On this episode of the Six Five – On the Road, hosts Patrick Moorhead and Daniel Newman are joined by Dell Technologies‘ Kyle Dufresne and Ihab Tarazi, who dive into the State of AI Adoption with Dell Technologies. This discussion unravels Dell’s positioning and strategy in the evolving AI landscape, exploring their unique value proposition, as well as, upcoming insights from Dell Technologies World.
Their discussion covers:
- Dell’s opportunities in GenAI following NVIDIA GTC
- The advantages of partnering with Dell for enterprise AI solutions
- Anticipated highlights and what to expect at Dell Tech World
Interested in learning more? Watch the Dell AI Factory with NVIDIA from GTC – Six Five – On the Road with Varun Chhabra.
Learn about how Dell Technologies accelerates your journey from possible to proven by leveraging innovative technologies, a comprehensive suite of professional services, and an extensive network of partners at https://www.dell.com/ai.
Register for #DellTechWorld in Las Vegas to hear how Dell is unleashing the #AI revolution and igniting the power of technology at https://dell.to/3PCBdBR.
Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.
Or listen to the audio here:
Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we ask that you do not treat us as such.
Transcript:
Patrick Moorhead: The Six Five is back and we are talking about our favorite subject that is AI. We’re talking enterprise AI here. Dan, it’s been two weeks since GTC, and we’re still talking about the reverberations of that event. We’re talking about implementation inside of enterprises. We’re talking about competition. We’re even debating networking. It’s great. Great for analysts to have this type of interest and pandemonium.
Daniel Newman: Yeah Pat, look, we are at the beginning of a multi-decade, probably once in a generation technological shift that is going to change the way we live, the way we communicate, the way we work. I don’t think we even fully recognize how significant this is, but it is so exciting. I mean, I think we talked about this on other pods, Pat, but we are seeing stadiums filled with people interested in the next chip coming out. You and I like to say, silicon is back baby, chips are back baby, but it’s systems, it’s services, it’s consulting, it’s design, and in the end it’s going to drive so much productivity and efficiency, Pat. GTC was a moment, but there’s so many more moments, like, I don’t know, maybe Dell Technologies World, where we’re going to hear more about this.
Patrick Moorhead: Exactly right. And one of the things we like to do is say, where are we on this map? So we’re going to be talking the state of AI adoption with two folks from Dell, Kyle and Ihab. Welcome to The Six Five, great to talk with you. Ihab, it was great to know that you check into The Six Five every once in a while. This is great.
Ihab Tarazi: Good show.
Patrick Moorhead: Yeah. No, it’s wonderful. And anytime anybody ever says, oh, I follow you on social media, I apologize first, but it tends to be a good avenue to educate people and create conversation. So Kyle, Ihab, welcome.
Kyle Dufresne: Thank you for having us.
Ihab Tarazi: Thank you for having us.
Kyle Dufresne: We appreciate it.
Daniel Newman: Yeah. We have a lot of fun here, Pat, and it’s been humbling, exciting, to see the podcast explode. You and I have been in airports and we’ve been walking down streets in cities and people are like, oh, we love the show. And AI of course has been a propellant, I mean, look, you guys heard both you, Ihab and Kyle heard the preamble, Patrick, myself. This enthusiasm comes from these opportunities as an analyst to just literally spend every single day on a topic that everybody is so excited about. It’s markets, it’s school teachers, it’s doctors, it’s lawyers, whatever industry you’re in you’re going, oh my gosh, it’s all about to change. And I’d love to start off with getting a little bit of a generative AI lens from the both of you. Coming out of GTC, what is Dell’s opportunity? I get it with AI, I get it with GPUs, I can get that, we’ve been doing this for a while, but the generative AI opportunity, how are you guys sort of communicating that to your customers?
Kyle Dufresne: Well, it’s interesting. And Ihab, it’s okay, I’ll start and maybe you can add on. If that’s okay?
Ihab Tarazi: Sounds good, yeah.
Kyle Dufresne: And thank you for having us. This is the most exciting time. You can see Ihab and I are pretty old, which means we’ve been around a long time and that means nothing else. And Ihab and I were just having a discussion just a few weekends ago on an evening, and we’ve never seen anything like this in the marketplace. And you hear buzzwords like generative AI, and it’s wonderful to see this happening, but the big things we took away from GTC and from the subject of GenAI, is the companies that do this literally are going to change the world. And more importantly, the companies that don’t do it are going to be left behind.
