking at the entire AI infrastructure stack, everything from GPUs to CPUs to HBMs to NAND to storage. And they realize they have to take a much longer term time horizon in terms of how they plan. So we are working very closely with them. We're definitely locking in the sort of extra byte supply that they need. But what we're working through with them is still the commercial construct of how we would price something further out in time. Right.