Dan Wright, Armada — theCUBE + NYSE Wired: AI Factories - Data Centers of the Future

[00:00:00] Speaker ?: Welcome back to The Cube, here at our NYSE studios. [00:00:09] John Furrier: I'm John Furrier, your host of The Cube. This is our AI Factory series, where we're talking to the leaders who are building out the next generation of large-scale systems, networks, and ultimately the infrastructure on the AI side that's going to create a lot of value and also create value extraction for the new kinds of AI native applications. Dan Reitz here, co-founder and CEO of Armonitz, here on The Cube coming in remote from the Bay Area. Great to see you, thanks for coming in. [00:00:33] Dan Reitz: Thanks for having me, great to see you, John. [00:00:34] John Furrier: You know, you guys are doing some really cool stuff over there. You're kind of setting the table. I can kind of see the dots connecting. I was, well, we'll get into it, but explain to what you guys are doing. You're really creating the system for what I would say AI factories, infrastructure-like trend to essentially connectivity all over the world and obviously connecting even from space. you got to deal with SpaceX and Starlink. And so the world's networked. Network is the operating system for these AI factories, but that's just within the data center. When you go outside the data center, you guys are doing some compelling work. Explain what you guys are doing and some of the things you got momentum around. [00:01:14] Dan Reitz: Yeah, the whole idea with Armada and what we're doing is distributed compute. So taking AI factories to the edge, utilizing whatever power is available and co-locating them with the data, as well as autonomous technologies and robotics and drones and other capabilities that people want to deploy now at scale in critical industries like oil and gas, mining, you know, telco, but then also in the public sector, working with the Navy, for example. This is public where we're out there working with them, doing, you know, cutting edge AI, cutting edge autonomy in the middle of the ocean or any other edge location. [00:01:53] John Furrier: How big is the company? Give some stats on just some of the data, how big you guys are, funding and progress on the product. [00:02:02] Dan Reitz: Yeah, so we've raised about $250 million from some of the top investors in the world, Founders Fund, Lux Capital, Valor, many others, as well as Microsoft made a strategic investment in the company and we are partners with them. And we also work very closely with SpaceX and what we had done is every time, you know, SpaceX rolls out in a new geo with Starlink, we're the first mover in terms of the infrastructure because we can deploy our modular AI factories faster, cheaper, more flexibly than anybody else. And in, you know, a few years since we founded the company, we're now well over 300 people and we're growing very fast. A lot of our people are up in Bellevue near Seattle and we were actually just named one of Seattle's top startups. Yeah. But, you know, right now I'm here in the Bay Area and we have an office here in the Presidio and we're all over the world. We even have a presence in UAE. [00:03:01] John Furrier: Awesome. Well, thanks for doing that. One of the things I want to get into is this, you look at the roadmap and Jensen Wong is really good at this at all his events. He lays very transparent about the roadmap to the supply chain, but also for the ecosystem. Obviously AI agents, physical AI is kind of like the three levels. Physical AI, they talk a lot about with robotics. We see robo taxis. You mentioned, you know, SpaceX and Starlink, a lot of these remote and a large scale computing environments, highly accelerated. We are in an era of AI factories, which are essentially new kinds of data centers, purpose-built for the kinds of scale requirements for the tokens, for the intelligence, for the ages, and ultimately for the physical digital kind of convergence of devices. Okay, great. Now, to make all that work, you got to network them, right? Right. You got to have compute. That's obvious. Compute's been checked the box there. We're getting more and more of that, but there's a lot more going on. Explain your vision on this, because I love how you have this distributed computing philosophy applied to the new architecture of factories, because the game is still the same in distributed computing, but the factories are a little bit different. Explain what's going on and what your strategy is. [00:04:20] Dan Reitz: Yeah, our whole strategy is to be the fastest, full-stack infrastructure in the world to deploy anywhere in the world, enabling, just as we were talking about, distributed compute. And the reason is there's a few fundamental things that are happening in the world that are going to continue to accelerate as we go through the rest of this decade and beyond. One is you have connectivity becoming ubiquitous. You know, Starlink didn't exist five years ago. Actually, it launched in public beta in November of 2020. Now it's in well over 150 countries. It's expanding every week, and the service is continuing to get better and better. But as you know, we also have private 5G with technologies like Nokia and