Dell Technologies World 2024 | Travis Vigil & Ihab Tarazi | Dell Technologies

Clips
News
More from Dell Technologies World 2024

Travis Vigil

SVP, ISG Product Management

Dell Technologies

Ihab Tarazi

SVP/Chief Technology Officer

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

Dell’s vision for the AI era: Integrating networking, compute and storage

As the artificial intelligence symphony nears its crescendo, it’s clear that companies must effectively combine storage, networking and compute to be effective, efficient and scalable.Discussing Dell’s recent AI-focused announcements with the company’s Travis Vigil and Ihab Tarazi.Dell Technologies Inc. is answering that demand with a comprehensive lineup of announcements, including its PowerStore Prime and PowerFlex storage offerings, as well as the XE9680 server class

play_circle_outline Focus on AI edition, new systems, storage, networking announcements

play_circle_outline Dell's Networking Strategy: Achieving AI Era Compatibility through Collaborative Partnerships with Broadcom and NVIDIA

play_circle_outline Dell's Diverse AI Services: Empowering Enterprises with Generative Solutions for Customers of all Sizes

play_circle_outline Quantum computing's promise and its potential use in specific workloads and sensing applications

Info

Travis Vigil & Ihab Tarazi | Dell Technologies

Travis Vigil

SVP, ISG Product Management Dell Technologies

Ihab Tarazi

SVP/Chief Technology Officer Dell Technologies

search

>> Good evening Dell fans and welcome back to the desert. We're here in fabulous Las Vegas, Nevada wrapping up our second day of live coverage of three days of Dell Tech World. It has been spicy, it has been fun, and it has been very interesting. My name's Savannah Peterson, joined by my fabulous co-host, Dave Vellante. Dave, what an evening. My gosh.

>> I know, two evenings in a row.

>> I know. We are the nightly news for enterprise tech now.

>> We haven't been outside in days.

>> Is there an outside world? I'm not even sure. Very excited, particularly for this point in the day to have two Dell superstars and also two very multiple-time CUBE alumni alums.

>> Yes.

>> Yeah. Travis Viejo, thank you so much for being here.

>> Thank you for having me.

>> And Ehab Terozi, it's always a pleasure to see your smiling face.

>> Thank you.

>> Love it. You made it right at the buzzer tonight just to grace us with your fabulous presence. Big week for Dell. Huge week for Dell. How's it been for you, Travis? What's the week been like?

>> It's been a whirlwind.

>> I bet.

>> It's the AI edition of Dell Tech World.

>> Certainly is.

>> And the announcements around the AI factory, the new systems, the new storage, the new networking have been fantastic and worthy of a Dell world all by themselves. But in addition to that, we have PowerStore Prime, the biggest release for PowerStore in the last four years. We had some PowerFlex announcements. It's just been chock-full of innovation and really a culmination of the last year of focusing on generative AI first, private cloud, multi-cloud second, and just the ISG organization is firing on all cylinders when it comes to innovation.

>> Ehab, we've talked about Dell's networking strategy and you're kind of changing the game there. Dell's never been a leader in networking. I'm sorry, don't hate me for saying that, but I'm sure you've thought about that and said, okay, how do we change the equation? And that's really what you're doing. Can you explain your strategy and why it fits in the AI era?

>> Yeah. I would say first of all, Travis and I have been working the network strategy for a long time in Dell. What we bet was on the future and AI validated all the decisions we made. What we bet on is that networking needs to be integrated with compute and storage and it needs to be very high throughput. And also, we also bet on open source and next generation software. So we are big partners of Broadcom and we're also working now closely with NVIDIA. The networking strategy we pick to get high performance switches with significant throughput and fully integrate the stack with compute is exactly what AI needs. What we showed today is how our strategy is just accelerating and scaling new components from Broadcom, Thor 2, Thor 3, co-packaged optics, Tomahawk 5 specifically that we launched a switch on is all part of that strategy. And if you saw, we also dialed in on NVIDIA with Spectrum X, Spectrum 4, InfiniBand Switching and the BlueField 3 DPUs. So it's really interesting to me because now with AI being as big as it is and how big of a use case that is and how AI acquires that integrated model, this not only helps us with AI, but helps us with enterprise because that is the future.

