Dell Technologies World 2024 | Pierluca Chiodelli & Greg Findlen | Dell Technologies

Clips
News
More from Dell Technologies World 2024

Pierluca Chiodelli

Vice President Engineering Technology Edge Computing Offers, Strategy & Execution

Dell Technologies

Greg Findlen

Senior VP, Product Management

Dell Technologies

Three insights you might have missed from theCUBE’s coverage of Dell Technologies World

Dell Technologies Inc. devoted much of its annual conference in Las Vegas this month to defining its artificial intelligence strategy. Through announcements and various executive presentations, the company highlighted a set of networking and infrastructure solutions designed to capitalize on the technology tsunami that has become AI.What that picture looks like now and how it could propel Dell forward was a central focus for theCUBE, SiliconANGLE Media’s livestreaming studio, as Dell executives, partners and industry analysts offered their insights into how AI is shaping the enterprise computing future during Dell Technologies World 2024

Exploring Dell’s AI Factory and its positioning for edge-centric workloads

One of the exciting new facets of artificial intelligence’s evolution is its integration at the edge. By bringing AI closer to the point of enterprise data generation, companies stand to unlock speed, efficiency, agility and scalability gains.Live at DTW 2024: Analyzing the Dell AI Factory with Pierluca Chiodelli and Greg Findlen.Dell Technologies Inc.’s AI Factory combines infrastructure, solutions and services for organizations to rapidly adopt and deploy scalable AI

play_circle_outline Introduction to Dell Tech World and the guests (Savannah Peterson, John Furrier, Pierluca Chiodelli, and Greg Findlen)

play_circle_outline The AI Factory: Unveiling the Similarities to Traditional Factories and the Crucial Infrastructure Requirements

play_circle_outline Exploring the AI Factory: Different Perspectives, Ecosystem Partnerships, and Customer-centered Services

play_circle_outline Empowering AI Capabilities: Unveiling the Crucial Role of NativeEdge in Tackling Edge Challenges

play_circle_outline Use cases for AI at the Edge, such as manufacturing and healthcare

Info

Pierluca Chiodelli & Greg Findlen | Dell Technologies

Pierluca Chiodelli

Vice President Engineering Technology Edge Computing Offers, Strategy & Execution Dell Technologies

Greg Findlen

Senior VP, Product Management Dell Technologies

search

>> Good afternoon, AI fans, and welcome back to beautiful Las Vegas Nevada. We're here at the end of day three of Dell Tech World, and it has been an absolutely fantastic week. My name is Savannah Peterson, delighted to be joined by the one and only John Furrier. John, thanks for being here today.

>> Thanks for having me on this great program.

>> Thanks for creating it. So we all good. We're playing AI. We're talking AI, and I'm really excited to talk about AI on the Edge with our two guests joining us from Dell today. Pierluca Chiodelli, thank you so much for being here.

Pierluca Chiodelli

>> Thank you.

>> And also Greg Findlen. Thank you. First time on theCUBE.

Greg Findlen

>> Thank you for having me.

>> How's the show going for you so far, Greg?

Greg Findlen

>> It's going great so far. We're seeing a lot of interest in the Dell AI Factory. It's the beginning and ending of every conversation, so it's been a great week so far.

>> Have you talked about anything else this week?

Greg Findlen

>> No.

>> Well, we've been glowing with the... I've been a huge vocal proponent of the AI Factory positioning. Every call I've been on with the Delphos, definitely hitching the wagon to NVIDIA-

>> The messaging is-...

>> is a great call. It sends a statement. It means something different. So what is it? What is the AI Factory at the Edge?

Greg Findlen

>> Well, I'll talk about AI Factory more broadly, and then I think we can pivot to the Edge. When we talk about the AI Factory, we draw the analogy of how it's similar to how a factory was built before, where you're taking raw materials, you're turning them into a finished product that comes out the other side. Only now, we're using raw data from an enterprise customer's data set to generate new outcomes that they couldn't have generated before. Huge new set of use cases for our customers. Super exciting. So we start with that analogy, and then we start talking about how do we get from those use cases to what is it going to take to go build that out? We talk about the infrastructure layer that's going to be required to support it, whether it's high-speed servers or performance storage or Edge devices to deliver that outcome. The massive set of ecosystem partners, we continue to expand to support these use cases with some of the new announcements that we made this week. Most critically is, I think, the services that sit on top of it that help customers on that journey, and we spent a lot of time this week talking about the Dell AI Factory with NVIDIA, which is a really good example of a specific set of capabilities that we're bringing together under that umbrella.

>> At GTC, Jensen Huang, CEO of NVIDIA, showed the super server, I call it. It was essentially a monster rack with a huge spine, a bunch of leaves out and basically clustered servers, in my opinion. They look like a bunch of servers clustered together. That's the future. That is, to me, going to be the kind of systems we'll be looking at. Is that the poster child for the AI Factory, or is it something more specific? Or is it just a collection of servers? How should we scope in our mind what the AI Factory is?

Greg Findlen

>> So I think it's important to look at it from an enterprise perspective. It's not just the servers, it's the complete set of capabilities that we have to bring together to build one of these solutions. So, when we talk to customers about the AI Factory, it is servers, it's networking, it's storage, it's data protection, it's workstations, it's laptops. What are all the set of capabilities? When you look at a large enterprise, there's a lot of different workloads, and they're each going to have their different requirements. So that will scale from what we announced with the Microsoft Copilot offerings and the new laptops all the way up to these massive scale-out systems that are being built to support very large workloads for training and things like that.