Ericsson. We have SD-WAN with Cradle Point and Peplink and others. And so finding the right way to enable perfect connectivity in these different locations, that's optimized both for performance and cost and security has never been easier, and it's only going to continue to get easier. So ubiquitous connectivity, that's one. The second thing is you have data exploding at the edge right now. 80% of data is created outside of, and stored outside of traditional hyperscale data centers. And the question is, how do you process that data efficiently to enable technologies like robotics, like real-time response using drone technology, like, you know, real-world AI at the edge? And then the third thing that's happening is that the AI models are becoming more and more powerful when people are looking at agents. How do I apply agents not just to the back office or to maybe coding some software, but to actually solve important problems on the ground in operations for critical industries, and then for things like defense? And we're enabling those things. And what's unique about Armada is that we can do it, again, faster, better, cheaper, more flexible than anybody else, and we do it full stack. So that it's like a cloud-like experience at the edge. Anybody who can use our tech, anybody can use our technology, and it's as simple as using the cloud. [00:06:28] John Furrier: Yeah, I think you guys are on something really big here, and I want, we're going to be putting out a report, I'm going to be putting out a report in February around a thesis that we have, and I think, I want to get your reaction, and I think you're right on the trend line there, is that the edge will be hyper-converged. And what I mean is that all the wireless protocols, whether it's licensed spectrum or unlicensed spectrum, will collapse in with AI factories as a compute-slash-intelligent device to connect wireless so they can mesh up, and also provide seamless access to anyone coming in, which, you know, is pretty obvious, and that's not too hard to understand. But what that enables with a data-centric model-specific or model-agnostic approach is that once you connect those factories, a lot of new things can happen. So models could move to the edge, new kinds of provisioning of services could be enabled, a whole new set of use cases emerges versus the old classic. You know, move data to the edge node, processes at highly available or high availability. I mean, old data models change, so now you have a new data-centric edge versus, say, a voice split it with data, an old telecom approach. And then also, like, think about, like, how do you bring AI to telecom, to networking, like real AI? So I think the model side of this really changes the game to the networking equivalent. What's your reaction to that, and how would you explain that to someone as kind of the new edge, AI on the edge, if you will? [00:08:08] Dan Reitz: Yeah, I think you're on the money there, and, you know, I'd explain it to somebody in a couple of different ways, which, when you take a step back and think about it, it becomes obvious that this edge, this hyper-converged edge, is inevitable. You know, one is, you look at where the data is, follow the data, and there's more and more sensors everywhere. Whether you're talking about, you know, critical infrastructure, dams and bridges, or even cell towers, pipelines, mining conveyors, you go on and on. All of these things have sensors all over them that are generating massive amounts of data. In the public sector, you have, you know, drones being used as a first responder, for example, we work with a lot of states that are using it to respond to natural disasters, like the fires that we saw in Southern California last year, or, you know, the recent flooding in Alaska. People want to be able to use all of that data that's being generated in real-time, and they also want to be able to take the most powerful AI models and run those in real-time at the edge, and so that requires distributed infrastructure to enable that. And then the other thing is, you've got to think about, well, what is the biggest blocker to AI today? What is the biggest blocker to AI today is the power, and power, energy, is also distributed, and so why wouldn't we utilize all the energy that's available, whether it is, you know, stranded natural gas, or it is excess energy that might be available at cell towers or, you know, different utilities? And so, our whole thesis is, just as you said, that the edge is going to become something that is not an edge thing, it's going to become the whole thing, and that's because, ultimately, that's where all the data and the power is, and it's going to be the most efficient architecture as we move forward. [00:10:04] John Furrier: Yeah, we've got our new studio here in New York Stock Exchange, and I'll tell you that I had some conversations here in New York about all the wasted energy in the buildings. So, I'd envision a Metro factory, and, again, this is distributed computing, you could have many nodes in the network, as long as it's kind of traversing properly and connected properly. This is why I want to ask you about the networking piece, because, you know, if you look at the models, right, I was