Shield AI raises $2B at $12.7B valuation and acquires simulation leader Echelon Technologies to accelerate AI pilots

Speaker 1: I'm rooting for American open source AI. Same. Play that eagle sound. I'm happy about it. Good luck to the folks over. Let's bring in NVIDIA. Ryan. Let's bring in Shield AI. Ryan, how are you doing?

Speaker 6: Doing well. How are you?

Speaker 1: We're great. It sounds like you're really, really great, though, because you have some big news for us. Tell us what happened.

Speaker 6: We do have some big news. We just closed our latest financing, raised $2,000,000,000 at a $12,700,000,000 night. Hit it again. Hit it again. Hit it again.

Speaker 2: There we go. Two heads, one for each billion. There. Wait.

Speaker 1: Did you upsize the round? Have on my sheet 1,500,000,000.0. It seems like the That is a preferred equity component. So guess, technically, you

Speaker 6: can call it 1,500,000,000.0. Okay. Cool. And then as part of the round, completed an acquisition. Really? You know, the a company called Echelon Technologies, the leader in physics based simulation Mhmm. For, you know, high end aviation training

Speaker 1: Mhmm.

Speaker 6: And something that's integral, I think, to the future of AI pilots and autonomy.

Speaker 1: Yeah. So, I mean, with that acquisition, sort of reintroduced the shape of the company software, where are the integration points in the hardware, what do you build, what do you partner on, who are you selling to, is it all government, is there some commercial? I imagine the business has grown. What's the shape right now?

Speaker 6: Yeah. So we sell to The United States and friends of The United States. Okay. It's all all government, almost entirely Mhmm. Military. If you're not familiar with the company, if your viewers aren't aren't familiar with the company, we have two the the company's focused on protecting service members and civilians with intelligence systems. You had my brother on here very early when he started. He's a

Speaker 1: He's great.

Speaker 6: Navy seal. We started the company together. The mission is very near and dear to us. The way that we get after the mission is is twofold. Number one, building advanced aircraft. Mhmm. And number two, building AI pilots or advanced autonomy.

Speaker 2: Yeah. What were you doing while while he was

Speaker 1: In the Navy SEALs.

Speaker 2: Were you were you on the startup path always?

Speaker 6: Yeah. I was doing nerdy things, working at Qualcomm. I had started a company that did wireless charging. I sold that company to Qualcomm

Speaker 4: Cool.

Speaker 6: And then led the wireless charging efforts at Qualcomm. Then my brother, I was getting out of the SEAL teams and thinking what about what he wanted to do next. And long story short, he came into me and said, wanna bring the breast of what's the best of what's going on in the autonomous driving sector to the mission of protecting service members and civilians. His mission was clearing buildings of threats. It was a mission type that had killed more service members and civilians than any other mission type in the preceding twenty or so years. He felt like if we could take the best of AI and autonomy, bring it to the mission of protecting service members and civilians, it would be a really impactful thing for the world. Yeah. I thought it was an inspirational mission. I did think it was a stupid business. But my brother's a persistent person, and he he showed me the line, and I'm glad that he did. So

Speaker 1: Yeah. Take me through the state of autonomy in the flying vehicles that the military operates because there's so many different vehicles from helicopters to fighter jets to cargo planes. Then if I'm correct, the different branches have different versions of the same air airframe sometimes. And so are you selling one piece of software to multiple branches of the military? Are there differences between what you're selling? Are you focused on a particular area within the different airframes that are in service right now? Like where's

Speaker 6: the Good biggest question. So the state of the autonomy, I'll say like two components. Number one, from the market side, I think that everybody has reached a point where it's undeniable to the customer base that autonomy is one of the most important things in the future of security and stability. That wasn't the case ten years ago, but I'd say kind of at this point everybody's fully bought in on the need for the capability. With communications and GPS degraded, you need systems that are going to be able to see, think, and act at the edge even if they can't reach back and communicate effectively Yeah. With with the pilot on the ground. The state of the technology is is moving incredibly fast. You can build very advanced capabilities. We see, I would say very advanced autonomy deployed on the battlefield. Certainly we have very advanced autonomy. What we haven't yet seen is very large scale deployment of autonomous systems. And the focus of Shield AI when it comes to the software side is fast forwarding the proliferation of advanced autonomous systems within The United States and friends of The United States to protect service members and civilians at scale. The hard thing about AI and autonomy when it comes to aviation and frankly any weapon system is sitting at the intersection of enabling very high performance, achieving very high levels of assurance, and supporting very fast development cycles. Traditionally in this business, it's been pick maybe one or if you're lucky, of those three things. You achieve high performance, but it took you a really long time. Or it might not be something that you can certify as airworthy or certify it in a weapon system because, you know, there's just too many things that could go wrong. Or you get a very high level of assurance, but it's just really stupid autonomy and it took you a long time to build. Yeah. And so everything that Shield AI has focused on building is basically the industrialization of autonomy to fast forward the proliferation of autonomy by creating basically, you know, pipelines that enable developers and countries and companies to deliver high performance, high assurance, autonomy at the speed of relevance.

Speaker 1: I have a I have a few sort of random scattershot questions to help me understand the state of autonomy in military equipment. Do tanks have lane keep assist? Does cruise control exist on a on a, you know, c one thirty? Like, are are there are are are is is America taking a walk, crawl, run system or are we going to jump straight into, like, AI dogfighting? Because that feels like the Mount Everest of the autonomy challenge versus Yeah. Just like you're tired and your tank is rolling down the the the highway and you don't wanna bump into the person in front of you.

