Sovereign AI Infrastructure You Actually Own

Hey everyone, welcome back here to Techstrong TV. I met this last gentleman recently at SUSECON. Oh, it wasn't so recently.

I think we're already talking about next year's- How recently? Well, time flies in the AI world. Anyway, let me introduce you to Rhys Oxenham.

Rhys is the VP and GM of AI at SUSE. Rhys, welcome back to Techstrong TV, but not quite in person this time, but similar. Absolutely.

Thank you so much for having me back. I was just thinking recently, back in April in Prague at SUSECON, right? Yes.

And it used to be April to July was still recently, but in the world of AI now, that's like ancient history, right? So much has gone on since then. It's crazy.

The rate of change is like nothing we've ever heard. I think that you and I, we were talking about this in Prague, actually. Mm-hmm.

About the rate of change that we're seeing and how previous generations of change, yeah, there were lots of things that have happened in our careers, but nothing like the rate of change we're seeing now. No. We recorded this on a Monday.

People may watch this on Tuesday or Wednesday. But you come in every weekend, and it's like you need to catch up. There's so much- Absolutely ...

that went on from Friday to here. Rhys, before we get into SUSE, and we're going to talk SUSE AI factory and so forth, I wanted to spend a quick moment or two about your background. Sure.

As I mentioned, you're the VP and GM of AI over at SUSE. You didn't study AI too much in school, I'll bet. We didn't think it would be here, but how does one prepare?

Right. What's your journey been like? AI wasn't exactly at the top of the- Yeah ...

list of activities back when I was a student. Uh-huh. Of course, the mathematics behind it was, of course, very relevant.

I studied computer science at university, and so it was somewhat relevant there, but I absolutely couldn't imagine that I'd ever get into the world of artificial intelligence. To me, back then, AI was science fiction. Sure.

I know that a lot of people have spent a lot of time working on AI, but it's only really over the last, I guess, five to 10 years that it's become a little bit more mainstream. A little bit. Yeah.

So mainstream, in fact, though, that SUSE has a VP and GM of AI, which seems- Sure ... makes me think that there's almost like, if not a separate business unit, certainly a substantial effort specializing in AI. If you wouldn't mind, share with us, how does SUSE look at AI that there's a GM of AI, right?

There must be a product line- Yeah ... go to market, everything. Yeah, sure.

As you and I, we discussed before, AI is changing the world. It's in every single thing we do. Everybody you talk to will have had some kind of interaction with AI.

So on the one hand, it's very natural that I think every organization in the world is looking at how they can either use or build on top of AI. For SUSE, it almost came naturally to us. If you look at where the world is going, and you see advantages through building artificial intelligence into your existing technologies, be that to run AI applications or use AI to support initiatives inside of your organization, SUSE becomes the ultimate bedrock for running and supporting those applications.

I think you and I, we chatted a lot before around what are the underlying infrastructure foundations, and are we perfectly suited for actually supporting AI thanks to the last 10, 20, maybe even 30 years' worth of innovation. And so, yes, SUSE is really building towards this kind of future, and we see the world of AI through two lenses as we chatted about before, AI for infrastructure and infrastructure for AI. Right.

Infrastructure for AI is kind of like, how do we take those foundations and use them to run and support AI workloads, for running generative AI inference, machine learning operations. And then you have AI for infrastructure, and this is where organizations want to leverage AI as part of their day-to-day operations. So think about it as maybe transforming the way an organization looks at doing CVE patching.

How do they look for security vulnerabilities in their estate? And so SUSE is trying to apply its experience, its knowledge, and providing capabilities to really solve for both worlds. Get it.

I get it. We were talking a little bit before we turned the cameras on. It came up in this book that I'm finishing right now, I'm writing, and AI, when we talk about infrastructure for AI, the good news is that a lot of the foundational work that's been done over, let's say, the last 15, even 20 years, first with the cloud itself, right?

Mm-hmm. And then with the cloud native stack, if you will, the cloud native community, including, of course, Rancher, and everything Rancher brings to the table from SUSE, really has become the stack upon which AI runs, right? AI didn't-- Yes, NVIDIA has CUDA, and we're going to talk about NVIDIA and SUSE in a moment.

Mm-hmm. But the fact of the matter is it's running on that Cloud Native stack has become the AI stack for many, many companies. And look, that has to play right into your sweet spot, right?

Because running infrastructure on top of SUSE, Linux, on top of Rancher is right in your wheelhouse. Absolutely, it is. I've been kind of wondering about this question.