interviewing what's come out of the robotics series we're doing, is these entrepreneurs were building these robotics for, say, things like life sciences. You know, very used case, a wet lab, I got, I got a robot that has some precision, I'm swapping in and out, you know, some things in that test lab, turns out that this one entrepreneur built a great robot, and turns out he's selling it into other verticals because he used open source software, and what that basically means is, like in that scene in The Matrix, where he says, plugs in and says, upload how to fly a helicopter. And instantly, the domain-specific skills speak to that, so you're seeing the software-hardware relationship now coming to, say, the edge, where if there's a drone, or, say, first responder, or, say, some user retail example, why wouldn't you want to just insert domain expertise in any device, whether it's a robot, or an edge server? This is kind of, this is kind of where the software is going, Dan. [00:11:31] Dan Reitz: A hundred percent, right, and you think about what is exciting about that, I mean, you mentioned healthcare, you know, life sciences, those are some of the most exciting use cases, but then you think about other things that are going on in the world, like an example I like to give people is manufacturing, right? We can't hire enough people and train them to do all of the manufacturing that we need to do in America in order to, you know, de-risk our supply chain versus, you know, foreign adversaries. And so, the only way to actually do that is to do it through robotics. And that is what this technology enables is to think about an army of highly skilled, you know, manufacturing talent that is suddenly at your disposal and works 24/7. That's an exciting thing for, you know, I think both these companies that are deploying this technology, but also for the country. And then other areas, like we work with a lot of top mining companies that are focused on de-risking the supply chain for critical minerals, but similar dynamic there, it's like, how do you hire enough skilled people to be able to produce what we need to produce, given that today, China has 90% market share when it comes to the supply chain for, for example, rare earth magnets that are used to create smartphones and drones and electric vehicles. So, the only way we're going to be able to do that is through advanced AI automation and robotics at the edge in order to scale our capacity and that is what, you know, full-stack, hyper-converged edge, just as you said, enables. [00:13:07] John Furrier: It's interesting that manufacturing example, one thing that's also come up, just to riff on that with you a little bit, is that a lot of the manufacturing lines don't have to be purpose-built for a specific thing. You can have multi-function capabilities in manufacturing, you can actually run a line, change the skew, change the product, you know, within reasons versus the old days of build the factory, line up the robotics, static supply chain, build, so much more versatility. I mean, basically, it points out that your TAM is massive, so, what's your strategy, obviously, I mean, every market is going to have some physical, digital automation, there'll be hardware, software, robotics, there'll be a version of that, whether that's a retail experience for someone walking into a store, computer vision, download my models, connect to my home PC, what's in my shopping cart, oh, it's in aisle five to, you know, autonomous driving. These are real, this is a massive market, what's your strategy, are you guys going for the whole enchilada, are you guys going to come in and sequence to a position, and then kind of take territory, what's the vision? [00:14:18] Dan Reitz: Yeah, yeah, so we have big ambitions, and the idea is to be the hyperscaler for the edge, that's the vision for the company, and the way that we're going to do that is to follow what actually worked for the hyperscalers decades ago, which is, you create an ecosystem around a platform that enables lots of different workflows, different use cases, and you know, the way that we do that is that we have a bunch of partners, whether it's at the connectivity layer, whether it's at the application layer through our marketplace that can plug in to our platform, and then they also make money in this whole equation by adding value for these end customers. And so whether that is, you know, a partner like we just announced a partnership with OpenAI, bringing those models to the edge, or any other model company, or it is a traditional kind of industry specific provider like Halliburton, which we also have a partnership with, this is public. So for example, we want to be the edge partner that brings all of those capabilities to the edge, and that's also true for even the, you know, traditional hyperscalers. So for example, we have a close partnership with Microsoft, we embrace technologies like Azure Stack in their latest models and give them a home and infrastructure where those can run all at the edge. [00:15:41] John Furrier: Explain the full stack approach, I'm looking at your partner strategy here, you got channel partners, which is essentially people in kind of the domain, you got