>> Really, it's a complete ecosystem. It's a huge part of the conversation today, but it's not just one component. It's the whole shebang. Speaking of components and shebangs, let's talk about AI at the Edge. What have your conversations been like this week?

Pierluca Chiodelli

>> Yeah, no, he is absolutely right. We speak about factory AI. I think Michael said on stage, how you feed the beast. Right? And so, when you're thinking about that, AI at the Edge has been there for twenty-plus years. Now the AI at the Edge is not generative AI yet. It's more about models that they've been training for many, many years. Very precise. But the challenge is how, now that you have all this power, you can move this model and retrain the model. And so, there is a huge opportunity here, especially using NativeEdge because if I visit the manufacturing floor today, for example, and I have a solution that with camera that look at the operator, how he puts the cable and stuff like that. The model starts to drift. What is going to happen in that case, I have just a line. A serum goes up, the line stops, and the operator needs to decide this is a good image, not a good image. So, everything has stopped it, and I need to retrain that. How is happening today? It takes me two weeks. I need to send someone, call someone. He needs to come down with a USB, take the picture that they're wrong, go in another things, connect the things, and then, hopefully, he has a Factory AI. But in many cases, he doesn't have that, and he needs to do all of this again. So with NativeEdge and what we announced, obviously, with the name and NVIDIA is the ability to have a blueprint that can connect and transfer and create a unique CI/CD pipeline that connect Edge core and cloud. At that point, not only I can feed the beast, but I can also start to do generative AI. The generative AI that we're thinking at the Edge is not exactly what we think here. TheCUBE is its own AI, you just told us, but is also for the machinery inspection like that. Today, for example, if you're doing an inspection of a bridge, you focus only on the corrosion of the pipe or a very specific thing. Now, you need to go to a multi-model kind of things because you may have a branch tree, you have many other things. Normally, for retrain those model and make them more generative AI, you need to spend months. And so now we can connect to that Edge, we can bring the things. Last but not least, where the formation are important. Another concept that probably we spoke here more than 12 years ago is the data gravity. Where are the model important? The model are important, the data are important where you need to do something with that data. And guess what? At the Edge, if you have a manufacturing line, if you are in a store, the moment that you need to analyze and take action is there. So inferencing, it's absolutely key in that point.

>> The data gravity piece is interesting because, we want to come back to that because with data gravity, that's data. Now you've got metadata gravity, whole another level of... Because Edge, you need the metadata. You can't move all the data so the metadata is good to have on that.

Pierluca Chiodelli

>> By the way, that's introduced another thing that we will see in the years to come, right? When you want to keep the data where they are, first of all, you want to save only the data that you need and important. So you need to have the metadata, but the other things, you need to be able to federate so you don't move the data to another place, but you can use the power of the Edge to process all this data.

>> Let's put a little more color on NativeEdge. Because I think a NativeEdge. I think of edge-native like cloud-native. So that environment will be relevant obviously, and specifically the NVIDIA relationship-

Pierluca Chiodelli

>> Correct....

>> which announced a ten-to-one stock split this morning after their amazing earnings results. So, I mean, obviously doing well with NVIDIA. So the NVIDIA partnership with NativeEdge, and what does Edge native mean?

Pierluca Chiodelli

>> We didn't build a NativeEdge just for the stock, but obviously for solving the problem -

>> Your stock's doing good, too.

Pierluca Chiodelli

>> But the things that you see, right? What Greg's speaks about, the AI Factory, the AI Factory's still in the data center. Now when we go outside of the data center, of the wall of the data center is the land of no one. So you need to have the right hardware, but also there is no skill, so IT is not present there. And you need to support not one device, but thousand of devices. You need to have security, you need to have zero touch. NativeEdge addressed all of that at scale. So, allow us to go in place where there is no connectivity, but be able to bring whatever you need. With NVIDIA particularly, this is the first platform in the market that can do that. We can package this in the blueprint technology. If you remember we spoke about last year that we acquired a technology, we acquired a company, Cloudify, back in last year. We introduced that inside of NativeEdge. So not only NativeEdge allow you the zero trust and the security that you need for the manufacturing, the retail, but also allow you to transport all the packaging and all the receipt that you need for deploy AI, and this is very important. Because with the NVIDIA for example, and NIM and all these other AI things that we see, things change every day. So you need to have an ability to change only the component that you need. The security is another important thing. Someone calls you today. You are in a store, someone call you and say, "I have a problem with one of the package." It takes this to get to the package. You need to shut down the thing. With NativeEdge, you can just use the blueprint, redeploy the things and done.

Greg Findlen

>> It's an important tie back to the data center part of the AI Factory because you're going to be developing new models, you're going to be testing out new models, you might be fine-tuning it, but you need to be able to push it out at scale to your production sites, for example, or whatever your Edge location is for your business. So how these things combine together in the broader AI Factory is super important because there's a lot of things that need to happen in the core data center, and there's a lot that needs to happen out to the Edge and being able to coordinate those. When we bring these technology stacks together, it's super powerful for our customer.