And this is in the world of product development, sales productivity, the way we support our customers, the way we drive new capabilities across every single industry in the world, it’s literally changing things. And there’s hundreds and hundreds of use cases that are being developed, there are ecosystem partnerships that are being developed. And I think you’re right, this is like day one, and I think we’re on a multi-year journey of this day one. And it’s just so exciting to see the value that’s being created in the marketplace, and what the world is going to look like in the next five years.
Daniel Newman: For sure.
Ihab Tarazi: Yeah, I think it’s same for me, you said it so perfectly Kyle. And the excitement is enormous because, I think Dan, you said this before, the advancements in silicon are amazing, and when the H100 came out and being able to get eight of them together working with very high speed NVLink is such a game changer. It enabled applications never possible before, it’s a big moment, but it hasn’t stopped, every new day there’s another leap of silicon from somebody. So getting silicon innovation and then networking is hot again: software stacks, the models, the frameworks. It’s like every piece of the stack is alive and the customers are excited about it and able to generate benefits from it. It’s early stages, but the promise is incredible.
Patrick Moorhead: Yeah, it’s pretty exciting. And you talked about being old, I like to call it experienced. My bestie, Dan here, loves to call me a boomer. I’m a Gen X, that puts us in there-
Daniel Newman: Sure buddy. Sure.
Patrick Moorhead: The only thing that comes close to this was the build out of the internet and what it did. It created an entire wave of data center technologies and build out, an entire… All the PCs, right? There were internet PCs, there were non-internet PCs, just like we’re seeing AI PCs and things like that, so it is super exciting. And technologically, I love to focus on, and I’ve done this for 30 years, is the quadrangle, which is any major inflection in a marketplace. You have compute, memory, storage and networking, and all four of those have to be in sync.
Maybe there’s a fifth, which is accelerator/GPU, I don’t know, but those always have to be kept in balance. So let me go to the customer side here, I have to ask, why should customers work with Dell on their enterprise AI versus other competitors? And you can name them, not name them, but you have the on-prem folks that you compete with and you also are competing with the hyperscalers as well. And maybe Kyle, start off with you.
Kyle Dufresne: Yeah, sure. And thanks for the softball, we love to get that.
Patrick Moorhead: I don’t know, I usually don’t put it in there.
Kyle Dufresne: No, no, it’s a great question, and we are so grateful at Dell to be part of this journey. So first of all, it’s interesting your paradigm about the changes in technology because if you map the things from silicon, to the internet coming, to different speeds in the internet, so 4 and 5G, and you look at Dell technology’s growth over every one of those inflection points, you’re going to find a really tight correlation, and we see this as this next big thing. And if you look at the marketplace, we are uniquely positioned to help customers across the continuum of their AI journey. Yes, we make great servers and networking and storage, but this is really, truly, an outcome-based relationship motion, customers are in all different parts of their journey. We have customers who are still trying to figure out their business case and their use cases, and that’s a really important start.
And we help them really define those things, we help them weight them from business return and technical viability. How do we group them? Understanding where their data is. Is their data correct? Because everybody says, when you put data into the model, you’re going to get data out of the model. So if it’s bad going in, it’s bad going out. So helping people organize their data, understand where their data is, make sure it’s the correct data, what models? Ihab talked about, what models do we do? How do we group these different use cases so that we can use similar models? What are the best models to use? A lot of people are talking about large language models, but in the enterprise world, we don’t think that’s going to be used very much. There’s RAG models and all different types of use cases it’ll fit, but really working with customers to understand how to do that. How to create sandboxes? The biggest thing is, we see companies, even like Dell. At Dell we’ve looked at over 850 use cases, 850.
How do you organize? How do you get ready? How do you launch them into a sandbox type of environment, make sure they’re perfect and then launch them into the enterprise? I haven’t said the word server, storage and networking, but that’s when that comes in, and it’s just a part of what we do when we optimize that stack to work with these customers. So really it’s about the journey, it’s about getting business return. And we believe that AI is going to be done where the data is. And many of our customers have their data on-prem and I think it makes a lot of sense from a cost standpoint, from a data sovereignty standpoint, from a data protection standpoint, to continue doing that on-prem. And we’re going to be the perfect partner to work with them across that continuum. IhaB, would you expand on that?