hardware, infrastructure, technology, and then industry, industry, is that the ecosystem? And what's interesting in a technology partner and say, an infrastructure and a hardware partner? [00:16:00] Dan Reitz: Yeah, good question. I mean, with the infrastructure and the hardware partner, it's a pretty simple proposition, which is number one, we bundle their technology with our own, and then distribute it as a end product, you know, end solution, I should say, to the customer. Because ultimately, the people that are working in the factories, the people that are working in the hospitals, the people that are working on the oil rigs, just as an example, they are not IT people, typically, they are people that are just trying to solve a job, do a job better, and what we do is we take this full stack solution, and we give them superpowers. And so that's how we want it to work. The other nice thing with these hardware providers is that we then put that hardware in our marketplace, and it's like one-stop shopping for the edge for our customers. For example, I mentioned that we have a partnership with Starlink. Customer wants any Starlink terminal, any plan, they can go into our marketplace, boom, I want it over here, I want it at these five locations, and we just make it happen. So that is one. The second thing is with the software companies, you know, like, for example, we have a partnership with Halliburton for their landmark applications, they have all of the deep subject matter expertise from decades working in oil and gas. We don't have to reinvent the wheel. Instead, what we do is we make those capabilities available in our marketplace, and in the same way, it's boom, you know, I can, as a customer, make one click, I can run that at the edge natively, and then I get the value of that application, which I already understand because I use it at my more connected sites in these more remote edge sites. And then the same thing is true with these cutting-edge AI model providers, not just the big ones, but also emerging ones that want to create models that solve really high-value use cases at the edge. We create an ecosystem around there, and then we give them low-touch distribution, where the second they publish that model in our marketplace, it's available to all our customers all over the world. [00:18:01] John Furrier: So, since your physical AI kind of game here, whether it's robotics or whether it's whatever industry, oil and gas and others, you're relying on partners in the ecosystem to build those out on your behalf, similar to what AWS did with platform-as-a-service and infrastructure-as-a-service, let the SaaS market build the solution. Is that the right way to think about it? [00:18:23] Dan Reitz: Yeah, that's the default, is that if there is an existing model or application that solves a given problem, we just integrate and we partner. Because, again, we don't want to reinvent the wheel. As you said, there's infinite use cases here that are high-value and that are time-sensitive, and so we want to tackle them as quickly as possible. In certain cases, there's nobody else out there that has a subject matter expertise or that has built the application or the model, and in that case, we will build it ourselves. But we try as much as possible to partner because there are so many of these use cases, and what we want to be is the brain. Whether you're talking about an application or a model or a robot or a drone, we're the brain that makes it all work and coordinates all of this together all over the world. [00:19:13] John Furrier: So, obviously, infrastructure, NVIDIA, big partner, I mean, they're just awesome. Everything you're talking about, it's on their roadmap, physical AI. On the hardware, I noticed you got a sensors, you mentioned sensors earlier. That's highly involved, I mean, autonomous, anything, has sensors. What are some of the hardware areas you're looking at that you're looking to recruit partners on or see traction immediately? And what else is beyond NVIDIA? Obviously, they have a ton of edge going on with the Nokia deal that was announced in D.C. That makes a lot of sense to me. There's a huge national footprint focus in the U.S. Again, that ties nicely. I can tell Lucent, Bell Labs, that checks the box for me. It feels great. NVIDIA loves it. Where are your areas of focus for some of those key hardware partners? What's the new area you need to lock in? [00:20:07] Dan Reitz: Yeah, great question. So, we're already partnering with NVIDIA and Nokia, just to name a couple. But we want to partner with any hardware customer that our companies that we work with, as well as with, you know, the U.S. government and its allies that they're adopting, right? And then increasingly, we're actually being asked, you know, for example, what's the best connected camera by our partners, by our customers? And so, we can also help them when they're looking for something. But they, in many cases, you take a large, you know, oil and gas company or, you know, you take a large manufacturing company, they already have sensors all over the oil rigs, all over the refineries, all over the factories. In the case of the factories, they have cameras that they're using today. What we do is we just