>> Greg, we just interviewed one of your customers that did the HPC innovation labs, Northwestern Medicine. She was amazing. Their vision is the AI Factory in the location of the healthcare provider, the hospital. So they're doing X-rays. They do that 40% increase. They just had massive success. And by adding more GPUs, they can get multiple dimensions on the pictures. So CTs are next. So that's an example of the on-prem data value.

Greg Findlen

>> A hundred percent, and I think it really gets back to the use case piece here. Where it's very specific to each customer. So, Northwestern Medicine had a very specific problem that they were trying to solve, and increasing their productivity of their X-ray processing, and it worked spectacularly for them.

>> So, I have to ask you guys a question since you're here. On-premise is coming back. I mean, it's never really left, right? So okay, cloud had its day and still rocking and rolling. Amazon's doing great, Azure. But the on-prem debate, is it repatriation or net new? So maybe this is a repatriation for optimization spend or whatever, but AI is a whole other category. Jensen even said that as his keynote at NVIDIA GTC. This is a new category. So it implies net new spend or spend shifting. So how do you guys see that? Because the use case of the hospital place is a good one because it's a pain point that they can't get it done faster, but it's a net new capability, and their data's there. Not all enterprise data is in LLMs yet. So there's headroom for enterprises to launch their models.

Greg Findlen

>> For sure. There's so much enterprise data that's still sitting on-premise that is valuable. In many cases, it's the most critical data inside the enterprise. So, we do see this as a very new paradigm in data centers and kinds of workloads. So, there's a piece of this that is absolutely net new, and I think it'll be interesting to see, but when we look at it from a growth perspective, there's a lot of net new capability that's going to be required in the enterprise to support these workloads, and huge productivity gains and ROIs that will come from that investment as well.

>> Yeah, you guys are subscribers of theCUBE research, so you guys will get a copy of our report. We're going after that because we're trying to squint through the noise on the repatriation game, which we see some. But cloud's got a great advantage, Savannah. We talk about it all the time, but on-prem, the IP is in the workflows.

>> Mm-hmm.

>> The Edge is unexploited yet.

Pierluca Chiodelli

>> Edge is a great opportunity. The data are at the Edge. So the thing is, I listened timeline the interview before and also I was meeting with the hospital that say the same thing. They want to harvest from different MRI machine that they are not speaking together right now. So that's why they looking to use NativeEdge to be able to democratize this different ecosystem, and then they can collect all the data, and then they can put the AI capability to tag those data and do all the processes. So not only breaking the barrier that they have, but using what they already have. And all of us, we know, right? In our houses, in all our office, we have a ton of data. It's not that the data... Maybe we have a copy in other place, but most of the time we have 5, 6, 7, 8 different copies of the same thing and we make it tweak. So harvesting those data is very, very key. Also in Dell, we're doing that right now.

>> It's super critical. We've talked a lot about healthcare as an example. It's an outstanding example. Chris even told us, of all the data that healthcare facilities have, they only use 3% of that data, which is shocking when you think about the picture of holistic health. We could have if we had more of that. Can you give us some other examples outside of healthcare where you're hearing that customer craving for AI at the Edge in particular?

Pierluca Chiodelli

>> Yeah, no, absolutely. I think all the manufacturing, for example, that especially when you build a new manufacturing floor, is not anymore to just look at the things that happen every day, but is the ability to monitoring and describe an entire and connect all the machinery in the world for example. But the data, they don't leave. Right? For example, what I saw, a customer of Dell, that they have a control room that they produce machinery for the entire world. They can control all the machinery from there. They can also call the customer and say to him, "Hey, by the way, your plant in Dallas is going below the threshold of other system." And all of that is done with the AI, but without leaving the key data from the Edge. They actually not transport any data. By the way, the back end yes is on the cloud and all of that, but the data stay where they are. They have a capability to tag in and by the way, they use our fantastic infrastructures, do all of this. So it's very important.

>> Greg, from your perspective, are you looking at it from the AI Factory, from a data or infrastructure perspective or both? Because we've been saying on theCUBE that the action right now is infrastructure. Okay, chips, things are getting faster. But the data layer, the middle layer that's going to be disrupted, enabling... Disrupted in a good way. What's your lens? Where do you dig in for Dell?

Greg Findlen

>> So I look at it sort of holistically. From my perspective, I always start back at the use case. At the end of the day, this is about generating new use cases for customers. In order to go do that, they need the infrastructure for sure. They need the ecosystem of capabilities through software partners. They need the services. But at the end of the day, it comes back to the data. When we talk to customers about prioritizing use cases, one of the most common things that we highlight as they go and prioritize is always start your use case prioritization on where do you have good data? Because if you don't have good data, it's going to dramatically slow down your AI transformation efforts because you're going to spend a lot of time getting there. So there's a lot of opportunity there, especially as you look forward from traditional structured kinds of data management to the kinds that are going to be needed for how do you manage document and ingest into these LLM systems? So it's an interesting new space that's going to continue to get a lot of focus.