Is this just circumstantial, or is it the perfect storm of events with regards to having all of these pieces on hand? So you know what? A mentor of mine, my friend Brad Feld, who's a pretty well-known VC, and he started Techstars and a bunch of other-- He'd done so many things, Brad.

He always used to tell me that 99% of technology's evolutionary, not revolutionary. And when you look at this, and I go into details on this in my book, it's actually built on the bones of the dot-com era. All that fiber that was laid- Sure ...

a trillion, two trillion back in those dollars, in those days' dollars, of fiber laid, and it took us 20 years to use, gave rise to the cloud, right? Because now Amazon and Google and Microsoft and the other cloud providers, they built the cloud on top of that ubiquitous amount of bandwidth that was out there, right? You couldn't have it.

And then also hypervisors, right? Virtualization. Yeah.

And then Cloud Native on top. But go ahead. I think there's lots of different layers and different sort of scenarios you have to take into consideration.

Fiber, of course- Mm-hmm ... like actually having the backbone to provide all this- Sure ... interconnectivity.

Absolutely right. You mentioned hypervisors, virtualization technology. I would say that a lot of the lessons that we've learned from the likes of edge computing- Yeah ...

telecommunications, low latency, whether it's real-time trading or real-time radio access network. There's a lot of capabilities that we've had to build out that ultimately today support the next generation of artificial intelligence-based workloads. And so I'm edging towards the perfect storm.

Like this is the absolute right time, and we have all of those capabilities to really support these. And so for us, tying it back to SUSE, we have a 30-plus-year history of deploying Linux into mission-critical environments in many of those settings that I just talked about. Sure.

Build on top of it with Rancher and Cloud Native, our experience in Edge and Telco, and now our kind of foray into managing artificial intelligence. That to me is, first of all, a great opportunity for us, but we're able to leverage, to your point, many years of experience. What came before.

Sure. Absolutely. Absolutely.

I think it's key to the whole thing. Reece, if you don't mind, we teased a little bit about NVIDIA. Recently, SUSE partnered with NVIDIA on something called SUSE AI Factory.

Mm-hmm. Tell us about it. Yeah.

Sure. So SUSE AI Factory, we actually made this announcement in April, during SUSECON. So SUSE AI Factory is our kind of flagship private enterprise AI offering.

And just to clarify what we mean by private, this doesn't mean it's reserved for private clouds. It means private in the sense that as an organization that consumes it, it's purely within their control. Yes, of course, they can deploy it on a private cloud.

They can deploy it out at the edge, at a cell tower. They can run it in the public cloud. But it's entirely within their control.

So what we try and do with this is we try and turn this into more of an assembly line experience for our customers. They can deploy their AI workloads, manage the lifecycle of them, handle the underlying infrastructure, handle observability, security guardrails, and we basically make that experience as turnkey as possible for customers that want to run infrastructure for AI. I love it.

And talk about the NVIDIA relationship in this. Yeah, sure. So yeah, that's probably the critical bit that I missed.

So on top of SUSE AI Factory, we have an offering called SUSE AI Factory with NVIDIA. Now, this takes NVIDIA AI Enterprise and embeds it deep into that factory experience that I was talking about. So NVIDIA AI Enterprise is NVIDIA's flagship AI software suite.

It's where they bake in all of the latest and greatest capabilities that you see all these regular announcements about. So this is your NIMs, your optimized models, your Nemotron toolkits, frameworks, agentic management solutions, and they make it available as part of NVIDIA AI Enterprise. And what we do is make those capabilities available to customers as part of our solution.

So it's an embedded solution, which is a critical distinction. So we're responsible for how it gets deployed, how it gets lifecycle managed, all of the end-to-end validation, the user experience, and critically, we provide the support for it. Yes, of course, we have a back end to NVIDIA when we need to escalate challenges too, but there's one vendor that our customers need to directly interface with, and that's SUSE.

Got it. Reece, another aspect or a big initiative, let's say, from SUSE is the sovereignty movement. Yeah.

Right? And SUSE has a leading role playing across that. And I noticed in the announcement around SUSE AI Factory with NVIDIA was this notion of private AI across cloud, on-prem.

Mm-hmm. And really that plays into the whole sovereignty thing, right? We want to be able to run AI free from anyone or any entities, any government ability to curtail or within reason, obviously, if we're doing something that's so terribly wrong, but do you know what I mean?

Yeah. The whole reason behind sovereignty, right? A little independence there.

How does this play into that? Yeah. So it absolutely ties into what I was just talking about with regards to private enterprise AI, where you ultimately choose where and how you deploy it.