>> Well, it's interesting because networking's never been a centerpiece, but now it is, and it's not like as Michael was talking about, we are going to disrupt ourselves or be out of business in five years. So this is a nice target to disrupt. You can go hard after this. How do you think it's going to change the equation in the market dynamic?

>> I think generative AI offers a whole new landscape for networking. And if you look at the issues that folks are trying to solve as they implement small, medium, or large scale or generative AI deployments, the GPU to GPU communication is a whole new ball game. And we really have a strategy that's focused on providing a streamlined solution for that use case with a high degree of automation. And it's a great compliment to our server and storage portfolio to be able to package up compute, networking, and storage for not just a net new use case for networking, but a net new use case in general where customers are really looking for that integration and for that automation.

>> And also now networking extends to the NIC and inside the server. So that's a big change to the market. Selling a switch by itself is not really easy anymore. You really have to provide a system. And then you add to that the fact that AI is expanding at a massive speed, you have to jump to 800 gig next and do this next. So it's a completely different market dynamic. I'll tell you a story. I talked to a customer today about networking and they said, "Oh, we have a really good solution from an additional vendor." I'm not going to say who. And I said, "What's your throughput? What is your GPU to GPU throughput?" And we said, "Well, we really don't know." I said, "Most likely your throughput, you're given away half of the utilization of the GPUs that already scarce and expensive." And that kind of a change in metric is going to have a huge impact in the market.

>> And we were talking at MWC that there aren't a lot of companies that can do both the sort of internal communication across the XPUs and the external switching. And you guys are consumers of both of the leading companies that do that. NVIDIA and Broadcom, you've got great relationships with both of them. These are really hard technologies to get right.

>> They're really hard technologies. And I mean, you're bringing up a great point. It's all about this ecosystem. We've heard it on stage, open ecosystem. We're talking about open source. And it's not just these one-off, to your point, a switch or a component or one single player, it could be irrelevant six months from now. What's going to matter is these ecosystems playing together to provide inclusive solutions to your customers. And I mean, I got to give you guys a shout-out. Last week, Morgan Stanley saying that you are the AI stack of the enterprise future. It's a pretty bold claim. You're obviously seeing the stock smile as a result of that deservedly so, which is exciting. But I thought that was really interesting timing given we're leading into your week. You have all these announcements going on, and your keynote yesterday is literally all of the AI celebrities on the planet hanging out and patting each other on the back, and then we get to hang out with y'all. So I'm curious, you guys have been anticipating this moment for a while. How long have you been working on your collaboration, making sure that everything went together with all these partners to get us to this party that we're at today basically?

>> I think back to a year ago, and I think we were just entering in sort of this inflection. ChatGPT was just starting to enter everybody's consciousness and we announced Project Helix.

>> We talked about it a lot this week.

>> And at the same time, you go back two years from there. And that was the point in time where we made the decision to create the XE-9680. So it was two years prior to a year. So three years ago where we made that investment and we could see a lot of pieces coming together. We could see that the GPUs were going to be more critical going forward. But I don't think I'd be telling the truth if I said we knew exactly how big it was going to get over this last year.

>> That's kind of what I was wondering.

>> Right. Yeah. And I think Ehab can attest from last year to this year, it has been just very agile development, constant focus on providing solutions, an intense amount of focus on large-scale training. We think that the inferencing boom and the enterprise boom of generative AI is going to follow. We were thinking that we're on the front end of that, but I think it was good foresight combined with good luck. And this last year, I think we've really been agile in addressing the opportunities, quick in addressing the opportunities, introducing new technology at a rate that we have to have to keep up with the pace of the industry.

>> That's my follow-up question there. How do you keep up? I mean, you're not only playing ball, you're leading the game and helping everybody stay in the arena.

>> Yeah, I think Travis and I, again spend a lot of time on this stuff. A good example. We worked on the XE-9680, our most popular best product we launched. We started to do that more than two years before Gen.AI came up, probably three. And part of it was AI was just evolving and NVIDIA was starting to work on those chip sets, the H100. But then what we announced today with the NINETY-9680L only took nine months. So first process was two to three-year development, this one, nine months. And so what we did, we created a very modular platform because if you notice also at the AMD event, we announced the MI-300X, and we announced day one we're taking orders unlike anyone else, so very modular platform to be able to handle all of that, very high-speed innovation. I think probably the biggest new thing is we're learning... Everything we're learning, we're incorporating into our own map very quickly, technology and product.

