CES 2026 | Marc Hammons & Travis North, Dell Technologies

Clips
More from CES 2026

Marc Hammons

Sr. Director, CSG

Dell Technologies

Travis North

Sr. Distinguished Thermal Engineering Technologist

Dell Technologies

Savannah Peterson

Principal Analyst & Host

SiliconANGLE Media, Inc.

play_circle_outline Introduction to Dell's CES coverage with exclusive interviews about AI and hardware.

play_circle_outline Introduction of a new thermal interface material, Element 49, for better heat transfer.

Info
Transcript

Marc Hammons & Travis North, Dell Technologies

Marc Hammons

Sr. Director, CSG Dell Technologies

Travis North

Sr. Distinguished Thermal Engineering Technologist Dell Technologies

Savannah Peterson

Principal Analyst & Host SiliconANGLE Media, Inc.

HOST

In this interview from CES 2026, Marc Hammons and Travis North from Dell Technologies join theCUBE’s Savannah Peterson to showcase the hardware powering the next generation of local AI. The conversation highlights the new Dell Pro Max lineup, specifically the GB10 and GB300, designed to bring "AI Everywhere" by enabling massive models to run securely on-device. Hammons demonstrates an agentic workflow using the GB10 to summarize news feeds locally, while North breaks down the breakthrough liquid cooling engineering required to manage 1,000-watt GPUs. Together... Read more

explore Keep Exploring

What is being discussed in the Dell CES coverage segment involving Savannah, Marc, and Travis? add

What new thermal interface material has been designed for the dye, and how does it improve the cooling efficiency compared to previous materials? add

bolt Powered by CUBE AI

Marc Hammons & Travis North, Dell Technologies

search

Savannah Peterson

>> Good afternoon, Nerd Fam, and welcome back to our Dell CES coverage. Very excited to be bringing a series of exclusive interviews to you as a part of our coverage of the event. My name's Savannah Peterson, joined with a repeat guest and a new guest for me today. Marc and Travis, thank you so much for hanging out.

Marc Hammons

>> Thanks for having us.

Savannah Peterson

>> We're going to be talking about AI, but more importantly, we're going to be talking about the hardware that hours that AI, which is really cool. I'm going to open it up with you, Marc. Generative AI, you've got some really interesting workflows going on here. Tell me what's happening.

Marc Hammons

>> Oh, yeah. Awesome. So, thanks, Savannah. We've got with us today, the Dell Pro Max with GB10. It's our newest little workstation powerhouse.

Savannah Peterson

>> It is a little workstation powerhouse too, just the sizes.

Marc Hammons

>> It's pretty fantastic. What you see here is two stacked together, so 128 gigs of RAM, unified memory across the 20 cores in the CPU. And then our Blackwell GPU. Pretty fantastic performance. We're bringing all kinds of models down on this device.

Savannah Peterson

>> Small, but seriously mighty.

Marc Hammons

>> Seriously mighty. And we scope this out for maybe education, maybe data scientists, but it can go a lot further beyond that.

Savannah Peterson

>> Ooh.

Marc Hammons

>> We start to think of this thing as an appliance. You can imagine putting it in a small or medium business, maybe on your shelf, adjunct to your central compute that you use day in and day out. You can bring your own models down. You can do all kinds of fantastic things right there on that device.

Savannah Peterson

>> That is awesome. What's going on here? Can you give me an example of one of those models?

Marc Hammons

>> Yeah. Let me give you this here. This is kind of an agentic workflow, so we've taken our application and we've embedded something called n8n.io, which is a really great application that allows you to kind of stitch together some little basic functions and build an entire workflow. Now, this is something we all like to do. We like to go out, we like to read news. We have our favorites. We might want to curate that. We might want to go and maybe get summaries of those articles, so that we can dive in deeper. So, I want that summarization before I'm going to invest the time and take 10, 15 minutes to read it. This will do that for me, so I can set this up so that it'll run here on my GB10. It'll go out and fetch the workflows. I'm going to go ahead and start it. So, it's going to go to a pre-configured set of RSS feeds. It's going to ingest those feeds. It's going to vectorize them. It's going to store them down in a database. It's going to sit here and do that on a periodic basis. And then I'm going to pick another workflow that on a separate timer will come in, go and pull those workflows out of that vector store and then feed them to an LLM and summarize and normalize them, put them into an email and send that to me. That way I just get the email, pop it open, read through those summaries, and then I can dive in and figure out what I want.

Savannah Peterson

>> So, this is why you always seem so smart and put together?

Marc Hammons

>> Exactly. It's all about this machine. It has nothing to do with me.

Savannah Peterson

>> And then behind you here, we have this beauty. Travis, talk to me about this.