Ihab Tarazi: Yeah. I think you did pretty good setup of the engagement. We see two types of customers for the most part. One type who has their own software stack, they’re developing a model, they’re building software, they’re building their own GenAI to optimize their business. And those customers are very savvy, there the engagement is about what GPU or CPU to pick, it’s not easy to navigate. And the CPU world, more as well, is you double performance every two years. We’re getting double performance four times a year here. And then the inferencing server, the software itself is more than doubling the performance. You saw at GTC, the last TensorRT-LLM from NVIDIA makes it a 5x performance, so that is a huge discussion.
What GPU to pick? How do I protect my investment? And second topic, how do I optimize network storage, compute, as you said, to make it a system that’s optimal with the highest throughput. The second set enterprise, as Kyle talked about, they need help. What use cases? Should I go… How do I create? Hybrid, public or on-prem? What’s the TCO? There, it’s more about building a token generation cloud. What’s the latency to first token? How many concurrent users, tokens, throughput? And we help them with that whole design both from an infrastructure as well as software and tools.
Daniel Newman: So one of the things that… So we’ve done a lot of research in data, Ihab and Kyle, on the overall state of AI. And one of the things that we’re finding is, one is, why and how vendors are being selected is changing in the era of AI. We actually saw a really high propensity to vendor attrition and new vendor selection, and we did a really significant survey. And the things that were coming to mind to people were effectively, one is, there’s a big question on capabilities. How capable are these companies and how efficient they are? So you’re seeing this sort of board edict flowing down in organizations, and we joke, we call it the do AI edict. The board’s telling the CEO and the CEO tells the CIO and the CIO goes to the… And so everyone’s kind of out.
And we also, by the way saw, this last quarter, I think it was Accenture, did 1.1 billion in the first half of the year of generative AI work. Huge number, more revenue than all of the generative AI VC backed companies on the planet right now, just out of Accenture. If you guys don’t mind me asking, it sounds like you’re talking to customers. I’m guessing you’re sort of in a spiderweb of engaging with other, you have cloud providers and you guys are in there and companies like SIs, these are partners of yours. You have chip makers in there telling their stories, you have ISBs. How complex is this spiderweb? And how do you kind of rise above right now and become a really world-class partner to these companies?
Kyle Dufresne: That’s actually such a great question. We are in the center of this right now, and I think you Ihab did a nice job kind of bifurcating the type of customers we see. We have the large cloud service providers, the CSPs who are buying from us, along with some of these ISBs who are creating these models. That’s one sales motion, and the demand that they’re seeing is tremendous. And for us, it’s all about high quality products, the ability to execute. When you look at Dell Technologies, it’s nothing new for us to say that we have the best supply chain in the world. We’re obviously a large customer with incredible scale, and we have the unique ability to deliver these complex technologies really quickly in the marketplace and have customers become operational very, very quickly. We think it’s a differentiator for us.
But the second piece that’s an interesting thing you’re seeing, we are seeing a group of customers on the enterprise side, which is a very different middle, and they come to us a little bit with a FOMO type of environment, fear of missing out, because there is a lot going on here, there has been some shortages in the marketplace. And in those cases we’re also able to deliver very quickly, but often we take a step back and really try to understand what those outcomes are going to be. Because for us these are marathons, it’s not a sprint, and we want to make sure that we are adding value over the long term. But you also mentioned that ecosystem, because it’s been incredible. There are the big companies like Accenture and E&Y and Deloitte and many, many other ones who are working with customers to understand their data, understand strategies, and we’re linked in very closely with those customers and how they’re going to market.
We’re forming these partnerships and have formed those. But on the other side, there are great partners with companies like IBM and many, many other ISVs who are creating individual solutions. And these come from very, very large customers to startups that are coming out every single day that are offering vertically integrated solutions. Dell is really at the fulcrum of a lot of this happening, and we’re a key partner to all these players in the marketplace. Because at the end of the day, we believe the winners in that space, in the enterprise, are the ones who are going to add the most value in the outcomes that we’re trying to deliver, and we think that’s such a key piece. So it is a spiderweb, and I think it is really, really complex, but it’s one of the things that we believe is one of our best values.
