Next in Tech | Ep. 270: Agentic revolution at Dell Technology World

View Full Transcript

Eric Hanselman   00:00:01

Welcome to NextTech, S&P Global Podcast with a world of emerging tech. I'm your host, Eric Hanselman, Chief Analyst at S&P Global. And today, we've got a special episode in which we're going to be getting perspectives from Dell tech world. There's a lot that is going on, not just in the Dell ecosystem, but a lot of the associated industries, and technologies that are wrapped up in this. And we've got the S&P analyst team, to talk us through some of the bits and pieces that we were seeing, and some of our perceptions about really what this is identifying, some of the trends we're seeing and the kind of impacts we expect this to have on the broader marketplace, so here we go. And I'm here with Brian Partridge at Dell Tech world. Brian, what's your take and what's going on this week...

Brian Partridge   00:00:53

It's been a really great week. They've packed a lot in. And I got to tell you, I've been coming to Dell Tech World for about the past 10 years. This one has felt different. It's felt a bit more consequential. I think where we are in the AI journey. So from what I've seen and heard, first of all, you got to tip the cap to Dell for the execution that they're displaying across portfolio expansion, ecosystem expansion and really sort of meeting the market where it needs to be. Key takeaways for me, big focus on sovereign AI, sovereign compute, the move to edge inferencing and private AI for inference workloads. I actually had an opportunity to moderate a session for Dell's OEM ecosystem. And a lot of that was about inference moving to the edge and then the implications of what comes next with physical AI and what's the role for the OEM and for Dell to enable those new experiences.

Eric Hanselman   00:01:50

This is really this transition that we're sort of hovering on, which is moving from the training world and inferencing and what those implications wind up being.

Brian Partridge   00:01:57

And it's amazing how fast this has all moved. I mean ChatGPT was 3.5 years ago. We've already been through maybe 2 eras of AI already. And now we're looking at this physical AI era, which is extremely complicated technically to make happen. You've got a lot of fragmentation of devices, data types, a lot of brownfield. So complex challenges, but really amazing opportunity when you think about it.

Eric Hanselman   00:02:23

Well, and all you have to do is sprinkle in a little supply chain difficulty into all of this, and that only complicates things all the more.

Brian Partridge   00:02:30

True, true. But when you think about supply chain complexity, it's one of those things that does play to Dell's strengths. That's always been a core capability of Dell, and they are using that to their advantage, their supply chain as well as their scale.

Eric Hanselman   00:02:44

No, good points. And a lot of things going on here. We'll point you towards a lot of the research that's coming out of our trip here.

Brian Partridge   00:02:50

Thanks, Eric. We'll talk soon.

Eric Hanselman   00:02:52

Thanks, Brian.

Brian Partridge   00:02:53

All right.

Eric Hanselman   00:02:55

We're here at Dell Tech World, and I've got William Fellows with me to talk about what we've seen and some of the perspectives about what's going on. William, what's the big news? Certainly a lot of AI, but a little bit more beyond that. What's your take?

William Fellows   00:03:11

My take is that Dell has hit a jackpot again, Mr. Dell. This just carries on going wave after wave. In the era of AI, Dell is now become a first-class citizen in the way that it really wasn't in the cloud era. Honestly, it's really found its mojo again, what with -- as it says, most organizations now doing AI on-premise. And this week at the show, it's landed OpenAI, it's landed Grok. It's landed some more elements of Google Gemini as on-premise offerings for customers. as well as that. The majority of compute that is selling is now going for inferencing, which is driving more use and more spending. And this company, it barely needs sales because stuff is flying off the shelf. In fact, it's pipeline backlog really is attributable to, I guess, what folks are calling the flashpocalypse that there simply isn't the amount of memory out there that's required to fulfill the orders. So overall, it's a very promising picture for Dell. But this is built on well, now decades of being the preeminent expert in building the supply chain, delivery mechanisms and all of those things that go along with it.

Eric Hanselman   00:04:40

Well, being here in Las Vegas, of course, they put all their chips in on supply chain. And this is, to your point, it's paying off at a point now where in the early stages of this whole shift, we were looking at shortages of GPUs. Well, now, of course, with all the memory concerns, everything else that's going on here, it's become that much more critical to be able to manage that. And then that really seems to be the thing that is paying off or at least generating such a big payout.

William Fellows   00:05:12

Exactly.