>> I was just going to add, I think that the initial good bet that we made was the need to be able to support a variety of different GPUs, a variety of different CPUs, both air and liquid cooling way back when. And that initial-

>> That was a good call.

>> That initial decision is what has allowed us to continue to iterate very quickly.

>> So 9680 is air-cooled and you got choice of Ethernet or InfiniBand, the 9680, I had XL, it's L, just L.

>> That's right.

>> Liquid cooling, Blackwell, you got density improvements, you got 2X energy efficiency and you got better networking. 12 PCIe slots.

>> That's right. 12, which is industry-leading. Again, that's huge.

>> And industry-leading density for you. And then you got rack solutions, both air and liquid cooling.

>> That's right.

>> You got a seventy-kilowatt air-cooled version, 64 GPU, a hundred kilowatt liquid cool for 72 Blackwell GPUs.

>> You should slow down on this one, Dave. That is leading.

>> Don't gloss over that part.

>> Don't gloss over that part.

>> No, I mean I'm emphasizing that's dense in a good sense. And then 130 kilowatt design on next gen ORV-III, which is a new form factor, right? That's a-

>> Yes. That's 21 inch. Yeah. But the 19 inch, both the 70 KW air-cooled and also the 100 KW liquid-cooled are very important for one reason. The industry is facing many challenges of data center capacity. It's facing challenges of power cooling. There's long lines for buying chillers. This is incredible for customers. Both of them, because they have this heat exchangers allow you to add them without having to add chiller capacity. And it's the most dense and efficient of this many GPUs in the market.

>> So here's my question. So over the last several years, all the hyperscalers have extended their depreciation schedule on servers by a couple of years actually. It went from I think four years now it's six years. And they do that because it hits the income statement. That's good. Here's my question. Given the pace of innovation and the continuous need for lower energy, lower power, will server cycles compress because of AI?

>> Dave, I think it remains to be seen. You're absolutely right in pointing out that the cadence for GPUs is very quick, and the innovation from our silicon partners is enormous. And what we've seen to date is that there is a definite desire, especially by some of the larger providers to go to the latest and greatest. I think as we start to build out, maybe we'll reach some sort of steady state. Love your thoughts here. As we sit here today, the focus is on deploying the latest because there's such a performance benefit, a density benefit, a capability benefit from going from previous generation to current generation.

>> Yeah. What's happening is a couple of things. First, customers don't know what to pick. Too many choices. We spend a lot of times helping them pick the right one. What accelerator do I go... Air cool, liquid cool. Do I wait for that one or take this one? But number two, we have a lot of discussion on our Apex model because that's the way to de-risk that whole depreciation topic for customers, and we're getting a lot more attention to that. So between those two, we're helping customers navigate this, but there's no doubt that the cycles are speeding up significantly.

>> And we don't see a slowdown.

>> We don't see a slowdown at all.

>> And on Apex, you own the paper or the customer owns the paper. It depends, right? You could do both ways?

>> Apex, it depends.

>> But I'm saying somebody owns the paper.

>> That's right.

>> And there was an article at Financial Times today that, and I haven't corroborated, but it said that Amazon has put the brakes on Hopper purchases and they're waiting for Blackwell, which says to me two things. One is maybe Blackwell's coming sooner than we thought, but the other is that maybe server cycles are compressing. That would be my bet. We don't know yet.

>> I think you're right.

>> It's going to depend on how much value the AI brings. If AI brings more value, you're going to be forced. Like we used to have to buy PCs every two years. We wanted the latest and greatest. It made us more productive. Could be a similar dynamic. I mean, I could argue the other way too. More power. Maybe they'll last longer. I don't know. It's going to be interesting to see.

>> There continues to be demand, continues to be more applications, more influencing demand. The demand is there from what we see and interest is significant. It's a matter of optimizing for what AI workloads use what accelerator.

>> The other thing I could see is that today's training systems become tomorrow's influences.