Daniel Newman: Yeah. By the way, previous question Kyle, you talked about data management. What we’re seeing is the biggest inhibitor to large-scale generative AI roll-outs inside of enterprise, is classic garbage in garbage out, we learned this in the ’80s. It’s the same and it’s amplified because now we have data domains, ERP and SCM with PLM, with every as a service you can think of, and it’s just harder. And I’m curious, Kyle or Ihab, how are you helping clients with data management? Again, this is what our research is suggesting, which it’s, listen, we’re 15 years into the public cloud and 75% of the enterprise data is still on-prem or on the enterprise edge, and there’s a reason. And what nobody is telling me is that they’re going to just magically start uploading to the public cloud all of this data to take advantage of generative AI. So how are you helping customers in their data journey?
Ihab Tarazi: I can take that, Kyle, if it’s okay. So data and storage are coming together for customers, and the best way to answer that question is, there’s three phases customers are in, in terms of how they want to manage the data and the data journey. I agree with you 100%, data is the biggest inhibitor, it’s the biggest question, biggest topic. The first phase we see are people who are trying to chain models. And the data challenge there is, how do you feed enormous amount of data on a continuous basis for these GPUs that are very hungry for data? And Michael coined this term, GPUs are hungrier for data than CPU, which we really believe and see from day-to-day operations. So step one for most of these people, how do you keep their GPUs fed with data? And the competitive advantage we have at Dell, you can’t do that by selling one component, you have to design the server.
Our XE9680 is designed with throughput and data in mind from the inside-in. Also, how you optimize your network. We have Dell networking, the expertise. And then, how you optimize storage and distribute at high-density. So it takes a lot. That’s the first data challenge, and we’re doing a lot of work on that. The second challenge is when you set up your retrieval augmented generation model, then you want to take a lot of files and put it in, and there you need automation software stack multi-cloud. It takes not only the infrastructure, but the multi-cloud software layer above it. And then you have to deploy these microservices and containers and optimize the whole system with vector databases and memory, as you said before, and storage.
So we launched at GTC, a retrieval augmented generation solution with NVIDIA. Again, the first one jointly with NVIDIA to launch a full solution, that’s part of our AI factor. The third piece is what people are looking at next. Now that I’ve done training, I’ve done the first use cases, how do I take all my data everywhere and be able to use it for inferencing and training anywhere? And that is a hard challenge. This is more of a large data issue, you need indexing, you need parallel file processing, you need many components, and the whole industry is still early on in solving that.
Patrick Moorhead: Yeah. While I have you, a lot of questions I’m getting for enterprises are, GPU versus accelerator and InfiniBand versus Ethernet. And I know we could probably talk an entire hour or two, but there’s pros and cons to both approach, and I just very quickly wanted to get your take. If you were asked these questions by an enterprise, what would your answers be?
Ihab Tarazi: We developed a tool to determine what to use. This is much more complex than… In a traditional way this would’ve been like an offer or some kind of a differentiation document, but this is very complex because if it’s generative AI and it needs a lot of data, most likely it has to use GPU. If it’s general AI, it’s more about how many concurrent users, and it’s different if it’s video or text chat. Also, is this an edge application? Is it centralized? So we see CPU, GPU, accelerator, all correct. It depends on the size of the model, the use case, is it a quantized model? Are you doing the AG? Where are you’re doing it? How much data? What’s stage? And we have a very complex calculation for that.
Same with storage, how many checkpoints you need to save, the frequency of checkpoints. So we’ve automated all of that where the discussion, as Kyle was saying before, what model are you using, what outcome? And then we determine, and this is a big part of the help we do, is help people navigate between all of them. So in terms of Ethernet versus InfiniBand, InfiniBand clearly is the lead today. People are used to it, they used to it from HPC, it’s ease of use. However, Ethernet is taking off as the future, and we’re collaborating with PodCom as well as NVIDIA both, on next-generation Ethernet solutions.