Eric Hanselman   00:05:13

Keeping in Vegas terminology in terms of being able to actually deliver infrastructure to be able to feed what is this seemingly insatiable appetite for AI infrastructure. But also if you take a look at some of the ecosystem around it, software capabilities, some of the rest of this, you mentioned that they're now looking at what had been cloud-only approaches, Gemini starting to put some of this on-prem. That starts to open a whole set of interesting doors and some new directions.

William Fellows   00:05:44

Absolutely. And of course, the thing that's not apparent here, that's really the existential threat that it has is -- but that's all under its ability to command because it's been the owner and the keeper of that universe for so long, whereas we used to come to these events and it would be, well, what is going to be the impact of AWS and Google and the rest of it? Well, all of that stuff has kind of moved to the sides now and really the game in town here is eating the need specifically for all this. That's not to say that it hasn't made some missteps and some wrong acquisitions along the way. And if we want to think about, for example, here at the event, I think Adel is now stepping up with a new kind of expertise in the software domain, which it previously hasn't had or it's done a good job in terms of talking to talk about it, but never walking the walk. And now we're seeing some of the fruits of its own expertise and some of its smarter acquisitions now coming to the surface. And I'm thinking in particular here of the Dell Automation platform, for instance. We've talked about all these components. That's whether it's the infrastructure, the GPUs, the AI factory and then all of the third-party system software and layers that are on it. Well, now you can put all that together and manage it with this emerging thing that they call the Dell Automation platform. And I think they've done actually a pretty good job since the acquisition of Cloudify, which is really forming the basis of this, and they're going to gradually move their customers who have been using this historical Cloud IQ product, which nonetheless has served extremely well. It's got 20,000 customers. They're all going to be able to have access to this new platform.

Eric Hanselman   00:07:45

Well, and it's -- I think that the promise of AI has always been this piece that is -- that it's going to help with automation. It's going to help with all the other capabilities around this. But putting AI to work in automation has been sufficiently complicated. That's been a challenge. That transition has been one that has been it's hard to achieve. But when you look at what the automation platform is actually putting together, those are pieces that when you think about some of those capabilities have been a particularly big deal, that challenge in terms of where they fit and how you actually can deliver that value is the thing that I think is really the -- probably the most valuable aspect of that in terms of really being able to leverage automation in that AI sphere and do it in a way that customizable within the kind of scope that enterprises can actually manage that.

William Fellows   00:08:43

Absolutely. And of course, they're going to -- they're being clever about it because they're going to allow organizations to plug in things that are already familiar to them around policy, around guardrails, around data classification.

Eric Hanselman   00:08:56

You get to actually bolt up the stuff you got with all the AI good on top of that.

William Fellows   00:09:02

I thought also Jeff Clark had a very good challenge to the industry because I think the point at which Dell stops is really -- so this whole world of tokenomics or inferencing and how people are going to be able to control or manage that. And clearly, the growth in the use of tokens on the platforms they're providing, that's a world which Dell is not going to inhabit specifically, I don't think. In other words, he's thrown out the challenge, hey, your token consumption is going through the roof. How are you going to be able to manage and control it? And I think that's the next chapter in all of this. I don't think that Dell is equipped at the moment to actually provide that. And I'm not sure that it actually wants to, but that's certainly where it acknowledges the industry is going to have to go to help that.

Eric Hanselman   00:09:58

Well, because we've been through this process, and I guess maybe it's a sign of the level of maturing we've yet, but we've been through this process in cloud with FinOps. And now it seems like we're heading the same direction with AIOps.

William Fellows   00:10:11

Absolutely. I guess the chief concern here is that not all tokens are created equally. And indeed, one task, if you do it in many different ways, can consume orders of different magnitudes of token. So perhaps we're talking here about looking at GPU use. I don't know. I think a lot of that is yet to be decided, but there's clearly a significant opportunity here for folks to dig into this, I would say.

Eric Hanselman   00:10:40

The cost management piece is part of it...

William Fellows   00:10:43

Absolutely.

Eric Hanselman   00:10:44

Well, a lot to dig into, but so much more that I'll encourage our listeners to take a look at the research that we're putting together on this and more to come.

William Fellows   00:10:54

Thanks, Eric.

Eric Hanselman   00:10:55

Thank you.

Eric Hanselman   00:10:57

And I'm here with Henry Baltazar, Henry, what's your take on everything we've been seeing here at Deltek World?

Henry Baltazar   00:11:02

Thanks for having me, Eric. It's typical Dell Tech World, a lot of platforms to front and center, especially I think the big launch for them this week was the PowerStore Elite, which is basically their fastest-growing platform. And with the new benefits, they're going up to higher performance, higher efficiency. And obviously, that's going to be way more important, especially with people struggling with cost of storage and high data growth and all these different issues.