>> You just factor data into the system architecture. All right, now getting back, I guess, more product-oriented, in my mind, a server and a PC were easy to understand. I got a PC, great XPS right here, server PowerEdge. I rack and stack them on top of rack switch. I feel good. That's IT. Now, in AI Factory, it's like you're the general contractor. You're building a lot of servers, a lot of networking, there's a lot of interconnects. We're seeing that at the chip level and then also at the system level. So what are customers buying from Dell? Is it a product? Is it a product number? Are you like a GC on a construction job? Are you building the factory? Take us through. I'm trying to visualize what the factory is. Is it just a collection of servers that drop into the hospital, or is it much more like bigger like the NVIDIA system that we saw, or a collection of little Edges. They're all connected with distributed computing. I get that, but what are you guys selling?

Greg Findlen

>> So, when we talk about the AI Factory, there's actually three different ways to consume it. You can consume it as a solution where we package everything together. Customers are maybe a little bit more advanced, might consume it as components that we sell to them. And some set of customers are going to be looking for it on a subscription basis or other consumption basis. But when we look at the customers who are looking at it on a solution basis, what we're trying to do is bring that all together for the customer. So it's a packaged set of server storage, east-west, networking, data protection, the software stack that goes above it in consulting services that go along with it to help customers around their journey get to the outcome they're going for. It's not a packaged product in the way that we would think of a VX block in years past, but it is a system specific set of configurations that sort of pre-engineered at a solution system level to get to customers for that outcome.

>> So it's got a spine and leaves, got all the networking in there, it's tied together, has multiple servers like clusters. One-

Greg Findlen

>> The whole system design, we're looking at cooling, we're looking at how do we rack and stack it, we're looking at the cabling, all the networking that's required so that we can deliver that as an integrated system.

>> We were talking earlier, Savannah and I, about the cooling pieces. Really relevant. Varun was on earlier, and we saw Jensen on stage. He's got a love in the cooling. You guys got that cooling product is pretty differentiating for Dell. It's been a big part of the factory.

Greg Findlen

>> Absolutely. Yeah. The XE9680 XL is our latest form factor. It gets us from a 6U form factor with air cool down to 4U, can handle the B200. So now we can hit in a single rack configuration with liquid cooling. Moves us along quite a bit towards-

>> Very impressive.

>> Way to bring the speeds and feeds to theCUBE.

>> That is exciting.

>> We love our speeds and feeds.

>> I know. No, I love it.

>> You will definitely be invited back, that's for sure.

>> Yeah.

>> No, we'd love the speeds and feeds. This is what's key right now. We are in a build-out mode. This is a product market. If you look at the changes, some of the other companies like Amazon Web Services, the leadership changes are happening. Product-led managers are on the front lines right now because it really is a net new opportunity, and it's build-out mode. So, products are critical, speeds and feeds, performance specking. We seeing a systems revolution going on.

Pierluca Chiodelli

>> Yeah, and not only that, at the Edge you see that the new emerging products, they also come in, they are needed for low power. If you see the T160 that we announced here, that is an enterprise tower that can be at the Edge. But more and more, you're going to see in the future all the, sorry, inferencing will surpass training. So you need a new set of power in the Edge that will have used GPU, but also a die kind of things.

>> We have, I and Savannah, have been talking on theCUBE many times since Supercomputing two years ago. Ethernet. I'm a pro Ethernet person, just for the record. I'm pro Ethernet, just saying that. I love the other one, InfiniBand, but that's a whole different discussion. Ethernet will win. So Ethernet will be the networking.

Pierluca Chiodelli

>> If you say so, yes.

>> Of course. No InfiniBand has a use case. They have different use cases.

Pierluca Chiodelli

>> Yes.

>> You can't really compare. And the monster configuration with NVIDIA, InfiniBand makes sense, but interconnect. But in the general at large, at the Edge, it's Ethernet.

Pierluca Chiodelli

>> Yes, absolutely. People did not even switch to 10 gig at the Edge. You need to think that you need to have a 5G capability, private 5G, LTE. You need to also switch from one band to the other. If you as, for example, welding machine that create interference with the thing. But also there is a new war there, right? It's like it's just in infancy to be transformed, right? So we will be, in the future, speaking about which factory we're going to build.

>> Actually, you guys are a great duo here because the factory and then the Edge together comes back to stuff that we've been covering, Savannah, client server kind of conversations. Device at Mobile World Congress or CES. Consumer devices, whether they're wearables or IOT devices on street poles or whatever are now connected into the network. So, you now have that neural network and full network. So the question is, what are we going to be seeing? Every show's turning into an AI show. Mobile World Congress is turning into AI show. What are we going to hear at Mobile World Congress? Because if this continues to go down, it's only a couple more months away in Barcelona, about a half a year.

Pierluca Chiodelli

>> Half a year.

>> But that's enough time to get stuff baked out. You're going to see more AI factories going on. So Telco, that's an on-premise. What are we going to be hearing at the Mobile World Congresses of the world as the devices start getting smarter?

Pierluca Chiodelli

>> I think what we going to hear-

>> We'll be there, by the way, at the queue.

Pierluca Chiodelli

>> Yeah. Yes. Maybe we're doing another interview there.

>> Yeah, of course.

Pierluca Chiodelli

>> I don't have a crystal ball, but I think what already seen and speaking with different Telco, there is definitely a huge interest in entering the enterprise space with a solution and bringing the AI to the Edge. And now Telco also defined the Edge in different way, but now it's starting to go to the functional Edge and the -

>> Far Edge.

Pierluca Chiodelli

>> Yes, and that's where we are playing. Now we have a great partnership that we announced with Nokia... Sorry, with... I don't remember right now.