Travis North

>> Yes, ma'am.

Savannah Peterson

>> Dell has a history of really owning patents and IP in the thermodynamic and cooling space. It looks like this is no exception. What's happening?

Travis North

>> Great question. This is our Dell Pro Max with GB300. And what you're seeing here is basically the cooling solution and the illustration of what we've done to really provide value. This solution, if you look at it, if you flow with me, it starts with a dual pump flow mechanism that gives us the complete redundant between Pump A, Pump B. These pumps are basically cooling a 1000-watt GPU, shown here in this big copper plate in the middle.

Savannah Peterson

>> Oh, yeah.

Travis North

>> And there's bypass channels that allow us to cool both the CPU and the CX8 in parallel. After that, all that fluid is basically captured and returned back to a common fluid return port, which we run through two tubes up to two separate radiators. We've got a 420 millimeter radiator at the top, a 360 millimeter radiator front. This whole solution is basically in parallel. And so you have to think about it just most of the solutions today are done in series, so it flows from CPU to GPU all in line. This is all flowing in parallel, so there's a very small pressure drop, which allows us to run these pumps at really low speed. And the one way to think about this, if you look at your PCI slots today and like gaming, you can see I've got my Alienware shirt on, basically three slots is about 400 watts. We are cooling six slots of power within an area in the small area of this cold plate. So, it's 1,000 watts on that GPU dot. Now, on top of that-

Savannah Peterson

>> That's no easy feat.

Travis North

>> No, it's not. It's crazy to think about where we were not two years ago .

Savannah Peterson

>> I was just going to ask you, at what point... Actually, I'm going to ask you, and then I'll let you keep going, because it's relevant. At what point did you realize this was even possible?

Travis North

>> I think if you look at the H Series, H100, H200 over to the B Series where it's starting to scale, that's really where we saw power densities skyrocketing. It really came from AI. What you want to be able to do is run a lot of data in parallel. And just think about it from a laptop, 40 trillion... 40 tops, right, 40,000 billion operations per second, you're running an eight billion parameter model. This one is targeted at like one trillion parameters, so it gives you the capability to run the biggest models locally on the system. This is all closed loop. And so basically it's plug and play. You drop it, plug it into the wall, you're done. There's nothing else that you need to do. You can run the models and it's completely cohesive. Additionally, on this, there is a new thermal interface material that we had to design for that same dye. And what you're seeing in the industry is that the system itself, that's like 25% of the cooling heat. Going from the silicon through the lid, through the thermal interface material into the cold plate, that is now like 75% of the challenge to cool. And so we have a new material called Element 49 that is a mixture of gallium and indium and a special component that is basically-

Savannah Peterson

>> The secret component you can't tell us about.

Travis North

>> I can't tell you about. I would love to, but I can't. It's about 4X the connectivity of what we're shipping on our Alienware product, the previous Element 31. And so that's a core focus for Dell. We're going to continue to focus on that core area. On top of that, it's EG35, which gives us the capability to basically ship this anywhere. So, there's freeze protection. In the data center world, they use a little propylene glycol, 25%. You usually have to fill it at the data center. So, they pull a vacuum, they backfill it because they want to be able to use that because they can't ship it in that condition, so it's a little bit better .

Savannah Peterson

>> Yeah. It's an interesting constraint.

Travis North

>> Yep, but this one isn't. So, it's designed for basically drop and run. We can ship it anywhere. It's local compute. Additionally, we did all of this for acoustics in addition to basically just the thermal compute. It's desk side, so you can run that trillion parameter model locally, firewalled off, and then you can do it in a good experience. Yeah.

Savannah Peterson

>> Honestly, just thinking about that is insane.

Marc Hammons

>> And even better, we can couple two of them together using the CX8 interface that will interconnect to this.

Savannah Peterson

>> Oh my goodness. What?

Marc Hammons

>> Absolutely. So, you got your own QSM-

Savannah Peterson

>> It's like its own little mini AI factory.

Marc Hammons

>> Absolutely.

Savannah Peterson

>> Wow. Okay. And you've done this extremely fast, right?

Travis North

>> Yes.

Savannah Peterson

>> The design process on this as well?

Travis North

>> And that is different. And our Chairman talked about it today, we built this prior to the Board being available, and so the intention was to pre-design. And so there was a lot of iterations that go with that type of process, but instead of a process taking years, it can take months. And so it allowed us to be very dynamic. We're going to continue with that process and scale. We have a bunch of IP, which patent pending, as always, with this process, but we are very proud. And we think that this is going to be a different shader for Dell as we scale. And I do expect that you'll see some of these innovations scale over to our other product lines over time.