The change in networking is two big things. Networking now starts inside the server, the GPU, you can’t just sell at switch, so therefore it puts Dell in a very competitive advantage to design it end to end. And the control is going to move to the NIC, over time. Number two, the software is not separate. You have to optimize your frameworks like Nickel or ICLE for the network topology. So you really have to work with the GPU partner to optimize the network. Again, that puts us in a very good position. It’s not possible to deliver performance if you only do the GPU or only do the network, and it’s tremendous change.
Daniel Newman: Yeah, it’s a really great way to sort of take us towards the end here. Pat and I love, on the show, to talk a little bit about what are the things the market is missing. No, I mean everyone gets compute, but it’s like the networking is really, really important. Again, you’re talking about the parallel computing that’s actually going on from GPU to GPU, and then of course you’ve got to talk system to system. And I like to use the analogy of Apple and Android. We’re going to have a Ethernet, you’re going to have InfiniBand, this is not going to be all one. This is not a winner takes all. Market loves a zero-sum game, it’s not…
And Dell is well positioned because you can do either, whatever, and it’s kind of nice to be able to be the catch-all. Look, we’ve got a minute or two left. DTW, I said it in the beginning. I don’t know if we’ll have an SAP center moment. But will we? Pat and I are hearing a lot of things, I’d love to see it, I heard Jensen might be on stage. What are you guys excited about? What do we want to know? And if you want, go ahead and give us the preview of all the announcements. Give us a little rundown before we send you onto your day.
Kyle Dufresne: I like my job a lot, so I’m going to Ihab tell you about all the announcements that are going to happen here at Dell World. I think the Jensen thing is public or semi-public, so I do think he’ll be there. And it was great, because he does see Dell as a key partner, you saw him at GTC being loud and proud at the keynote. I put Ihab right next to Michael, I sat next to him during the event. And no, I think you’re going to see a lot around AI. You’re going to see a lot about helping customers create more and more easy buttons on how to get started. And most importantly, how to get value. There’s been too often in our industry over many, many decades of people selling rainbows and puppy dogs, and we’re interested in selling very, very tightly managed ROIs on real use cases that are going to change these industries.
So we’re really trying to create as many easy buttons as customers join us in this journey, and I think you’re going to hear a lot about that. You’re going to see the entire floor, regardless if we’re talking about our PC business with all the NPUs that we’re adding to those for GenAI based PCs and workstations. People are running models now on workstations, it’s the most incredible thing we’ve seen, all the way through our biggest i and Core environments, working in data centers and not on the edge where a lot of this data is being created. So I think you’re going to see a lot about that there. And the show’s heavily going to be weighted towards AI, as you would expect.
Ihab Tarazi: And I think you’ll see us bring some of the partners we work with. You’re going to see a big leap forward with ecosystem partners, both on software and infrastructure, in solving AI for enterprises, especially on-prem. We’re going to make a big leap forward. You’re going to be very excited about the partners we’re bringing on stage and the announcements we’ll make. I personally feel it’s going to feel real to people, not just, here, it’s something in the future.
Kyle Dufresne: Right.
Daniel Newman: Kyle and Ihab, thank you both so much. We’re going to hold you to that. If it’s not awesome, it’s going to be a Friday morning conversation between Pat and I, but we expect that it will be. Appreciate you both joining us here on The Six Five On The Road. It’s the post GTC rundown, and we’re talking about Dell and the innovation that’s going on, not only at NVIDIA, but with Dell, with its partners, with its customers. Thanks everyone for tuning in. Hit that subscribe button. Join us for all our shows. Stick with Patrick and myself, but not on this episode, because we’re out of here.
Author Information
Daniel is the CEO of The Futurum Group. Living his life at the intersection of people and technology, Daniel works with the world’s largest technology brands exploring Digital Transformation and how it is influencing the enterprise.
From the leading edge of AI to global technology policy, Daniel makes the connections between business, people and tech that are required for companies to benefit most from their technology investments. Daniel is a top 5 globally ranked industry analyst and his ideas are regularly cited or shared in television appearances by CNBC, Bloomberg, Wall Street Journal and hundreds of other sites around the world.
A 7x Best-Selling Author including his most recent book “Human/Machine.” Daniel is also a Forbes and MarketWatch (Dow Jones) contributor.
An MBA and Former Graduate Adjunct Faculty, Daniel is an Austin Texas transplant after 40 years in Chicago. His speaking takes him around the world each year as he shares his vision of the role technology will play in our future.