Eric Hanselman   00:11:25

Availability.

Henry Baltazar   00:11:27

It's got to be higher efficiency and people are trying to squeeze as much value as possible soon. These type of upgrades, while we see them all the time are more necessary than ever.

Eric Hanselman   00:11:36

Well, and with the pressure that AI is placing on infrastructure, growing data volumes, these are all things that are starting to add up to, I guess, a critical mass, although I guess you can say we've already been that critical mass in terms of adoption and the infrastructure needs for the capabilities.

Henry Baltazar   00:11:54

And I think as these problems get bigger and bigger and more prominent, not just to the hyperscalers and GPU clouds and every other enterprise, I thought one of the more interesting things that was launched at the show or was expanded upon the show is DexScale storage strategy, which is really Dell's software-defined storage strategy where they're allowing larger customers to be able to deploy a singular hardware platform and be able to deliver specific software-defined storage personalities on demand. So you can go to an object scale, you could do PowerScale if you need scale-out mass or you could even do the light and fall system if I need high performance. But the good thing about this is they're allowing you to switch those personalities and not have to do a forklift upgrade on the hardware. So I think the longer-term implications of this is that's a great strategy for clouds and these other large-scale customers. But wouldn't it be great if they were able to trickle that down to normal customers? And because I mean, one of the big complaints we hear about Dell all the time is too many platforms, too many platforms, too many platforms. You have to worry about too many platforms and too many silos if you can consolidate those resources and enable the right personalities when you need that type of capability.

Eric Hanselman   00:13:00

Well, because it seems like we're living in a world in which none of these storage modalities are going away. And in fact, there's reinforcing for each of these different forms because of the different use cases, the different capabilities. So you really have to be in an all-the-above world.

Henry Baltazar   00:13:16

Yes. And you have to think about it also in terms of what's the secret power of public cloud storage or cloud storage type service. It's the ability to make -- it's not just the elasticity and demand. To me, it's the ability to make something go away. If you have something physical, I can't necessarily make it go away. Here, well, let's say we make changes and maybe we didn't need object storage, maybe I need more NFS or maybe I need even block storage. the extension for that they're announcing this week was on the addition of PowerFlex, which is coming, which will give them a block storage flavor to that exascale strategy. So longer term, when I look at this, that seems like it could be quite compelling, not just for the big customers, but for anybody because we're kind of sick of trying to juggle multiple silos and multiple proprietary platforms. It'd be nicer to have a nice path where I could buy one hardware and be able to augment it to what I actually need.

Eric Hanselman   00:14:03

Well, and as you're pointing out, just all you need is a little bit of supply chain disruption to reinforce the need to be that much more efficient or to make efficiency that much more important in terms of how you're actually managing your entire storage estate.

Henry Baltazar   00:14:15

Yes. And that's the other -- that's probably the other slightly in the back supply chain question came up a lot as well. I brought it up a couple of times. And quite frankly, Dell's competitor strength is, we buy more than everybody else. And by having the volume in those ties, that's why they're not as worried about supply chain issues. Likewise, who else has that type of benefit is only the hyperscalers -- and you could see that in the media company's earnings and how much goes to them. So Dell is trying to piggyback on that type of advantage and their scale and the advantage of having scale, and they're going to use that against a lot of the other competitors.

Eric Hanselman   00:14:50

A lot to think about. Well, thank you, Henry. That will point us towards a lot of our research, but this has been great. Thank you.

Henry Baltazar   00:14:56

Thanks for having me.

Eric Hanselman   00:15:00

And I'm here with  Gregg Macatee to talk about AI infrastructure.

Eric Hanselman   00:15:04

What are you seeing here? And there have been a lot of announcements. So curious what your take is.

Gregg Macatee   00:15:10

Yes, really remarkable progress by Dell over the last year between the new Power Edge servers, the storage systems, the parallel file system, Project Lightning coming to fruition and everything else they brought the picture really kind of brings the whole AI story into focus as enterprises work to become AI factories of their own.

Eric Hanselman   00:15:26

Well, if you think about a lot of the core capabilities, they've really evolved. You mentioned Lightning on the storage side. If you think about what they're standing up in terms of capabilities, that even extends out to some of the core functionality in the physical systems out to racks and management, improved cooling, improved operational capabilities, really a much more complete story at this point.

