Michael Dell's 40-Year Tech Journey: AI Supercomputers & Innovations | WSJ CEO Council

And the Wall Street Journal's Christopher Mims. All right. Oh, hot mic. Welcome to the Thunderdome. We put it right in the middle here.

Wow.

Yeah.

Is there a cage that's going to come down?

I mean, maybe. We'll see who can push the other one off the stage. This was— actually, I was told we would do this to better deal with the hecklers. So I don't know how your crowd work is, but we're going to find out tonight. So let's start with like a little bit of your origin story, because you've been at this long enough that I mean, while you are familiar to this audience, I think that, you know, you and your company have had 9 lives, really, or more. And your latest one is being this pick-and-shovel dealer in the AI gold rush. Is that fair to say?

Sure, I'll take it. Yeah. And what I find interesting— I'll allow it. Yeah, go ahead. Okay.

What I find interesting about that is it feels like such a logical extension of where you started, right? Like in your dorm room, souping up these stock IBM PCs. But now you're saying, hey, you're tweeting or whatever it's called now on X, look at this 120-kilowatt server rack that's the most powerful in the world, and we're the first to sell it to an end customer, right? Which is CoreWeave, the AI hyperscaler. I mean, do you see kind of a through line there? Do you feel like you're still getting to do what you loved when you started out in that dorm room in 1984?

I do. And I love technology. I've always been fascinated by its power and how it's been enabling, you know, humans to progress. And yeah, we started out as a, company upgrading, you know, IBM personal computers. And then we started, you know, designing our own computers. And since then, we've evolved to be the largest IT infrastructure company in the world. And so certainly, you know, when you, when you fast forward to 2025, right, with AI, The ingredients for AI are algorithms and data and computing power and networks and storage and latency and bandwidth. And those are a lot of the things that we do. So it's, it's a, it's, it's incredibly exciting time. I think it's a sort of revolution in how technology is being used in the world and certainly a big change in computer architecture and We have a lot of capabilities that are highly relevant, you know, whether you're doing a massive training model or you just want to unlock the power of the data inside your corporation, or, you know, you want a new PC that does a whole bunch of stuff your old one didn't do. And you—

I mean, Dell has seen steady growth, you know, for decades. But I imagine that this latest boom in some ways must have caught you by surprise. The AI boom, that is.

Yeah, I think the large language model wave, I don't think many people were expecting that to occur. And, you know, the amount of data in the world has been growing very, very fast. sort of the rate of the, the amount of data, you know, doubling has been shrinking, right, all the time. And we kind of knew that this data was really valuable, but nobody had really figured out how to unlock it. And, you know, back in 2016, we combined with a company called EMC, and we really solidified our position in the cloud and data center as the—

Because EMC is a storage company.

EMC was a storage company, and now, you know, we brought together the number one in all parts of infrastructure as one provider. That really respond— you know, customers really responded well to that.

You mean a provider of the real physical substrate?

Of the infrastructure, right? All the things that go inside that make it work. And, you know, certainly 10 years ago we had machine learning and earlier versions of AI that were, you know, some organizations were starting to use, but the large language models have just taken it to a whole nother level.

Why do you think that is? I mean, what you were telling me when we were talking backstage that, you know, a couple years ago you had 20 or 30 customers for your AI servers, and now you have 2,000. So that's incredible growth in 2 years. Why? What has changed that suddenly everybody can do this or wants to do it?

Well, you know, it's because these large language models are just an enormous breakthrough, and it's pretty easy for customers to apply data to them and to have pretty astounding results. And so organizations have figured out that, you know, 20% productivity efficiency opportunities are very much achievable with some sightings of 30% and 40%. And, you know, if you're one of the CEOs in this room and you hear about that, you wanna learn, you wanna learn about it and understand how do I get that inside my corporation? And, uh, you know, if, if I get it before my competitor gets it, then I have an advantage. If they get it, you know, before I get it, then I'm— then I've got a problem. And so, yeah, this, this wave is, is coming quickly, and we see it's the number one topic that our customers want to talk about all over the world.