>> That's correct. Yeah. We were having a conversation earlier where the investments made today you might see a tiering of that equipment down to rag or inferencing or something that's not necessarily needing the absolute maximum in performance and capability.

>> Yeah. Well, and retooling that, repurposing that. I mean, sustainability is a big goal of Dell's as well. You're talking about modular solutions, that drives that as well. If you can plug and play, remove things or even adapt to a different part of your stack and then upgrade as necessary when you need to, it really optimizes the cost as well as the experience for all of your customers. Should every enterprise be using Dell or are you a fit only for a particular type of customer?

>> The great thing about working for Dell, and I've been doing that for two decades as of a couple of weeks ago-

>> Congratulations....

>> is that we are able to service customers of all sizes. And you're starting to see it in the AI factory work that we're doing is some of the solutions are really focused on the cutting edge of large-scale training. But we truly believe that there is a second wave coming as generative AI makes its way into the enterprise. Customers realize that potentially a smaller purpose-built model trained on the right data is more beneficial to them than a larger model running in the public cloud. And if you look at the Dell server and storage portfolio, we service customers of all sizes. And so I think we'll see a variety of solutions inclusive of GPUs and CPUs all working on generative AI deployments of all sizes.

>> Is your CTO a new role or is it always been-

>> No, he's always been a CTO.

>> He's always been a CTO.

>> Because I always thought John Rose was CTO, but-

>> John Rose is the corporate CTO. He's the overall CTO. I'm the CTO for our AI compute and networking and the ISG business unit.

>> All the cool stuff.

>> Yeah.

>> Oh yeah, it is.

>> Thank you.

>> Bummer.

>> I'm having a good time. But I mean, John also is involved in the same technologies and he drives a lot of the long-term innovation, lots of research in that area as well.

>> So what do you think, we were talking to John about quantum and AI coming together. How do you look at that?

>> Yeah, we definitely see quantum starting to become promising. We see that it'll be a hybrid model where for specific workloads you would use quantum and then you use traditional CPU, GPU for the rest of it. It's just quantum is not deployable easy today. It's very big. It needs a lot of cooling, but we are very engaged in that research. At some point, it'll be an add-on to the optionality.

>> Jensen at GTC, we were in the analyst thing and he said, "We're probably the biggest quantum computing company that doesn't make quantum computers." That was interesting. And then he said something that was very controversial amongst the analysts. He said, "Well, pretty much anything you could do with Quantum, you could do with AI, except for encryption." Now we're like, whoa, wait a minute. I wish you were right here. We could ask him. But it seems like that didn't make sense.

>> There are new use cases with quantum. One of them is sensing.

>> Explain that.

>> And by that I mean it has the ability to sense at a molecular level things that would be very hard to do in an additional compute. I'll give you an example. A stadium full of people, if you want to know how many people have COVID, quantum can sense that very quickly compared to additional compute.

>> Based on symptoms?

>> Based on the ability to deconstruct the molecular construct of all the different people and what's there.

>> That's mind-blowing.

>> That's very complex to do. You would have to spend enormous money on a lot of GPUs and an additional compute. Too expensive to do, but very quick to do in a quantum, just as an example.

>> No, but that helps frame it. The quantum, to the extent that you can make it stable, is going to be more efficient relative to throwing-

>> Specific use cases. That sensing is different than what enterprises deal with. They may not need to do sensing.

>> Thank you for that. That's good.

>> Well, quantum's going to save us all from getting the conference cold we always end up getting when we come to Las Vegas every time. All right, last question for you both since you are such regulars on the show and we appreciate that very much. What are you going to be able to tell us or hope to be able to tell us when you join us at the next event? Which could be SC24. Could be Dell Tech World, could be... We see you along the way everywhere, so could be anything. Travis, I'm going to start with you.

>> I know it's going to be more, it's going to be bigger, it's going to be better. I hope to be able to tell you that the early traction that we're seeing in the enterprise is materialized. I hope to tell you that we see enterprise customers moving from POCs and early deployments into full-scale production. And like Michael talked about on stage, I think that what I'm most excited to be able to tell you is the thing that I have absolutely no idea what it would be because it's such a use case that we've never thought of.

>> I love that. Great answer.