>> Ericsson.

Pierluca Chiodelli

>> No. We announced a partnership where we can really bring our, for example, gateway, having public 5G for example, we already support LTE. All of that will be orchestrated by NativeEdge. So you can not only deploy the solution, but also, you can now connect the functionality to the network. I think that's very important because you are in some cases out in the nowhere. Right? There is no ethernet cable.

>> Right. Right. Well, we look forward to continuing the conversation then at MWC. This has been fantastic. Greg, Pierluca, thank you you so much for being here as always. And John, fantastic discussion as usual.

>> We love the Edge.

>> Yeah, we do. Living on the Edge, hopefully in a-

Pierluca Chiodelli

>> Far on the Edge.

>> Yeah.

Pierluca Chiodelli

>> There's more Edge -

>> We have a theater center at the main studio. That's our core Edge.

Pierluca Chiodelli

>> Thank you very much.

>> Yes, thank you. Thank you.

Pierluca Chiodelli

>> Fabulous.

>> Yes, fabulous. And thank all of you for tuning into our three days of live coverage here at Dell Tech World in Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.

Pierluca Chiodelli & Greg Findlen | Dell Technologies

search

>> Good afternoon, AI fans, and welcome back to beautiful Las Vegas Nevada. We're here at the end of day three of Dell Tech World, and it has been an absolutely fantastic week. My name is Savannah Peterson, delighted to be joined by the one and only John Furrier. John, thanks for being here today.

>> Thanks for having me on this great program.

>> Thanks for creating it. So we all good. We're playing AI. We're talking AI, and I'm really excited to talk about AI on the Edge with our two guests joining us from Dell today. Pierluca Chiodelli, thank you so much for being here.

Pierluca Chiodelli

>> Thank you.

>> And also Greg Findlen. Thank you. First time on theCUBE.

Greg Findlen

>> Thank you for having me.

>> How's the show going for you so far, Greg?

Greg Findlen

>> It's going great so far. We're seeing a lot of interest in the Dell AI Factory. It's the beginning and ending of every conversation, so it's been a great week so far.

>> Have you talked about anything else this week?

Greg Findlen

>> No.

>> Well, we've been glowing with the... I've been a huge vocal proponent of the AI Factory positioning. Every call I've been on with the Delphos, definitely hitching the wagon to NVIDIA-

>> The messaging is-...

>> is a great call. It sends a statement. It means something different. So what is it? What is the AI Factory at the Edge?

Greg Findlen

>> Well, I'll talk about AI Factory more broadly, and then I think we can pivot to the Edge. When we talk about the AI Factory, we draw the analogy of how it's similar to how a factory was built before, where you're taking raw materials, you're turning them into a finished product that comes out the other side. Only now, we're using raw data from an enterprise customer's data set to generate new outcomes that they couldn't have generated before. Huge new set of use cases for our customers. Super exciting. So we start with that analogy, and then we start talking about how do we get from those use cases to what is it going to take to go build that out? We talk about the infrastructure layer that's going to be required to support it, whether it's high-speed servers or performance storage or Edge devices to deliver that outcome. The massive set of ecosystem partners, we continue to expand to support these use cases with some of the new announcements that we made this week. Most critically is, I think, the services that sit on top of it that help customers on that journey, and we spent a lot of time this week talking about the Dell AI Factory with NVIDIA, which is a really good example of a specific set of capabilities that we're bringing together under that umbrella.

>> At GTC, Jensen Huang, CEO of NVIDIA, showed the super server, I call it. It was essentially a monster rack with a huge spine, a bunch of leaves out and basically clustered servers, in my opinion. They look like a bunch of servers clustered together. That's the future. That is, to me, going to be the kind of systems we'll be looking at. Is that the poster child for the AI Factory, or is it something more specific? Or is it just a collection of servers? How should we scope in our mind what the AI Factory is?

Greg Findlen

>> So I think it's important to look at it from an enterprise perspective. It's not just the servers, it's the complete set of capabilities that we have to bring together to build one of these solutions. So, when we talk to customers about the AI Factory, it is servers, it's networking, it's storage, it's data protection, it's workstations, it's laptops. What are all the set of capabilities? When you look at a large enterprise, there's a lot of different workloads, and they're each going to have their different requirements. So that will scale from what we announced with the Microsoft Copilot offerings and the new laptops all the way up to these massive scale-out systems that are being built to support very large workloads for training and things like that.

>> Really, it's a complete ecosystem. It's a huge part of the conversation today, but it's not just one component. It's the whole shebang. Speaking of components and shebangs, let's talk about AI at the Edge. What have your conversations been like this week?

