WEBVTT - Western Digital CEO Irving Tan Talks AI Storage Demand

0:00:02.480 --> 0:00:09.560
<v Speaker 1>Bloomberg Audio Studios, Podcasts, Radio News. We're getting back to

0:00:09.640 --> 0:00:13.200
<v Speaker 1>the AI hard disk drive story. I thing. It's great

0:00:13.240 --> 0:00:14.680
<v Speaker 1>to have you back on the show. You know, like

0:00:14.760 --> 0:00:18.640
<v Speaker 1>investors still believe in that story AI hard disk drive.

0:00:19.800 --> 0:00:23.360
<v Speaker 1>Your stock is under some pressure. Just reflect on the

0:00:23.400 --> 0:00:25.479
<v Speaker 1>moment that you tried to outline in this in this

0:00:25.560 --> 0:00:26.320
<v Speaker 1>earnings print.

0:00:27.240 --> 0:00:27.560
<v Speaker 2>Yeah.

0:00:27.680 --> 0:00:29.760
<v Speaker 3>Well, first of all, thank you very much for having

0:00:29.800 --> 0:00:32.639
<v Speaker 3>me back. If you look at the results that we

0:00:32.680 --> 0:00:35.599
<v Speaker 3>delivered in Q fourth it's called twenty six, they were

0:00:35.720 --> 0:00:39.000
<v Speaker 3>very strong. We had over forty percent revenue growth, grass

0:00:39.000 --> 0:00:42.720
<v Speaker 3>margin was above fifty four percent, very strong cash flow

0:00:42.800 --> 0:00:46.279
<v Speaker 3>generation as well. If anything, probably the expectations were very

0:00:46.320 --> 0:00:48.520
<v Speaker 3>high in terms of the increased amount of exo bikes

0:00:48.520 --> 0:00:51.080
<v Speaker 3>that we could supply into the market. But if you

0:00:51.120 --> 0:00:55.560
<v Speaker 3>take a step back, really the fundamentals for ongoing revenue growth,

0:00:55.600 --> 0:00:59.720
<v Speaker 3>margin expansion, cash flow generation, we mean very strong. On

0:00:59.840 --> 0:01:05.479
<v Speaker 3>the demand side, demand from hyperscalarists, neo clouds, physical means

0:01:05.600 --> 0:01:08.800
<v Speaker 3>very robust, and we continue to see over twenty five

0:01:08.840 --> 0:01:12.440
<v Speaker 3>percent kig or demand from a demand perspective over the

0:01:12.480 --> 0:01:16.319
<v Speaker 3>next five years. Simdenly on the supply side, we have

0:01:16.360 --> 0:01:19.639
<v Speaker 3>a very robust roadmap that we have put in place

0:01:19.959 --> 0:01:22.120
<v Speaker 3>that will be able to meet this demand. And so

0:01:22.160 --> 0:01:25.560
<v Speaker 3>it's pretty much a function of just quarterly transitions that

0:01:25.600 --> 0:01:26.199
<v Speaker 3>we're going through.

0:01:27.760 --> 0:01:30.040
<v Speaker 1>I think what you're touching on there is you know,

0:01:30.240 --> 0:01:33.759
<v Speaker 1>storage is distinct in many ways from what's happening in memory,

0:01:34.080 --> 0:01:38.240
<v Speaker 1>in some ways highly analogous. Right. There is a pricing

0:01:38.319 --> 0:01:43.360
<v Speaker 1>story there, there is a demand outpacing supply story, and

0:01:43.400 --> 0:01:48.880
<v Speaker 1>your industry facing the balancing act of expanding output manufacturing

0:01:48.880 --> 0:01:51.720
<v Speaker 1>capacity on a permanent basis or not.

0:01:52.240 --> 0:01:53.440
<v Speaker 2>How are you managing all of that?

0:01:54.680 --> 0:01:57.120
<v Speaker 3>Yeah, So first of all, thank you for highlighting that

0:01:57.200 --> 0:02:01.160
<v Speaker 3>it's a clear distinction between memory, which is like nan

0:02:01.320 --> 0:02:04.240
<v Speaker 3>and d RAM versus storage. Right, because memory and like

0:02:04.280 --> 0:02:07.680
<v Speaker 3>computers being recycled all the time for different workloads, but

0:02:07.760 --> 0:02:10.400
<v Speaker 3>the data that's being generated continues to be stored.