And whenever there's growth that's this fast and on this scale, uh, there's always a talk of, is this a bubble, or when is this going to cool off, or when is this bubble going to burst? I mean, do you have a sense— some have argued that, you know, some of the spending by some of the best-funded AI startups is just not sustainable given the revenue numbers that we're seeing. Um, do you have a sense that that there's gonna be a cooling-off period, that this bubble might burst? I mean, you had, I think, 60% growth in your AI server division last quarter. I can't even remember if that was year over year, but the quarterly growth was unbelievable. I mean, are you gonna be telling perhaps, you know, in a year or two, like, well, we saw that cool off significantly because of this, you know, the heat has come off of this?

Well, it was, it was 58%, uh, and the quarter before that it was, it was 80%. The quarter before that was 42%. But that's not AI servers, that's actually all servers. That includes traditional servers, which are also growing fast. We're, we're gaining share there. You know, to me it looks like we're in the first 10% of the S-curve adoption here.

Of AI?

Of, of AI. I, I also believe that any technology that is sufficiently game-changing by definition, there are going to be some excesses and some mistakes that are made. But that doesn't mean that it's not, uh, you know, game-changing and, and, and, uh, and super important. And so, um, I think, you know, there, there was a, uh, survey that came out, um, about a month or so ago from one of the big banks And they surveyed, you know, a couple hundred companies, and 90% of the, of the companies said that their AI projects had either met or exceeded expectations. 40% said that it exceeded expectations. And so I'm not seeing, you know, a lot of the 2,000 customers that we're selling these to come back and say, you know, we, we tried AI, it didn't really work, and we're not going to do that anymore, right?

I've, I've yet to Yeah, I mean, this is a place where it's useful to make a distinction. You're selling to customers who are savvy enough to come to you and say, we wanna buy our own AI hardware from you, uh, you know, much less we're building our, making our own software for AI, you know, using our own data. So these are pretty savvy customers, right? Because there has been, I think, some disappointment for end users of AI. who are just using something that is off the shelf or is popping up in a productivity app. I mean, there has been a lot of disappointment there. I mean, we've covered it, others. What's the divergence there between what you're hearing from customers where they're satisfied and they're gonna come to you for more versus folks in other applications where there's some disappointment?

The areas where I think customers are seeing success are where they tend to identify the 2 or 3 things that are most important in their business. And they organize all the data, they make it a big priority, and they really build out a whole process to improve that using technology. And so, You know, it's kind of easy to find a project that you could do with AI, but to really—

What are some starters? What do people like to, when they're coming to you for the 101?

You know, the approach that we recommend is first you start with what problem are you trying to solve? And it better be a big needle-moving problem. Like, you know, let's look at the major processes inside your company And let's go after the 2 or 3 of those that use the most number of people. And how do you organize all the data? Because in many cases the data was on, you know, hundreds of individual islands. It has to be brought together. You have— it takes some time to organize it all. And then you've got security and regulatory issues. But once you do all that, you can unleash some pretty epic improvements and you can start to reimagine the business because now you're enabling all sorts of new use cases.

Does this tend to be back office processes? Is it customer service? Like where are the places where the AI is best suited? Right. Because there are certainly applications where it's not. Large language models are not good at math, for example, famously and weirdly.

So where are— although they are getting better, the, you know, the most common use cases that we're seeing are coding assistance, you know, in software development, which is also interestingly improving the quality of code that's already been written, customer service and, uh, you know, sales and marketing content, many, many back office functions. You know, we have an interesting use case around documentation where we create a lot of documents inside our company, and these tools are spectacularly good at creating documents in a given form. And then, you know, we do it in 180 countries, so we need to translate them. And, uh, you know, that process can get, you know, way more efficient for us. I'll give you another example in our case, which is one that, that, uh, you see very often inside big companies. We have 30,000 people that do support, and we have an enormous database of previous support calls. We have all these white papers and, you know, warranty data and, uh, you know, lots of documentation that's been written about how to solve problems, Jira databases, And now we have this tool called Next Best Action that sits on the right-hand side of the screen. And so as somebody contacts us, the Next Best Action tool is listening to the call and it tells the agent what is the absolute best way to solve that problem. And it gets smarter and smarter every day.

And this is called Next Best Action?