Pierluca Chiodelli

>> Yeah, no, he is absolutely right. We speak about factory AI. I think Michael said on stage, how you feed the beast. Right? And so, when you're thinking about that, AI at the Edge has been there for twenty-plus years. Now the AI at the Edge is not generative AI yet. It's more about models that they've been training for many, many years. Very precise. But the challenge is how, now that you have all this power, you can move this model and retrain the model. And so, there is a huge opportunity here, especially using NativeEdge because if I visit the manufacturing floor today, for example, and I have a solution that with camera that look at the operator, how he puts the cable and stuff like that. The model starts to drift. What is going to happen in that case, I have just a line. A serum goes up, the line stops, and the operator needs to decide this is a good image, not a good image. So, everything has stopped it, and I need to retrain that. How is happening today? It takes me two weeks. I need to send someone, call someone. He needs to come down with a USB, take the picture that they're wrong, go in another things, connect the things, and then, hopefully, he has a Factory AI. But in many cases, he doesn't have that, and he needs to do all of this again. So with NativeEdge and what we announced, obviously, with the name and NVIDIA is the ability to have a blueprint that can connect and transfer and create a unique CI/CD pipeline that connect Edge core and cloud. At that point, not only I can feed the beast, but I can also start to do generative AI. The generative AI that we're thinking at the Edge is not exactly what we think here. TheCUBE is its own AI, you just told us, but is also for the machinery inspection like that. Today, for example, if you're doing an inspection of a bridge, you focus only on the corrosion of the pipe or a very specific thing. Now, you need to go to a multi-model kind of things because you may have a branch tree, you have many other things. Normally, for retrain those model and make them more generative AI, you need to spend months. And so now we can connect to that Edge, we can bring the things. Last but not least, where the formation are important. Another concept that probably we spoke here more than 12 years ago is the data gravity. Where are the model important? The model are important, the data are important where you need to do something with that data. And guess what? At the Edge, if you have a manufacturing line, if you are in a store, the moment that you need to analyze and take action is there. So inferencing, it's absolutely key in that point.

>> The data gravity piece is interesting because, we want to come back to that because with data gravity, that's data. Now you've got metadata gravity, whole another level of... Because Edge, you need the metadata. You can't move all the data so the metadata is good to have on that.

Pierluca Chiodelli

>> By the way, that's introduced another thing that we will see in the years to come, right? When you want to keep the data where they are, first of all, you want to save only the data that you need and important. So you need to have the metadata, but the other things, you need to be able to federate so you don't move the data to another place, but you can use the power of the Edge to process all this data.

>> Let's put a little more color on NativeEdge. Because I think a NativeEdge. I think of edge-native like cloud-native. So that environment will be relevant obviously, and specifically the NVIDIA relationship-

Pierluca Chiodelli

>> Correct....

>> which announced a ten-to-one stock split this morning after their amazing earnings results. So, I mean, obviously doing well with NVIDIA. So the NVIDIA partnership with NativeEdge, and what does Edge native mean?

Pierluca Chiodelli

>> We didn't build a NativeEdge just for the stock, but obviously for solving the problem -

>> Your stock's doing good, too.

Pierluca Chiodelli

>> But the things that you see, right? What Greg's speaks about, the AI Factory, the AI Factory's still in the data center. Now when we go outside of the data center, of the wall of the data center is the land of no one. So you need to have the right hardware, but also there is no skill, so IT is not present there. And you need to support not one device, but thousand of devices. You need to have security, you need to have zero touch. NativeEdge addressed all of that at scale. So, allow us to go in place where there is no connectivity, but be able to bring whatever you need. With NVIDIA particularly, this is the first platform in the market that can do that. We can package this in the blueprint technology. If you remember we spoke about last year that we acquired a technology, we acquired a company, Cloudify, back in last year. We introduced that inside of NativeEdge. So not only NativeEdge allow you the zero trust and the security that you need for the manufacturing, the retail, but also allow you to transport all the packaging and all the receipt that you need for deploy AI, and this is very important. Because with the NVIDIA for example, and NIM and all these other AI things that we see, things change every day. So you need to have an ability to change only the component that you need. The security is another important thing. Someone calls you today. You are in a store, someone call you and say, "I have a problem with one of the package." It takes this to get to the package. You need to shut down the thing. With NativeEdge, you can just use the blueprint, redeploy the things and done.

Greg Findlen

>> It's an important tie back to the data center part of the AI Factory because you're going to be developing new models, you're going to be testing out new models, you might be fine-tuning it, but you need to be able to push it out at scale to your production sites, for example, or whatever your Edge location is for your business. So how these things combine together in the broader AI Factory is super important because there's a lot of things that need to happen in the core data center, and there's a lot that needs to happen out to the Edge and being able to coordinate those. When we bring these technology stacks together, it's super powerful for our customer.

>> Greg, we just interviewed one of your customers that did the HPC innovation labs, Northwestern Medicine. She was amazing. Their vision is the AI Factory in the location of the healthcare provider, the hospital. So they're doing X-rays. They do that 40% increase. They just had massive success. And by adding more GPUs, they can get multiple dimensions on the pictures. So CTs are next. So that's an example of the on-prem data value.

Greg Findlen

>> A hundred percent, and I think it really gets back to the use case piece here. Where it's very specific to each customer. So, Northwestern Medicine had a very specific problem that they were trying to solve, and increasing their productivity of their X-ray processing, and it worked spectacularly for them.

>> So, I have to ask you guys a question since you're here. On-premise is coming back. I mean, it's never really left, right? So okay, cloud had its day and still rocking and rolling. Amazon's doing great, Azure. But the on-prem debate, is it repatriation or net new? So maybe this is a repatriation for optimization spend or whatever, but AI is a whole other category. Jensen even said that as his keynote at NVIDIA GTC. This is a new category. So it implies net new spend or spend shifting. So how do you guys see that? Because the use case of the hospital place is a good one because it's a pain point that they can't get it done faster, but it's a net new capability, and their data's there. Not all enterprise data is in LLMs yet. So there's headroom for enterprises to launch their models.