0:02:10.440 --> 0:02:11.600
<v Speaker 2>It isn't getting deleted.

0:02:11.639 --> 0:02:14.320
<v Speaker 3>So the amount of data storage requirements is just compounding

0:02:14.880 --> 0:02:17.640
<v Speaker 3>over time, which puts a lot of more demand on

0:02:17.680 --> 0:02:22.120
<v Speaker 3>our products on an ongoing basis. And how we're trying

0:02:22.120 --> 0:02:25.280
<v Speaker 3>to solve for that is really true. Technology. We're not

0:02:25.480 --> 0:02:29.079
<v Speaker 3>investing in increasing the number of drive units we're producing.

0:02:29.680 --> 0:02:32.960
<v Speaker 3>We continue to invest in technology to increase the amount

0:02:32.960 --> 0:02:35.560
<v Speaker 3>of hits and media that we produce, because that's the

0:02:35.600 --> 0:02:40.000
<v Speaker 3>big driver of capacity in a drive. So we're currently

0:02:40.000 --> 0:02:42.679
<v Speaker 3>shipping a thirty two terrabyte hard drive as on top

0:02:42.680 --> 0:02:45.359
<v Speaker 3>of the line drive, but starting this quarter will be

0:02:45.360 --> 0:02:47.640
<v Speaker 3>shipping a forty terra byke drive. So that's an ability

0:02:47.639 --> 0:02:51.160
<v Speaker 3>to drive over thirty percent capacity to customers without adding

0:02:51.160 --> 0:02:52.160
<v Speaker 3>any unit capacity.

0:02:53.000 --> 0:02:55.880
<v Speaker 1>Could you go into how that works, you know, bring

0:02:55.880 --> 0:03:00.200
<v Speaker 1>a forty terabyte drive to market, the engineering complexity of it,

0:03:00.919 --> 0:03:02.880
<v Speaker 1>and how much do the customers have to sort of

0:03:03.360 --> 0:03:07.760
<v Speaker 1>re architect their infrastructure to accommodate for that new generation

0:03:07.880 --> 0:03:09.239
<v Speaker 1>of HDD.

0:03:10.160 --> 0:03:11.240
<v Speaker 2>Yeah, well, hard drives.

0:03:11.280 --> 0:03:14.080
<v Speaker 3>Actually, as an engineer, I always use hard drives as

0:03:14.080 --> 0:03:16.520
<v Speaker 3>a movel of engineering. We use tutas or the periodic

0:03:16.560 --> 0:03:19.920
<v Speaker 3>table in our product, so as we move from capacity

0:03:19.919 --> 0:03:22.919
<v Speaker 3>point to capacity point, there's a lot of re engineering

0:03:22.960 --> 0:03:25.440
<v Speaker 3>that goes on into the recipe of our media designs,

0:03:25.880 --> 0:03:28.240
<v Speaker 3>the recipe of our hit design. So it's quite an

0:03:28.280 --> 0:03:31.200
<v Speaker 3>engineering feed. Obviously, we have great teams that have years

0:03:31.200 --> 0:03:33.720
<v Speaker 3>of experience doing that. At the same time, we work

0:03:33.800 --> 0:03:36.680
<v Speaker 3>very closely with our customers to make sure we go

0:03:36.720 --> 0:03:41.240
<v Speaker 3>through very robust qualification processes because customers are very sensitive

0:03:41.280 --> 0:03:45.480
<v Speaker 3>to the quality and reliability of their data storage capability.

0:03:47.040 --> 0:03:50.960
<v Speaker 1>The other sort of highly analogous factor that's in front

0:03:50.960 --> 0:03:54.320
<v Speaker 1>of you is long term agreements with the memory market.

0:03:54.320 --> 0:03:56.960
<v Speaker 1>We see long term agreements coming in as well. So

0:03:57.040 --> 0:04:00.240
<v Speaker 1>you have these in place like nine thirty two, twenty

0:04:00.280 --> 0:04:03.200
<v Speaker 1>thirty one, how do you know how to price an

0:04:03.240 --> 0:04:06.120
<v Speaker 1>agreement today going that far into the future.

0:04:07.240 --> 0:04:10.480
<v Speaker 3>It's a really good question. We do have ltas in

0:04:10.520 --> 0:04:13.680
<v Speaker 3>place locked in till twenty nine. We have customer demand.