There was something I said on the keynote stage was, if you can't choose how and where and when you deploy your intelligence, you can never really control it, and you certainly will never own it. Yep. And so, yeah, the whole sovereignty argument is incredibly important, and that is ultimately the message that we're trying to deliver.

Choose where you deploy it, absolutely, but don't ever be restricted in terms of your access to your intelligence. The third-party frontier models, incredible technology, but ultimately, they are not yours to-- You may one day be subject to price increases or export control restrictions, as we saw recently. Yes.

And so, we're trying to enable customers to really take full control over that infrastructure, run intelligence on their terms, within the span of their control, where nobody can ever take that away from them. Rhys, talking about running AI, whether it's for sovereignty reasons or just your private AI across cloud, on-prem, and sovereign environments. I know in addition to the NVIDIA release or announcement, you guys recently announced something with Vultr as well, didn't you?

Yeah, absolutely right. So this again comes down to our core message of offering customers complete freedom of choice when it comes to where and how they run their AI infrastructure. For many of our customers, they're going to run it on premises.

They're going to run it at the edge. They're going to run it right near to the data source. But for many of our customers, they will want to run it on trusted technology ecosystems.

And what we mean by that is public cloud offerings. So we made an announcement with Vultr, at the Raise AI Summit just a couple of weeks ago, where it's now possible to deploy and manage SUSE AI Factory with NVIDIA on Vultr infrastructure. So it's across 32 data centers worldwide, rapidly onboard, use whatever you want to run on top of that infrastructure, pay as you go.

So really, it's about providing flexibility and freedom of choice to our customers. You bring up an issue, and I'll bring it back to our first opening conversation. You come in on a Monday morning these days in the world of AI- Mm-hmm ...

and the world's changed again. So over the weekend, actually, last Friday, I think it was, the folks announced, the people behind Kimi, Moonshot AI- Mm-hmm. Yeah ...

announced the latest version of Kimi, which is almost on par with the frontier models, right? Somewhere- K3, right? Yeah.

They're claiming it's somewhere between Fable and Sol, right? Mm-hmm. I don't know, but the important thing about it is it's what we're calling open-weight model, which means it can be hosted, run independently to a certain extent.

" Or how do you view the whole open-weight model, whether it's Mistral or any of them? Sure. So we are completely supportive of that direction.

So yes, the new models that are coming out, I've always said since day one of me taking responsibility for this role is do not underestimate the power and the innovation that comes out of the open-source community. Sure. And so we're seeing with these models, whether it's Kimi K3 or Mistral or a wide variety of open-source or open-weight models, is that they give, they enable that operational resiliency.

Nobody can take that away from them. You're not subject to third-party access control or the potentially ratcheting up costs. You deploy that within your infrastructure.

You gain predictable costs, and you ultimately own that infrastructure. So I would say open source is a way of, A, giving you access to innovation, but B, freeing you from vendor lock-in and dependency on third parties that you absolutely cannot control. Something I've said about open source-- Actually, I didn't say it.

I heard it said, and I've repeated it for many years, which is the beauty of open source, you have free as in freedom and free as in beer. Absolutely. There's never anything wrong with free beer or freedom.

Rhys, I got to do a little housekeeping here. Mm-hmm. People who want to get more information about SUSE AI Factory and SUSE AI Factory with NVIDIA, what's their best on-ramp?

Yeah. Standard cop-out answer. com.

We have a whole page that's dedicated to our AI solutions. We go into depth around SUSE AI Factory, SUSE AI Factory with NVIDIA. We've got a bunch of demos, data sheets.

I've been doing my best to meet with as many possible people as I can to really share the word. So, yeah. Search SUSE AI Factory.

You're going to find me doing lots of different things. The team is putting out a whole bunch of content, blogs. We're doing deep dive demo videos.

So yeah, there's plenty of content out there. I love it. Hey, Rhys, I realize it's a little later in the day.

Well, not too late your time, but thank you. No, it's almost 3:00. Yeah.

It's still working hours. It's not like you had to forego dinner. But Rhys, I want to thank you for coming on Techstrong TV.

Give my regards to all of our friends at SUSE. We miss them, and I guess we'll be hearing more about next year's SUSE Con soon enough, right? Yeah, absolutely.

I'm sure you will. Mm-hmm. I know the team is hard at working on the presentations and the content and figuring out all of the preparations.

So, yeah. I'm sure we will be hearing a lot more very soon. All righty.

Thanks for coming on. Rhys Oxenham, VP and GM of AI at SUSE here on Techstrong TV. We're going to take a break, but we've got a lot more Techstrong TV coming your way, so stay tuned.