Greg Findlen

>> For sure. There's so much enterprise data that's still sitting on-premise that is valuable. In many cases, it's the most critical data inside the enterprise. So, we do see this as a very new paradigm in data centers and kinds of workloads. So, there's a piece of this that is absolutely net new, and I think it'll be interesting to see, but when we look at it from a growth perspective, there's a lot of net new capability that's going to be required in the enterprise to support these workloads, and huge productivity gains and ROIs that will come from that investment as well.

>> Yeah, you guys are subscribers of theCUBE research, so you guys will get a copy of our report. We're going after that because we're trying to squint through the noise on the repatriation game, which we see some. But cloud's got a great advantage, Savannah. We talk about it all the time, but on-prem, the IP is in the workflows.

>> Mm-hmm.

>> The Edge is unexploited yet.

Pierluca Chiodelli

>> Edge is a great opportunity. The data are at the Edge. So the thing is, I listened timeline the interview before and also I was meeting with the hospital that say the same thing. They want to harvest from different MRI machine that they are not speaking together right now. So that's why they looking to use NativeEdge to be able to democratize this different ecosystem, and then they can collect all the data, and then they can put the AI capability to tag those data and do all the processes. So not only breaking the barrier that they have, but using what they already have. And all of us, we know, right? In our houses, in all our office, we have a ton of data. It's not that the data... Maybe we have a copy in other place, but most of the time we have 5, 6, 7, 8 different copies of the same thing and we make it tweak. So harvesting those data is very, very key. Also in Dell, we're doing that right now.

>> It's super critical. We've talked a lot about healthcare as an example. It's an outstanding example. Chris even told us, of all the data that healthcare facilities have, they only use 3% of that data, which is shocking when you think about the picture of holistic health. We could have if we had more of that. Can you give us some other examples outside of healthcare where you're hearing that customer craving for AI at the Edge in particular?

Pierluca Chiodelli

>> Yeah, no, absolutely. I think all the manufacturing, for example, that especially when you build a new manufacturing floor, is not anymore to just look at the things that happen every day, but is the ability to monitoring and describe an entire and connect all the machinery in the world for example. But the data, they don't leave. Right? For example, what I saw, a customer of Dell, that they have a control room that they produce machinery for the entire world. They can control all the machinery from there. They can also call the customer and say to him, "Hey, by the way, your plant in Dallas is going below the threshold of other system." And all of that is done with the AI, but without leaving the key data from the Edge. They actually not transport any data. By the way, the back end yes is on the cloud and all of that, but the data stay where they are. They have a capability to tag in and by the way, they use our fantastic infrastructures, do all of this. So it's very important.

>> Greg, from your perspective, are you looking at it from the AI Factory, from a data or infrastructure perspective or both? Because we've been saying on theCUBE that the action right now is infrastructure. Okay, chips, things are getting faster. But the data layer, the middle layer that's going to be disrupted, enabling... Disrupted in a good way. What's your lens? Where do you dig in for Dell?

Greg Findlen

>> So I look at it sort of holistically. From my perspective, I always start back at the use case. At the end of the day, this is about generating new use cases for customers. In order to go do that, they need the infrastructure for sure. They need the ecosystem of capabilities through software partners. They need the services. But at the end of the day, it comes back to the data. When we talk to customers about prioritizing use cases, one of the most common things that we highlight as they go and prioritize is always start your use case prioritization on where do you have good data? Because if you don't have good data, it's going to dramatically slow down your AI transformation efforts because you're going to spend a lot of time getting there. So there's a lot of opportunity there, especially as you look forward from traditional structured kinds of data management to the kinds that are going to be needed for how do you manage document and ingest into these LLM systems? So it's an interesting new space that's going to continue to get a lot of focus.

>> You just factor data into the system architecture. All right, now getting back, I guess, more product-oriented, in my mind, a server and a PC were easy to understand. I got a PC, great XPS right here, server PowerEdge. I rack and stack them on top of rack switch. I feel good. That's IT. Now, in AI Factory, it's like you're the general contractor. You're building a lot of servers, a lot of networking, there's a lot of interconnects. We're seeing that at the chip level and then also at the system level. So what are customers buying from Dell? Is it a product? Is it a product number? Are you like a GC on a construction job? Are you building the factory? Take us through. I'm trying to visualize what the factory is. Is it just a collection of servers that drop into the hospital, or is it much more like bigger like the NVIDIA system that we saw, or a collection of little Edges. They're all connected with distributed computing. I get that, but what are you guys selling?

Greg Findlen

>> So, when we talk about the AI Factory, there's actually three different ways to consume it. You can consume it as a solution where we package everything together. Customers are maybe a little bit more advanced, might consume it as components that we sell to them. And some set of customers are going to be looking for it on a subscription basis or other consumption basis. But when we look at the customers who are looking at it on a solution basis, what we're trying to do is bring that all together for the customer. So it's a packaged set of server storage, east-west, networking, data protection, the software stack that goes above it in consulting services that go along with it to help customers around their journey get to the outcome they're going for. It's not a packaged product in the way that we would think of a VX block in years past, but it is a system specific set of configurations that sort of pre-engineered at a solution system level to get to customers for that outcome.