Savannah Peterson

>> Yeah. It's nice that you can crosspollinate that way.

Travis North

>> Absolutely.

Savannah Peterson

>> Thank you, Travis. I'm curious, what sort of use cases? Why are we making these? Who's going to use it?

Travis North

>> Well, I can go first. If you talk about the medical imaging, that MRI machine, you want to do it local. They do the image. You have the HIPAA requirements, all of your data requirements. It's local on the system. You plug it in with the machine, you're done. And so it goes directly to the doctor, here's the output. That's a primary one. Marc, I'm sure you could talk to better ones than I can.

Savannah Peterson

>> Well, healthcare is a great one, just to be very clear.

Marc Hammons

>> It really is. Those use cases where, "I want that data local, it's got to remain private, I don't ever want it leaving." You can imagine there are federal use cases that are very, very similar where very sensitive things are being kept. Or maybe just in the enterprise, I may have built my own model and I want to stand that up, and I don't want that to go to the cloud. I don't want my weights going to the cloud. We have to protect those. That's where all my IP is tied up. So, those are use cases that we see day in and day out where a device like this really gives us that transformative capability without having to have all the networking infrastructure, all the storage infrastructure, everything that necessitates-

Savannah Peterson

>> Basically all fit in there.

Marc Hammons

>> That's right. You have to manage all of that in an enterprise setup. Now it's all in one compartmentalized device. Pretty fantastic.

Savannah Peterson

>> It's really fantastic. It's exciting.

Travis North

>> Yeah. I think to build onto factory, if you look at the scaling, if you looked at what Marc talked about earlier with GB10, you've basically got a full stack from small billion parameter model up to trillion. You can do it locally out of any wall outlet of your house, generally, if it's designed for that. And then after that, you can scale .

Savannah Peterson

>> It's just crazy. That's just nuts to think in a few years I could have one at home to power... . That might be good. We'll figure that out. I got one final question for you gentlemen. Dell's clearly having a moment. They're putting the consumer front and center. We're seeing a lot more enthusiasm and excitement on the marketing side. What do you wish the world knew about Dell hardware? Marc, I'm going to start with you.

Marc Hammons

>> Well, I really wish they knew about the fantastic engineering that goes into these solutions. Travis and I were actually talking about that very thing just earlier today. As a consumer, you never really realize all of the effort and engineering that goes into these products, how we work extremely hard to make them thinner, lighter, more performant, more powerful, and still fit the client envelope. It's got to plug into a 110 socket. It's got to meet certain acoustic characteristics. It's got to meet thermal characteristics, so you can put it on your lap or touch it with your hands. People don't really realize, a lot of times, the engineering that's required to make that happen.

Savannah Peterson

>> Yes, and the amount of IP that you generate as a result of that.

Marc Hammons

>> Oh, yeah.

Savannah Peterson

>> I've always been so impressed by you on that front. It's cool. What about you, Travis? What do you wish everybody knew?

Travis North

>> I think Marc nailed it, but I'd also add reliability. I don't think a lot of people understand that when we go test these products, it's 2,000 hours. We do 40C ambient testing. There's thermal shock. There's a chamber where it's 40. They drop it to five. On mobile, we've done videos on this in past, but like water spray.

Savannah Peterson

>> I bet that's a fun word-

Travis North

>> Water spray.

Marc Hammons

>> It's a great area at Dell.

Savannah Peterson

>> Yeah, I was going to say, wow, next time I'm around, you guys got to show me.

Marc Hammons

>> When you get to hit the rugged with a water cannon, it's great.

Savannah Peterson

>> Yeah. You're just a little stressed out at work some days.

Marc Hammons

>> That's right. Exactly.

Travis North

>> The smash room, we're just going to let you drop stuff.

Marc Hammons

>> Yeah, exactly.

Travis North

>> But I think the reliability is a big one because you don't see it, but on most of the boards, everything's derated, so every capacitor, every inductor, every FET is thermal coupled. And they actually take that and they do a derating for the full life of the product. And so that's a lot of the value that you get when you buy something from Dell. It's not just, "Hey, I'm packaging somebody's silicone hardware." It's the fact that you know it's going to last three to five years. That's the core difference.

Savannah Peterson

>> Yeah.

Travis North

>> It is.

Savannah Peterson

>> And at that point, who knows what computers and devices will be talking about.

Travis North

>> That's true.

Savannah Peterson

>> It's completely crazy. Travis and Marc, thank you so much. I know it's a busy week and this was absolutely delightful and impressive.

Travis North

>> Thank you.

Marc Hammons

>> Thanks for this.

Savannah Peterson

>> Yeah. And thank all of you for tuning in to our exclusive coverage of Dell at CES 2026. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.