>> If I go to last year, we announced Project Helix, look what happened. Look how much new stuff we uncovered between now and then. And now we have the full AI factory. In the middle of the last 12 months, we didn't talk about Meta and Hugging Face and we just announced massive progress with both of them. I think those are the highlights for me of today of our announcements. So we cannot predict a year from now, but if we continue at the pace, you can only predict how big of a change next year will be.

>> Those are very diplomatic answers and also very lovely. Ehab and Travis, thank you so much for being here. You're always fantastic guests. We love seeing you, especially towards the end of the day to keep the energy up. Dave, great job today. You're just absolutely wonderful and thank all of you for tuning in. Be sure and let us know what you're thinking in the comments. We're here at Dell Tech World in fabulous Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.

Travis Vigil & Ihab Tarazi | Dell Technologies

search

>> Good evening Dell fans and welcome back to the desert. We're here in fabulous Las Vegas, Nevada wrapping up our second day of live coverage of three days of Dell Tech World. It has been spicy, it has been fun, and it has been very interesting. My name's Savannah Peterson, joined by my fabulous co-host, Dave Vellante. Dave, what an evening. My gosh.

>> I know, two evenings in a row.

>> I know. We are the nightly news for enterprise tech now.

>> We haven't been outside in days.

>> Is there an outside world? I'm not even sure. Very excited, particularly for this point in the day to have two Dell superstars and also two very multiple-time CUBE alumni alums.

>> Yes.

>> Yeah. Travis Viejo, thank you so much for being here.

>> Thank you for having me.

>> And Ehab Terozi, it's always a pleasure to see your smiling face.

>> Thank you.

>> Love it. You made it right at the buzzer tonight just to grace us with your fabulous presence. Big week for Dell. Huge week for Dell. How's it been for you, Travis? What's the week been like?

>> It's been a whirlwind.

>> I bet.

>> It's the AI edition of Dell Tech World.

>> Certainly is.

>> And the announcements around the AI factory, the new systems, the new storage, the new networking have been fantastic and worthy of a Dell world all by themselves. But in addition to that, we have PowerStore Prime, the biggest release for PowerStore in the last four years. We had some PowerFlex announcements. It's just been chock-full of innovation and really a culmination of the last year of focusing on generative AI first, private cloud, multi-cloud second, and just the ISG organization is firing on all cylinders when it comes to innovation.

>> Ehab, we've talked about Dell's networking strategy and you're kind of changing the game there. Dell's never been a leader in networking. I'm sorry, don't hate me for saying that, but I'm sure you've thought about that and said, okay, how do we change the equation? And that's really what you're doing. Can you explain your strategy and why it fits in the AI era?

>> Yeah. I would say first of all, Travis and I have been working the network strategy for a long time in Dell. What we bet was on the future and AI validated all the decisions we made. What we bet on is that networking needs to be integrated with compute and storage and it needs to be very high throughput. And also, we also bet on open source and next generation software. So we are big partners of Broadcom and we're also working now closely with NVIDIA. The networking strategy we pick to get high performance switches with significant throughput and fully integrate the stack with compute is exactly what AI needs. What we showed today is how our strategy is just accelerating and scaling new components from Broadcom, Thor 2, Thor 3, co-packaged optics, Tomahawk 5 specifically that we launched a switch on is all part of that strategy. And if you saw, we also dialed in on NVIDIA with Spectrum X, Spectrum 4, InfiniBand Switching and the BlueField 3 DPUs. So it's really interesting to me because now with AI being as big as it is and how big of a use case that is and how AI acquires that integrated model, this not only helps us with AI, but helps us with enterprise because that is the future.

>> Well, it's interesting because networking's never been a centerpiece, but now it is, and it's not like as Michael was talking about, we are going to disrupt ourselves or be out of business in five years. So this is a nice target to disrupt. You can go hard after this. How do you think it's going to change the equation in the market dynamic?

>> I think generative AI offers a whole new landscape for networking. And if you look at the issues that folks are trying to solve as they implement small, medium, or large scale or generative AI deployments, the GPU to GPU communication is a whole new ball game. And we really have a strategy that's focused on providing a streamlined solution for that use case with a high degree of automation. And it's a great compliment to our server and storage portfolio to be able to package up compute, networking, and storage for not just a net new use case for networking, but a net new use case in general where customers are really looking for that integration and for that automation.