>> So it's got a spine and leaves, got all the networking in there, it's tied together, has multiple servers like clusters. One-

Greg Findlen

>> The whole system design, we're looking at cooling, we're looking at how do we rack and stack it, we're looking at the cabling, all the networking that's required so that we can deliver that as an integrated system.

>> We were talking earlier, Savannah and I, about the cooling pieces. Really relevant. Varun was on earlier, and we saw Jensen on stage. He's got a love in the cooling. You guys got that cooling product is pretty differentiating for Dell. It's been a big part of the factory.

Greg Findlen

>> Absolutely. Yeah. The XE9680 XL is our latest form factor. It gets us from a 6U form factor with air cool down to 4U, can handle the B200. So now we can hit in a single rack configuration with liquid cooling. Moves us along quite a bit towards-

>> Very impressive.

>> Way to bring the speeds and feeds to theCUBE.

>> That is exciting.

>> We love our speeds and feeds.

>> I know. No, I love it.

>> You will definitely be invited back, that's for sure.

>> Yeah.

>> No, we'd love the speeds and feeds. This is what's key right now. We are in a build-out mode. This is a product market. If you look at the changes, some of the other companies like Amazon Web Services, the leadership changes are happening. Product-led managers are on the front lines right now because it really is a net new opportunity, and it's build-out mode. So, products are critical, speeds and feeds, performance specking. We seeing a systems revolution going on.

Pierluca Chiodelli

>> Yeah, and not only that, at the Edge you see that the new emerging products, they also come in, they are needed for low power. If you see the T160 that we announced here, that is an enterprise tower that can be at the Edge. But more and more, you're going to see in the future all the, sorry, inferencing will surpass training. So you need a new set of power in the Edge that will have used GPU, but also a die kind of things.

>> We have, I and Savannah, have been talking on theCUBE many times since Supercomputing two years ago. Ethernet. I'm a pro Ethernet person, just for the record. I'm pro Ethernet, just saying that. I love the other one, InfiniBand, but that's a whole different discussion. Ethernet will win. So Ethernet will be the networking.

Pierluca Chiodelli

>> If you say so, yes.

>> Of course. No InfiniBand has a use case. They have different use cases.

Pierluca Chiodelli

>> Yes.

>> You can't really compare. And the monster configuration with NVIDIA, InfiniBand makes sense, but interconnect. But in the general at large, at the Edge, it's Ethernet.

Pierluca Chiodelli

>> Yes, absolutely. People did not even switch to 10 gig at the Edge. You need to think that you need to have a 5G capability, private 5G, LTE. You need to also switch from one band to the other. If you as, for example, welding machine that create interference with the thing. But also there is a new war there, right? It's like it's just in infancy to be transformed, right? So we will be, in the future, speaking about which factory we're going to build.

>> Actually, you guys are a great duo here because the factory and then the Edge together comes back to stuff that we've been covering, Savannah, client server kind of conversations. Device at Mobile World Congress or CES. Consumer devices, whether they're wearables or IOT devices on street poles or whatever are now connected into the network. So, you now have that neural network and full network. So the question is, what are we going to be seeing? Every show's turning into an AI show. Mobile World Congress is turning into AI show. What are we going to hear at Mobile World Congress? Because if this continues to go down, it's only a couple more months away in Barcelona, about a half a year.

Pierluca Chiodelli

>> Half a year.

>> But that's enough time to get stuff baked out. You're going to see more AI factories going on. So Telco, that's an on-premise. What are we going to be hearing at the Mobile World Congresses of the world as the devices start getting smarter?

Pierluca Chiodelli

>> I think what we going to hear-

>> We'll be there, by the way, at the queue.

Pierluca Chiodelli

>> Yeah. Yes. Maybe we're doing another interview there.

>> Yeah, of course.

Pierluca Chiodelli

>> I don't have a crystal ball, but I think what already seen and speaking with different Telco, there is definitely a huge interest in entering the enterprise space with a solution and bringing the AI to the Edge. And now Telco also defined the Edge in different way, but now it's starting to go to the functional Edge and the -

>> Far Edge.

Pierluca Chiodelli

>> Yes, and that's where we are playing. Now we have a great partnership that we announced with Nokia... Sorry, with... I don't remember right now.

>> Ericsson.

Pierluca Chiodelli

>> No. We announced a partnership where we can really bring our, for example, gateway, having public 5G for example, we already support LTE. All of that will be orchestrated by NativeEdge. So you can not only deploy the solution, but also, you can now connect the functionality to the network. I think that's very important because you are in some cases out in the nowhere. Right? There is no ethernet cable.

>> Right. Right. Well, we look forward to continuing the conversation then at MWC. This has been fantastic. Greg, Pierluca, thank you you so much for being here as always. And John, fantastic discussion as usual.

>> We love the Edge.

>> Yeah, we do. Living on the Edge, hopefully in a-

Pierluca Chiodelli

>> Far on the Edge.

>> Yeah.

Pierluca Chiodelli

>> There's more Edge -

>> We have a theater center at the main studio. That's our core Edge.

Pierluca Chiodelli

>> Thank you very much.

>> Yes, thank you. Thank you.

Pierluca Chiodelli

>> Fabulous.

>> Yes, fabulous. And thank all of you for tuning into our three days of live coverage here at Dell Tech World in Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.