0:04:13.760 --> 0:04:16.120
<v Speaker 3>Customers are coming us to put ltas in place to

0:04:16.160 --> 0:04:19.640
<v Speaker 3>support demand from twenty nine, thirty and thirty one. They

0:04:19.640 --> 0:04:23.359
<v Speaker 3>have really good visibility in terms of the storage requirements

0:04:23.440 --> 0:04:25.599
<v Speaker 3>that they need because they're looking at their entire AI

0:04:25.720 --> 0:04:27.360
<v Speaker 3>infrastructure stack, everything.

0:04:27.000 --> 0:04:29.159
<v Speaker 2>From GPUs, the CPUs to.

0:04:29.400 --> 0:04:33.680
<v Speaker 3>HBMS to storage, and they realize you have to take

0:04:33.720 --> 0:04:36.440
<v Speaker 3>a much longer term time horizon in terms of how

0:04:36.440 --> 0:04:36.800
<v Speaker 3>they plan.

0:04:36.920 --> 0:04:38.599
<v Speaker 2>So we are working very closely with them.

0:04:39.040 --> 0:04:42.560
<v Speaker 3>We're definitely locking in the sort of exerbyte supply that

0:04:42.600 --> 0:04:44.720
<v Speaker 3>they need, but what we're working through with them is

0:04:44.760 --> 0:04:47.159
<v Speaker 3>still the commercial construct or how we would price something

0:04:47.440 --> 0:04:48.400
<v Speaker 3>further out into right.

0:04:49.360 --> 0:04:52.080
<v Speaker 1>So let me look at it another way. Somebody comes

0:04:52.080 --> 0:04:54.560
<v Speaker 1>to you and says, serving I'm a big customer and

0:04:54.600 --> 0:04:57.279
<v Speaker 1>I want to lock you in till twenty thirty one.

0:04:57.680 --> 0:04:58.680
<v Speaker 1>Why might you say no?

0:05:00.160 --> 0:05:02.440
<v Speaker 3>Well, what we were saying at is, if you need

0:05:02.520 --> 0:05:05.400
<v Speaker 3>six hundred exobites of storage in twenty thirty one, we

0:05:05.440 --> 0:05:08.360
<v Speaker 3>will deliver that to you. How we actually price it,

0:05:08.560 --> 0:05:10.680
<v Speaker 3>we're still in negotiations with them where we might have

0:05:10.720 --> 0:05:13.919
<v Speaker 3>a base price for a percentage of that volume, and

0:05:13.960 --> 0:05:16.720
<v Speaker 3>then a stop tiar pricing structure as we increase the

0:05:16.760 --> 0:05:18.520
<v Speaker 3>capacity we deliver to them.

0:05:18.960 --> 0:05:20.920
<v Speaker 1>I don't mean to trivialize this in any way, but

0:05:21.160 --> 0:05:23.960
<v Speaker 1>you know, like it's the hard dis drive moment, you know,

0:05:24.040 --> 0:05:27.280
<v Speaker 1>make hard dis drive great again. What's that been like

0:05:27.320 --> 0:05:30.640
<v Speaker 1>for you to put the technology at the forefront of

0:05:30.680 --> 0:05:31.760
<v Speaker 1>discussion at the moment?

0:05:32.880 --> 0:05:34.200
<v Speaker 2>Well, it's really exciting.

0:05:34.240 --> 0:05:36.560
<v Speaker 3>I mean people thought that hard drives were a thing

0:05:36.560 --> 0:05:39.720
<v Speaker 3>of the past, but it's shown to be a fundamental

0:05:39.760 --> 0:05:42.839
<v Speaker 3>piece of the AI infrastructure stack with eighty percent of

0:05:42.960 --> 0:05:45.560
<v Speaker 3>all storage in the cloud, and AI is continuing to

0:05:45.640 --> 0:05:47.840
<v Speaker 3>recite on hard drives both today and going forward. So

0:05:47.880 --> 0:05:50.120
<v Speaker 3>it's a very exciting time for us. It's a great

0:05:50.200 --> 0:05:52.000
<v Speaker 3>challenge and I always tell my teams. It's a great

0:05:52.040 --> 0:05:54.640
<v Speaker 3>responsibility that we have to deliver to the industry to

0:05:54.760 --> 0:05:57.680
<v Speaker 3>enable this once in a lifetime opportunity that AI is creating.