>> And also now networking extends to the NIC and inside the server. So that's a big change to the market. Selling a switch by itself is not really easy anymore. You really have to provide a system. And then you add to that the fact that AI is expanding at a massive speed, you have to jump to 800 gig next and do this next. So it's a completely different market dynamic. I'll tell you a story. I talked to a customer today about networking and they said, "Oh, we have a really good solution from an additional vendor." I'm not going to say who. And I said, "What's your throughput? What is your GPU to GPU throughput?" And we said, "Well, we really don't know." I said, "Most likely your throughput, you're given away half of the utilization of the GPUs that already scarce and expensive." And that kind of a change in metric is going to have a huge impact in the market.

>> And we were talking at MWC that there aren't a lot of companies that can do both the sort of internal communication across the XPUs and the external switching. And you guys are consumers of both of the leading companies that do that. NVIDIA and Broadcom, you've got great relationships with both of them. These are really hard technologies to get right.

>> They're really hard technologies. And I mean, you're bringing up a great point. It's all about this ecosystem. We've heard it on stage, open ecosystem. We're talking about open source. And it's not just these one-off, to your point, a switch or a component or one single player, it could be irrelevant six months from now. What's going to matter is these ecosystems playing together to provide inclusive solutions to your customers. And I mean, I got to give you guys a shout-out. Last week, Morgan Stanley saying that you are the AI stack of the enterprise future. It's a pretty bold claim. You're obviously seeing the stock smile as a result of that deservedly so, which is exciting. But I thought that was really interesting timing given we're leading into your week. You have all these announcements going on, and your keynote yesterday is literally all of the AI celebrities on the planet hanging out and patting each other on the back, and then we get to hang out with y'all. So I'm curious, you guys have been anticipating this moment for a while. How long have you been working on your collaboration, making sure that everything went together with all these partners to get us to this party that we're at today basically?

>> I think back to a year ago, and I think we were just entering in sort of this inflection. ChatGPT was just starting to enter everybody's consciousness and we announced Project Helix.

>> We talked about it a lot this week.

>> And at the same time, you go back two years from there. And that was the point in time where we made the decision to create the XE-9680. So it was two years prior to a year. So three years ago where we made that investment and we could see a lot of pieces coming together. We could see that the GPUs were going to be more critical going forward. But I don't think I'd be telling the truth if I said we knew exactly how big it was going to get over this last year.

>> That's kind of what I was wondering.

>> Right. Yeah. And I think Ehab can attest from last year to this year, it has been just very agile development, constant focus on providing solutions, an intense amount of focus on large-scale training. We think that the inferencing boom and the enterprise boom of generative AI is going to follow. We were thinking that we're on the front end of that, but I think it was good foresight combined with good luck. And this last year, I think we've really been agile in addressing the opportunities, quick in addressing the opportunities, introducing new technology at a rate that we have to have to keep up with the pace of the industry.

>> That's my follow-up question there. How do you keep up? I mean, you're not only playing ball, you're leading the game and helping everybody stay in the arena.

>> Yeah, I think Travis and I, again spend a lot of time on this stuff. A good example. We worked on the XE-9680, our most popular best product we launched. We started to do that more than two years before Gen.AI came up, probably three. And part of it was AI was just evolving and NVIDIA was starting to work on those chip sets, the H100. But then what we announced today with the NINETY-9680L only took nine months. So first process was two to three-year development, this one, nine months. And so what we did, we created a very modular platform because if you notice also at the AMD event, we announced the MI-300X, and we announced day one we're taking orders unlike anyone else, so very modular platform to be able to handle all of that, very high-speed innovation. I think probably the biggest new thing is we're learning... Everything we're learning, we're incorporating into our own map very quickly, technology and product.

>> I was just going to add, I think that the initial good bet that we made was the need to be able to support a variety of different GPUs, a variety of different CPUs, both air and liquid cooling way back when. And that initial-

>> That was a good call.

>> That initial decision is what has allowed us to continue to iterate very quickly.

>> So 9680 is air-cooled and you got choice of Ethernet or InfiniBand, the 9680, I had XL, it's L, just L.

>> That's right.