plug into those and we integrate with those because that is the lowest friction way to create value for the customer. And what we're about at Armada is really fast time to value and creating so much value that then, you know, we want to go on to the next workload and the next workload and the next workload. And then what you get in a year or, you know, maybe 18 months is you have a connected rig. You have a connected factory because that is ultimately where this becomes game changing for all of these types of companies where you have everything connected. And actually, if you open up our platform, the front page of the platform is what we call our fleet map. And in that, we show you all of your connected assets all over the world and then you can manage them separately, I should say, together in your network operations center, your NOC. And what we see is in the future, just like Jensen says, every company is going to have its own AI factory and every country is going to have its own AI factory. But even down to the site level, you'll be able to see all of your connected sites, you'll be able to manage them centrally from a single pane of glass and Armada is going to be that pane of glass that enables that. [00:22:06] John Furrier: Well, Dan, you and your team, a great vision. I totally agree with this whole network of factories. Distributed computing is a paradigm that you just swap out computer, server with a bunch of servers, large-scale stuff, and connect them to something else that's a factory. And you have a network factory model. And, again, the intelligence and the AI is the unique thing that brings it to the network that we've never seen before. [00:22:34] Dan Reitz: Yeah, well, thank you. Yeah, we totally see things the same way. I think that it's going to be amazing and it's going to blow people's minds where the world is headed over the next few years. And we're really excited to see it happen. [00:22:45] John Furrier: Final question for you. Obviously, in the cloud game, again, this is my opinion, but I'll just say it. You know, the telco guys really dropped the ball on the cloud. They could have extracted a lot of value there. Instead, Apple and Google marketplaces sucked all that value out. I think with the factories, if you own the networks and have the data like a telco or a carrier, you can really make the edge work. And Starlink, Prometheus at AWS, these are things coming, terrestrial and landline, Ethernet, with wireless spectrum and unlicensed spectrum. There's a huge opportunity there for value creation extraction for networking operators. What's your reaction to that? [00:23:31] Dan Reitz: I totally agree. And I'll take it one step further, which is I think it's almost a mandate for telcos at this point. Because you probably saw recently Starlink just, you know, purchased $18 billion of spectrum from Echo Star. You know, five years from now, they're going to have, you know, connectivity everywhere that's as good as fiber in the ground. Well, what do you do if you're a telco? You utilize the assets that you have. You already have a very established presence in these markets around the world. You already have distributed power. What is the best thing to do? Enable distributed compute. Enable distributed AI factories and become a provider of intelligence to all of these different parts of the world and critical industries. And so I think we're seeing that, I think, as we go through the next few months leading up to Mobile World Congress and then throughout 2026, you're going to see that continue to accelerate. Because whoever makes that transition the fastest has a huge, huge opportunity and whoever doesn't is at risk. [00:24:34] John Furrier: I mean, the hyperconverged network brings up new services. Think about the corporate services they could provide through license spectrum. You know, I'm on a VPN. I want to have an app. I'm in public, public, private, hype, you know, multi-domain, you know, multi-tenant, large-scale clouds coming. Sovereignty. [00:24:56] Dan Reitz: Sovereignty. I mean, it's happening all over the world. People don't want just AI factories. They want distributed AI factories. And even more than that, they want sovereign distributed AI factories. So that is a huge opportunity and one that I think, you know, Telco is really well situated. [00:25:10] John Furrier: Well, you got a lot of great investors, Dan. And this won't be our last conversation. I think what you got is at least a good five-year head start, in my opinion, from everybody else. So congratulations and thanks for coming on our AI Factory series. [00:25:23] Dan Reitz: Thanks, John. I really enjoyed it. And yeah, looking forward to many more conversations. [00:25:27] John Furrier: I'm Jeff over at the Cube. This is our AI Factory series. Again, it's just the beginning, scratching the surface. Jensen said it's going to be a complete reset of the semiconductor industries. The growth will be 10x. It's going to grow significantly. As the AI infrastructure continues to accelerate, the next waves of scale kicks in. And again, edge, core, cloud, space, networking is going to be a big part of it. So we'll keep you covered here on the Cube. Thanks for watching. Thanks for watching.