Gregg Macatee   00:15:52

Yes. And what really impressed me is what we heard so much from Dell about is the whole customer Zero story end-to-end how they're working to make their customers' lives better. You're getting from at the very outset, being able to survey them, get a sense of their state, what are their problems with their operational challenges, what business opportunities they have and kind of walking them through the entire process end-to-end, ultimately becoming the factories themselves.

Eric Hanselman   00:16:13

The extent to which they're walking the walk, they are actually putting these capabilities to work in their environment as well. I thought that was the interesting part of some of these use cases, where they're actually leveraging AI functionality. And of course, they've got the infrastructure to support it. Yes.

Gregg Macatee   00:16:30

And another thing to point out is just the whole Agentic story. I know we're still pretty early innings overall, I would say, for customers at large, but that was a really, really big theme during both keynotes yesterday and today.

Eric Hanselman   00:16:41

Well, all things Agentic are clearly where we're actually starting to put AI to work. And that translation though, it's got to have the supporting infrastructure to manage it. And it's interesting looking at some of the -- both the private AI messaging, a lot of the move towards really rationalizing what enterprise infrastructure for AI has been all about. We start thinking about what that means today, a hybrid takes on a very different set of meetings and consequences for which you got to build.

Gregg Macatee   00:17:13

Yes. And something else is really impressive. Dell kind of meeting the customer where they are in their journey being open and hybrid both equally. I think there were plenty of exhibits about the overall rising token costs. The overall cost of the visual token might be going down, but the overall consumption is going up exponentially faster, leading to these overall increases in costs. And Dell recognizes that.

Eric Hanselman   00:17:36

Not to go right back into talking about Devon's paradox again, but it's the move towards making per token cost slower, raising consumption. I wonder whether or not we're getting towards a point at which concerns about cost start to do the same kind of things that we did in cloud and do sort of a FinOps or at least an operational cost management kind of approach.

Gregg Macatee   00:18:01

Yes. We kind of hear that from customers across the board. I think kind of the #1 inhibitor and getting back to Jeff's paradox is just that upfront cost to be able to get into the business.

Eric Hanselman   00:18:11

Yes, that's true because you're actually establishing that base footprint is not an insubstantial cost.

Gregg Macatee   00:18:18

No. No, it really is very, very expensive. And again, getting back to that hybrid story, Dell meeting the customer where they are, it's a combination of on-prem and in the cloud, open models, closed models that organizations are ultimately going to use to be able to fulfill their objectives.

Eric Hanselman   00:18:34

Well, I thought you were mentioning the keynotes, Jeff Clark's comment that token cost is going to be a line item on budgets, and this is something that organizations are going to have to be able to work through and get sufficiently capable at managing to be able to deliver...

Gregg Macatee   00:18:52

Ready for. Yes, we hear about these use cases like coding. Everyone's coding in the business these days, but what's that bill going to be as you get more used to it and get that experience around it, what's that going to end up being for the customer?

Eric Hanselman   00:19:05

But clearly, a lot of capabilities that are necessary to be able to get to all of the promise that is the AI and the gentic future.

Gregg Macatee   00:19:14

Yes. That's just one part of the story. We'll certainly see how it plays out as organizations gain more experience using these technologies. But definitely really Dell was on the identifying that in both keynotes.

Eric Hanselman   00:19:26

A lot of good things. All right. Well, so much more to see, but links to the research below. So...

Eric Hanselman   00:19:32

Thanks, Greg.

Gregg Macatee   00:19:33

Thanks, Eric.

Eric Hanselman   00:19:38

That's a great set of perspectives from the team. And I'll add my take as well. When you think about the many announcements that have taken place, for me, one of the most impactful is the Dell Automation platform. And it's not about the issues around automation. I think the ecosystem integration pieces around this are really important and impactful. But the bigger part relates to some of the issues around tokenization and the token economics or tokenomics that we've been talking about. And the idea that automation plays a key role in the way in which we manage costs in the age of AI to be able to automate the process of selecting the appropriate models and appropriate token costs for the particular workload that's being accomplished and to be able to schedule that, to be able to manage it and be able to handle some of those challenges in terms of how all of this really gets delivered in a much more cost environment. We've had the movement towards FinOps in cloud capabilities. We need to be able to head towards that in the AI age as well. But that's it for this epsiode of Next In Tech. Thanks to our audience for staying with us, and thanks to our production team, including Sophie Carr, Feranmi Adeoshun, and Kyra Smith and Dylan Scheible with Marketing and Events teams. I hope you'll join us for our next episode because there is always something next in tech.