Yeah.

So when can I get this for my own life?

It's like you've got a little genius, you know, on your shoulder and it's telling you exactly how to solve this problem.

Uh-huh.

Now, we're actually taking these tools and we're putting them in our more complex products so that if you're an IT person, you know, inside an organization and you're responsible for making all these things work, this tool will help you make sure your configurations are right or answer any of your questions And, you know, it's genius-level knowledge that is beyond the capability of any single human.

So it's the Wall Street Journal, so I'd be remiss if I didn't ask you a few things about the news. You were part of President Trump's CEO Council the last time he did that, you know, brought a bunch of CEOs like yourself together. Has there been any call for that again yet? Is there gonna be a CEO Council 2 at the White House that you know of?

I don't know. You should ask the president. I don't know.

So, but if you get an invite, will you and Tim Cook and everybody else— well, you can't speak for Tim Cook, but would you go back and be part of that advisory?

We'll see.

Okay.

Maybe.

Okay. So, you know, Intel's one of your biggest partners. I know that you are, you know, some of Intel's challenges I think could be a big threat for some companies, maybe not as much for you because you partner now with so many software and hardware providers, right? Like you guys are really tight with NVIDIA.

Yeah.

and you are using Meta's AI model, right, on your servers, 'cause it's open source.

And we use AMD chips and Qualcomm chips also.

So you're like Switzerland, you'll use anybody's stuff.

If it's good, yeah. We won't just use any, I mean, it's gotta be really good.

Okay, all right. So in this age of AI, there's this really curious phenomenon where Meta, of all companies, is becoming this really powerful player in terms of open source AI, and people are using it, and I hear about it all the time from the folks that I talk with. Is it, are they almost like the, what Linux, you know, as an open source operating system was? Is that what Llama is to AI?

I think it's a, it's a reasonable analogy, maybe at that sort of frontier model level. But you also see an enormous kind of dispersion and variety of these open and closed source models, large and small models, models for specific domains. And there's just, uh, it's just brimming with innovation.

And so your customers using all these, are they Do they come to you for advice?

They often come to us, you know, asking for help and, you know, what are the tools that they need to use to be able to solve the problems that they're trying to solve. And we also build these architectures so that, you know, if one company has a great model, but then somebody else comes along and it's better, we just use that one. So we're not hard-coded to really any, any of the models. They can be, and I think they're gonna evolve tremendously. They're super powerful, they're very impressive, but they're also the worst they'll ever be.

Hmm.

And they're gonna be way better in a couple of years. And most of the data, let's say, that is inside, you know, the organizations of the CEOs that are here, hasn't been touched by these models. It's private data.

It's proprietary data. But there's good reason why that data hasn't been touched yet. Hmm? There's good reason that data hasn't been touched yet.

Exactly.

Laws and stuff around that data.

And, and, and you know what, what, uh, what those 2,000 customers that we're working with are figuring out is I can take my private data, use one of these models, and I can unlock the power of that inside my corporation. And I've kind of kept that data. I want to bring the AI to where my data is. I don't want to send my data off to somebody else.

Now, with great power comes great responsibility. If you're enabling these companies to consolidate all that data, make that knowledge available to everybody across the company, I mean, there's a lot of pitfalls there. I mean, you could have folks in one part of the company suddenly accessing secret HR data because the helpful chatbot can't wait to cough it up. I mean, how do you help people set guardrails in those contexts? Because it seems like there's a lot of pitfalls.

Yeah, I mean, you have to organize the, the security and the privacy and even the, the responsibility in accordance with, you know, how you would manage sensitive information within your corporation and, and with the, the laws of the land.

So let's, let's talk about Elon Musk. You have a direct connection to Mr. Musk because you provided a lot of the servers that went into the new Colossus.

Yes.

Is that the name of it? I just heard it described as an AI supercomputer.

Uh, it's, it's, it's called Colossus.

And it's in Memphis, Tennessee.

Correct. Yes, I've been there.

You've been there?

It's very cool.

What does it look like? What's it feel like to go there? Is it a giant hall of computers?

Looks like a great Dell commercial is what it looks like.

Is it going to be featured in a Dell commercial?

