Dan Wright, Armada | theCUBE + NYSE Wired: AI Factories

Hello out those studio connecting Silicon Valley and Wall Street. >> I'm John Furrier co-host of The Cube here with Dave Vellante my co-host. >> Welcome back to The Cube studio here at the New York Stock Exchange. I'm Gemma Allen co-host of NYC Wired AI factories and today we are going to talk about AI factory in a box on the edge. Joining me now to unpack exactly what that means is Don Wright co-founder and CEO of Armata. Welcome Don. >> Hey, thank you. Great to be here. >> So is that your first rodeo at NYC Wired? I know you've been on the show before. >> Yes. >> But when we catch up with folks these days it feels like a century has gone by in just a couple of months. >> It's amazing. >> Things are moving so quick, right? Talk to me a little bit about what's happening at Armata Don. What has been going on since you were last on the show? >> Well, it's crazy. I mean when we started Armata was about 3 and 1/2 years ago there was no Anthropic, there was no XAI, there was none of these, you know, other labs out there. But we knew that there would be a lot more of them and that companies would want to use those models and the biggest constraint was going to be infrastructure and energy. And so what Armata does is we bring these AI data centers, AI factories in a box to the energy wherever it is and enable sovereign AI all over the world. And we're scaling extremely fast. We actually just today announced our 10 megawatt galleon. We call our modular data centers galleons like ships. >> Mhm. >> And it's called Orion. >> Okay. >> And that's been getting a ton of interest. Like inundated with my team with new leads wanting that that product. And we're deploying these all over the world. And then the other thing obviously we're here at NYSE. We're announcing with NativeX Coil the kind of free open exchange where you can invest in compute the same way that you can invest in other commodities like electricity. >> And it's all down to megawatt, right? We had those guys on the show earlier. >> Yes. >> Colin Powell they're going to commodify exactly what that means. What >> That's right. >> Let's talk a little bit about what's happening on the edge. We talk to a lot of companies in this space. I feel like it has grown a lot even since just this time last year. >> Yeah. >> There's a lot, you know, it was a space that was relatively unknown, somewhat sketchy. I don't know if they're sketchy terrain, too. >> But now it seems like it's a very competitive space. There's a lot happening. I think a lot of folks are starting to realize, okay, we're moving from, you know, needing inference in data centers and, you know, in these kind of mass outlets to actually also needing that same level of deterministic to probabilistic on the edge, right? Talk me through some of your customer cases. Like give me some examples of customers that you maybe weren't even working with a year ago. >> Yeah, so we call our model the hyperscaler for the edge, and what we've done is we've continuously redefined the edge. When we started the company, when people talked about the edge, they meant the edge of the cloud provider's networks. >> Mhm. >> But that only covers about 30% of the world. There's 70% of the world where there's oil rigs, there's mines, there's battlefields, there's natural disasters where first responders need to use data from drones in real time to save lives. And that's our focus as a hyperscaler for the edge is bringing it where others won't and filling all the gaps in infrastructure. And just to give you a few examples, we work with, uh, you know, for example, the first responders in Alaska, helping them process data from drones in real time. They used to have 28 hours of latency cuz there were no data centers in Alaska. >> Wow. >> And that's to respond to avalanches and floods. Avalanches and floods don't wait 28 minutes, let alone 28 hours. So you have to be able to respond in the instant. We work with the US Navy. We deployed last year, uh, offshore, running completely on ship's power, air-gapped at the edge. Um, we are now deploying with allies. We were public, um, about some work we did in the Middle East when the Itaviz data center got hit. We were deploying these to enable resilient distributed compute. And then energy, uh, for example, we work with Aker BP in the Nordics and Aramco. And then with our larger form factors like the Leviathans, which are 2 MW a unit, and the Orions, which are 10 MW a unit. Um, you know, a couple of uh real real-time examples, we're actually here today with Foss a fall, which is a big uh company coming out of Norway that wants to do a gigawatt of capacity using mainly hydroelectricity, almost uh all hydroelectricity, by 2030. And so, they're deploying our Leviathans in order to scale up quickly because we can deploy in months versus years with traditional data centers. And then, you know, on the other side uh in Australia, we're deploying with Windy Sea. Last year, they had 7.2 terawatt hours of curtailed energy. They've got tons of wind and solar energy. >> Wow. >> And so, what we did is we just bypassed the grid, which is the bottleneck. We took the the capacity to the energy where it lives to enable AI factories there in the Outback. >> Wow. Let's talk about the tech and what's unique on the real edge, right? The true edge, like these very remote areas. I actually met BG last night on the event. >> Yeah. >> It cool funny the guy, right? >> Yes. Very. >> guy. Um, our Norwegian friend. But, we know that like the edge has its own constraints and its own, I guess, also opportunities in the perspective of lower power, lower compute, right? You don't necessarily need, you know, the same brand of H100 that you need in other places. You can You don't need a Toyota to, you know, drive in a racetrack, whatever it However analogy has been given. Talk me through kind of what you're seeing from the perspective of what that means for like diversification of, you know, the tech stack. >> Yeah. >> Like, what are you seeing You actually work with Nvidia. You work with a lot of those, I'm sure, large chip players. Break it down though. Like, explain to me, you know, what Where How are things trending there? And then, I want to get into talkonomics and obviously, Native X. But, first one I want to What you're selling. >> Yeah. So, first of all, what we sell is is AI in a box, right? So, we do the full stack. We do the hardware. We do the software. We do GPU orchestration. We have a marketplace of all the latest models. Run those air-gapped at the edge. Um, if you've got your Nvidia chips already or your Dell servers, we do kind of all of the integration, deliver it as a turnkey solution, and then we do all the critical infrastructure monitoring and management. So, it's as easy at the as the cloud, but at the edge or what we call the new edge, that 70% of the world that doesn't have the big hyperscale