>> Liquid cooling, Blackwell, you got density improvements, you got 2X energy efficiency and you got better networking. 12 PCIe slots.

>> That's right. 12, which is industry-leading. Again, that's huge.

>> And industry-leading density for you. And then you got rack solutions, both air and liquid cooling.

>> That's right.

>> You got a seventy-kilowatt air-cooled version, 64 GPU, a hundred kilowatt liquid cool for 72 Blackwell GPUs.

>> You should slow down on this one, Dave. That is leading.

>> Don't gloss over that part.

>> Don't gloss over that part.

>> No, I mean I'm emphasizing that's dense in a good sense. And then 130 kilowatt design on next gen ORV-III, which is a new form factor, right? That's a-

>> Yes. That's 21 inch. Yeah. But the 19 inch, both the 70 KW air-cooled and also the 100 KW liquid-cooled are very important for one reason. The industry is facing many challenges of data center capacity. It's facing challenges of power cooling. There's long lines for buying chillers. This is incredible for customers. Both of them, because they have this heat exchangers allow you to add them without having to add chiller capacity. And it's the most dense and efficient of this many GPUs in the market.

>> So here's my question. So over the last several years, all the hyperscalers have extended their depreciation schedule on servers by a couple of years actually. It went from I think four years now it's six years. And they do that because it hits the income statement. That's good. Here's my question. Given the pace of innovation and the continuous need for lower energy, lower power, will server cycles compress because of AI?

>> Dave, I think it remains to be seen. You're absolutely right in pointing out that the cadence for GPUs is very quick, and the innovation from our silicon partners is enormous. And what we've seen to date is that there is a definite desire, especially by some of the larger providers to go to the latest and greatest. I think as we start to build out, maybe we'll reach some sort of steady state. Love your thoughts here. As we sit here today, the focus is on deploying the latest because there's such a performance benefit, a density benefit, a capability benefit from going from previous generation to current generation.

>> Yeah. What's happening is a couple of things. First, customers don't know what to pick. Too many choices. We spend a lot of times helping them pick the right one. What accelerator do I go... Air cool, liquid cool. Do I wait for that one or take this one? But number two, we have a lot of discussion on our Apex model because that's the way to de-risk that whole depreciation topic for customers, and we're getting a lot more attention to that. So between those two, we're helping customers navigate this, but there's no doubt that the cycles are speeding up significantly.

>> And we don't see a slowdown.

>> We don't see a slowdown at all.

>> And on Apex, you own the paper or the customer owns the paper. It depends, right? You could do both ways?

>> Apex, it depends.

>> But I'm saying somebody owns the paper.

>> That's right.

>> And there was an article at Financial Times today that, and I haven't corroborated, but it said that Amazon has put the brakes on Hopper purchases and they're waiting for Blackwell, which says to me two things. One is maybe Blackwell's coming sooner than we thought, but the other is that maybe server cycles are compressing. That would be my bet. We don't know yet.

>> I think you're right.

>> It's going to depend on how much value the AI brings. If AI brings more value, you're going to be forced. Like we used to have to buy PCs every two years. We wanted the latest and greatest. It made us more productive. Could be a similar dynamic. I mean, I could argue the other way too. More power. Maybe they'll last longer. I don't know. It's going to be interesting to see.

>> There continues to be demand, continues to be more applications, more influencing demand. The demand is there from what we see and interest is significant. It's a matter of optimizing for what AI workloads use what accelerator.

>> The other thing I could see is that today's training systems become tomorrow's influences.

>> That's correct. Yeah. We were having a conversation earlier where the investments made today you might see a tiering of that equipment down to rag or inferencing or something that's not necessarily needing the absolute maximum in performance and capability.

>> Yeah. Well, and retooling that, repurposing that. I mean, sustainability is a big goal of Dell's as well. You're talking about modular solutions, that drives that as well. If you can plug and play, remove things or even adapt to a different part of your stack and then upgrade as necessary when you need to, it really optimizes the cost as well as the experience for all of your customers. Should every enterprise be using Dell or are you a fit only for a particular type of customer?

>> The great thing about working for Dell, and I've been doing that for two decades as of a couple of weeks ago-

>> Congratulations....