No plans to do that. But yeah, I mean, it's an amazing engineering feat. And, you know, the— I mean, he placed incredible demands on us and many others. And we were honestly super excited to be working on that project. And, you know, we continue to be engaged with them and continuing to build out their capabilities. Very impressive.

Is the fact that you're neighbors, did that help you get that project going? Did he just kind of walk over and knock on your door?

Well, he, I mean, we, I see him, he's in Austin, right? So yeah, I mean, he came over, I went over to his place, he came over to our place, we showed him, you know, the servers and what we were doing. And we've been working with their team at Tesla and xAI for some time. And, but certainly it helps to be in Austin. Austin's great.

And so, As that gets built out, I mean, do you think that there is an— I mean, you're clearly an AI optimist. Like, there's— these models are the worst they're ever gonna get. Is there in some sense then at the high end, at the AI supercomputer end, do you think there's an arms race where whoever has the biggest AI supercomputer wins in the end?

There's no question there's an arms race, right? And there are, you know, at least 4 or 5, maybe 6, you know, companies that are playing in that arms race. And, uh, you know, they're all rushing forward to build out their algorithms, build out their computing capability, and continuing to layer on additional innovations on top of the, you know, models that exist.

So Austin, Texas, People forget that the integrated circuit was invented by Texas Instruments, right? It wasn't just all these boffins in Silicon Valley. You, you early on in your career were encouraged to move to Silicon Valley, right?

No.

Oh, okay. Are you sure? I swear you said that in your book.

No, no, I don't think so.

Okay. I don't think so, no.

I was born in Houston. I went to Austin to go to the University of Texas, and I started in my dorm room, my freshman dorm room.

Right. But did you ever feel tempted to move to the West Coast because you were building a computer company, or you were like, I'm always an Austin stalwart?

No.

So you've always been there. And I mean, I think if there's any single person who's responsible for the kind of state of Austin's current tech boom, I feel like it's you as the sort of Macy's anchor tenant of the giant sprawling tech mall that is Austin. What's your perspective on that now? I mean, it really just feels like it's booming. There's cranes everywhere. You drive out of the airport, the first thing you see is the Cybertruck factory. Elon Musk is building a company town in Bastrop. I mean, do you sort of look at that and just feel like Ozymandias and like, oh, I survey what I've done here?

I've never had that thought, but it's certainly been fun to see Austin grow and to be a part of that. There are, I think, 5,000 startups now in Austin. My fundamental belief is that whenever you find great companies, there's a great university right nearby.

Mm-hmm.

Austin has the University of Texas, but Austin, you know, Texas also has like, uh, you know, 1.8 million college students, uh, the second of any state, you know, in the country. Um, you know, it's got more Fortune 500 companies than any state in the country now.

And you keep picking up new ones, even well-established ones keep moving to the state.

Well, apparently people are figuring it out, you know, so Yeah, this is true.

Figuring out what exactly? Like that you have this nexus of talent and pro-business environment and the rest?

That's all true, yes.

How does Dell interface with that? Has it been tough for you to hold on to talent as all these new folks move to town?

It really hasn't been that tough. You know, I think part of it is you have a lot of young people in, in, in the, you know, uh, so there's sort of a, an area, uh, you grew up in San Antonio, so you know, you know, we have San Antonio, Austin, Houston, Dallas, and now these are, uh, 4 of the 10 largest cities in the country. And, you know, 1 out of 10 children born in America is born in Texas. And so, um, Yeah, it's a great place, and there's plenty of room for more companies. We're still allowing new people to move, last time I checked.

There's still plenty of land.

Yeah, still plenty, still plenty. Yeah.

So in the brief time that we have—

It's the greatest country in the world. Texas is?

Yeah. Well, I have a British friend. He calls Texas the America of America. Yeah, so clearly that's what's going on here.

Well, I mean, if you, if you, you know, you go do this in the Wall Street Journal, you know, make a chart over the last 30 or 40 years and compare the economy of Texas with the economy of the United States without Texas, and you'll notice a distinct trend. Texas has been doing quite a bit better for a long time.

So is Texas the next California in terms of tech?

We don't want to be the next California.