data centers. Um, and, you know, it's actually really good timing because the the chips, every generation they're getting more compact, they're getting more powerful. And so, in a given box, you can just do a lot more. And companies want to do a lot more. A lot of them, like, let's take, uh, you know, Aker BP, uh, kind of cutting-edge energy company in the Nordics. They want to do fully autonomous rigs over the next few years. You're going to need a lot of AI, you're going to need a lot of robotics. They have tons of data on those rigs, but you have to be able to deploy the capacity there locally behind their firewall because they also are very, um, sensitive to IT cyberattacks. They have a hard segregation between IT and OT. And so, we're really good fit there. But, the Gallion family is perfect because we say, "Okay, well, what workloads do you want to run?" And then we have a form factor for whatever you want to do. If you want to do something at massive scale, you know, with the Orion now, you can scale into the hundreds of megawatts. And if it's more of, uh, you know, "Hey, I need maybe a distributed compute, uh, kind of hub-and-spoke solution where I've got some large nodes and some very small nodes just for inference." We can do that, too. >> We've been talking a lot about how you measure performance, how you measure output, right? Native X. >> Yeah. >> Perfect conversation around that because those guys are essentially saying, "It's not GPU per hour. It's like megawatt per unit. It's how how much AI am I getting for my GPU, right?" As opposed to, which is, again, you know, I think the industry is kind of catching up to that, you know. >> But, when we think about the world of the edge, it's actually quite unique because, again, you know, you talked about various form factors there. You're in very difficult terrain, though. You probably have challenges with latency and resilience and performance that, you know, are very unique to the environment and location you're in, right? So, >> Yeah. >> creating that opportunity to connect. But again, when you're when the market has been selling everything in kind of a ubiquitous way, >> Yeah. >> you know, it does kind of create somewhat of a challenge. So, this opportunity with Native X, like, how do you unpack that? You know, we talked to the guys like, does Jensen want this? >> Yeah. >> an interesting answer. >> Absolutely. >> What do you think? >> I mean, I think when you look at it, some of the gigawatt-scale projects are coming up against some headwinds, right? There's a lot of pushback against those. Um but we look at it and say, um and we actually did a a white paper on this last year, uh before a lot of that was happening. There's 6 gigawatts plus of stranded energy just in the US alone. >> Wow. >> And then, you know, we talked about 7.2 hours of uh you know, uh 7.2 trillion hours of curtailed uh you know, energy in Australia. And lots of parts of Europe, it's the same thing. So, there is a lot of power available, but it's about co-locating the compute with the power. And then with Native X, what we're doing is we're actually creating a market that makes these projects uh easier to monetize, right? Which ultimately unlocks project financing, which allows them to move faster, which is very, very important because the US also needs to think about it from a national perspective. We got to win this AI race. It's the one race that you can't lose. And so, I think using our capital markets, and then what our model is doing is finding that energy wherever it lives and sort of being the infrastructure layer in between them. Uh you Jensen's excited about it, but we're excited about it cuz ultimately I think this is a way that we can win. Um and it's like all of the benefits of AI data centers with none of the drawbacks. We use behind-the-meter power, so it doesn't impact taxpayer electricity bills. It's a closed cooling system, so you're not impacting cooling. Um but what you are doing is you're creating a lot of new jobs, and you're also creating a lot of economic growth uh for these states and for the country. >> And when you think about this future world, do we have, you you GPUs a commodity? We have a benchmark that we're measuring against. From the perspective of how that impacts your customers, your clients, it obviously gives them a level of predictability and expectation, right? In what's called like good put. But, it also gives them like a financial predictability. >> Yeah. How and what are you kind of hearing because we hear a lot about, you know, tokenomics, the stresses, the fact that it's somewhat of a wild west, that perhaps people are getting like misaligned products sales, right? Like in terms of what they need versus what they're buying. Like what you know, your your industry's obviously very niche. I'm sure those relationships are extremely tight. What are you seeing and kind of feeling in the market? >> Yeah, what we're seeing is like a lot of companies are interested in using some of the capacity for their own internal use cases. All of them are trying to do more with AI. They're trying to do more with automation. But, then if you can say on top of that, if there's capacity that you're not using, you can immediately monetize it and create new revenue streams, that's very compelling, right? You think about um you know, a lot of energy companies. We work with a lot of energy companies. A lot of them were dipping their foot in the water when it came to Bitcoin, but then obviously there's a lot of volatility that can come with that. The beautiful thing with AI is it's higher margins and you know, the curve is going up and I don't see any end in sight because all of these AI labs and these latest models, they're just hungry for the compute. And so, if you can just make that available and also unleash the public markets, there's almost endless demand. >> Okay, so last question to you. The road ahead. So, you mentioned you came this close to ringing a bell at some point versus Cisco swept in. >> With up dynamics, yes. >> And I'm sure it was a very lucrative opportunity, too. What is the plan here for Amada? I mean, you mentioned you seem to have a huge thumb, some big brands you're already in partnership with. Are we going to see you ring the bell at the NYSE then? >> I mean, I mentioned our mission is to bridge the digital divide and and be that hyper scalar for the edge that is the hyper scalar for the 70% of the world that doesn't have these data centers today. That's a huge vision. And then we fulfill that vision and that mission I think we're going to end up here at some point. So we're looking forward to that. >> Well, please invite us to the party. >> Absolutely. >> Sam Reich, thank you so much for joining us on NYSE Wired. >> Thank you very much for having me. >> I'm Jane Allen here at the Cube studio at the New York Stock Exchange. This is NYSE Wired AI Factories. Thanks for watching.