>> is that we are able to service customers of all sizes. And you're starting to see it in the AI factory work that we're doing is some of the solutions are really focused on the cutting edge of large-scale training. But we truly believe that there is a second wave coming as generative AI makes its way into the enterprise. Customers realize that potentially a smaller purpose-built model trained on the right data is more beneficial to them than a larger model running in the public cloud. And if you look at the Dell server and storage portfolio, we service customers of all sizes. And so I think we'll see a variety of solutions inclusive of GPUs and CPUs all working on generative AI deployments of all sizes.

>> Is your CTO a new role or is it always been-

>> No, he's always been a CTO.

>> He's always been a CTO.

>> Because I always thought John Rose was CTO, but-

>> John Rose is the corporate CTO. He's the overall CTO. I'm the CTO for our AI compute and networking and the ISG business unit.

>> All the cool stuff.

>> Yeah.

>> Oh yeah, it is.

>> Thank you.

>> Bummer.

>> I'm having a good time. But I mean, John also is involved in the same technologies and he drives a lot of the long-term innovation, lots of research in that area as well.

>> So what do you think, we were talking to John about quantum and AI coming together. How do you look at that?

>> Yeah, we definitely see quantum starting to become promising. We see that it'll be a hybrid model where for specific workloads you would use quantum and then you use traditional CPU, GPU for the rest of it. It's just quantum is not deployable easy today. It's very big. It needs a lot of cooling, but we are very engaged in that research. At some point, it'll be an add-on to the optionality.

>> Jensen at GTC, we were in the analyst thing and he said, "We're probably the biggest quantum computing company that doesn't make quantum computers." That was interesting. And then he said something that was very controversial amongst the analysts. He said, "Well, pretty much anything you could do with Quantum, you could do with AI, except for encryption." Now we're like, whoa, wait a minute. I wish you were right here. We could ask him. But it seems like that didn't make sense.

>> There are new use cases with quantum. One of them is sensing.

>> Explain that.

>> And by that I mean it has the ability to sense at a molecular level things that would be very hard to do in an additional compute. I'll give you an example. A stadium full of people, if you want to know how many people have COVID, quantum can sense that very quickly compared to additional compute.

>> Based on symptoms?

>> Based on the ability to deconstruct the molecular construct of all the different people and what's there.

>> That's mind-blowing.

>> That's very complex to do. You would have to spend enormous money on a lot of GPUs and an additional compute. Too expensive to do, but very quick to do in a quantum, just as an example.

>> No, but that helps frame it. The quantum, to the extent that you can make it stable, is going to be more efficient relative to throwing-

>> Specific use cases. That sensing is different than what enterprises deal with. They may not need to do sensing.

>> Thank you for that. That's good.

>> Well, quantum's going to save us all from getting the conference cold we always end up getting when we come to Las Vegas every time. All right, last question for you both since you are such regulars on the show and we appreciate that very much. What are you going to be able to tell us or hope to be able to tell us when you join us at the next event? Which could be SC24. Could be Dell Tech World, could be... We see you along the way everywhere, so could be anything. Travis, I'm going to start with you.

>> I know it's going to be more, it's going to be bigger, it's going to be better. I hope to be able to tell you that the early traction that we're seeing in the enterprise is materialized. I hope to tell you that we see enterprise customers moving from POCs and early deployments into full-scale production. And like Michael talked about on stage, I think that what I'm most excited to be able to tell you is the thing that I have absolutely no idea what it would be because it's such a use case that we've never thought of.

>> I love that. Great answer.

>> If I go to last year, we announced Project Helix, look what happened. Look how much new stuff we uncovered between now and then. And now we have the full AI factory. In the middle of the last 12 months, we didn't talk about Meta and Hugging Face and we just announced massive progress with both of them. I think those are the highlights for me of today of our announcements. So we cannot predict a year from now, but if we continue at the pace, you can only predict how big of a change next year will be.

>> Those are very diplomatic answers and also very lovely. Ehab and Travis, thank you so much for being here. You're always fantastic guests. We love seeing you, especially towards the end of the day to keep the energy up. Dave, great job today. You're just absolutely wonderful and thank all of you for tuning in. Be sure and let us know what you're thinking in the comments. We're here at Dell Tech World in fabulous Las Vegas, Nevada. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.