Speaker 6: Yeah. Well, one of the amazing things about the US military and militaries in general is the the spectrum of capability that's in the inventory. There are literally planes that are 80 years old and still fly. Yeah. And then you have stuff that's that's fresh off the line. So you have the full spectrum of stuff that has the equivalent of lane keep assist Mhmm. To the stuff that is fully manual.

Speaker 1: Sure.

Speaker 6: A lot of the focus is on bringing advanced autonomy to the newest systems because that's sort of the easiest to cut in.

Speaker 1: That makes sense.

Speaker 6: And with the amount of technological change and the huge step up in volumes of systems being deployed Mhmm. In places around the world. There's just like ample opportunity to cut that software in as new things are coming off the line. We are starting to see some interest in sort of backwards application of advanced autonomy to legacy systems because certainly things are around for a long time And the more advanced autonomy you can apply kind of throughout your force structure, the more advantage you're going to have. And so certainly we see developments in that area as well. Talk

Speaker 1: to me about how your progress tracks with the AI progress in the big labs that we see news from every day. Like the models of the big labs, the LLM companies are getting bigger. They're GPU constrained. They're data constrained. Is that are the advances in the more like consumer focused, coding focused AI labs, is that transferring and accelerating your progress either because you're reading research papers like Google put out one about optimizing KV caches and I could imagine that applies to some The transformer paper might apply to your business. Or at least are you getting benefits from like, yeah, like we're able to use coding models to advance what the software that we're writing at Shield AI?

Speaker 6: The benefits are across the full stack. Okay. From development to deployed edge systems. The thing that we do is focus on enabling the intersection of high performance and high assurance at So the same you know it's like these models that are used throughout the consumer space and the amount of progress, see very high levels of performance, you see very fast iteration speed, and then you have things like hallucinations that can be counterproductive. Which is okay if you're writing an essay and make some stuff up. But if you're trying to disambiguate between blue force and red force and you're going to make you know, consequential decisions Yeah. Those things are are tougher to deal with.

Speaker 4: Or if

Speaker 6: it's gonna result in your airplane flying into the dirt, tougher to deal with. And so we absolutely use those things and we bring it together with frameworks that impose very high levels of assurance

Speaker 4: Yeah.

Speaker 6: On the decision making of the algorithms.

Speaker 2: What how are how are the various like Shahed systems? Are they are are they primarily what what do we know about them? Are they primarily remote piloted? Are they using are is there any element that's actually autonomous or are they just kind of picking a spot on the map Yeah. And sending it off and kind of just saying like, hopefully we hit our target, at least we'll create some chaos. Like, what Yeah.

Speaker 6: Good good good question. So typically, are GPS guided munitions. So they they pick a spot on the map and they send it. And then it's got you know, they they can have some terminal guidance that helps it, you know, like hit something of of value. Terminal

Speaker 2: what is terminal terminal Sorry.

Speaker 6: Terminal guidance is just to make sure that, you know, a weapon is tracking toward something of interest and and so someone And so

Speaker 2: if if one of those goes into us into an area where there's gamming or denial Is it just gonna, like, still do its best to hit the original target? Like, how what do we

Speaker 6: It it depends on how sophisticated it is. I would say that, like, Shahed's by and large and and and our GPS guided munitions. But some munitions will have cameras that like so it doesn't fly into blank pavement. It might be able to see like a truck or something else that would look like a valid military target. They just make small course corrections that enables like a a weapon to hit that instead of an empty parking lot would be the type of corrections that can be made during terminal guidance, as opposed to, like, the the long transit, portion of the mission. But, you know, those types of systems and the same so the same challenge exists in The United States when if you're if you're just picking targets off of maps without, like, good intelligence and good updates and the ability to provide, course corrections based on, you know, what you you're you're seeing on the edge, you can end up, you know, wasting a lot of munitions with little effect.

Speaker 2: Mhmm. What's been your reaction to the news out of Barksdale, Louisiana? I've seen a bunch of reporting this week around, some unknown drones flying around the airbase. I don't know how much you can share there, but, like, this seems like something where an opportunity to assess some enemy's capability, but it feels like there's drones are like, the thing I've been processing is, like, it's it seems like, you know, historically, you built like an exquisite system like an f 35 and you have a pilot in it, you can't exactly go and test it in in an environment in an enemy environment very easily because you'll, you know, risk starting a conflict or losing the the plane, all these different things. Whereas with inexpensive drones, like, why why would an adversary not just send them out around the world to start testing their capabilities in, a live setting because worst case scenario, you lose the asset. But, you you know, it's it's not super consequential when you look at these overall budgets. So

Speaker 6: Yeah. Well, good good good question. One, I'm I'm actually not tracking the latest news in in Louisiana. I'll go look it right after the fact, but I can tell you counter UAS is a major problem for governments globally. These things can be hard to deal with and if you engage them kinetically, the question of collateral damage becomes one that is an important consideration. The small drones that can harass military bases, yes, are like, you know, other countries can do the same with their largest pieces of hardware. And that's why you see the Chinese like running major military exercises around Taiwan. You know, they could stick to quadcopters, but they choose to use the full spectrum of their capabilities to intimidate and shape the environment. But you did call out something that I think is also very important, which it can be hard to fly your capabilities around and understand their true potential. When you think about some of the security challenges in The Pacific, the distances are so vast. And then if you look at the range space in America, it's relatively confined. And so how do you rehearse and think through some of the problem sets when your range space is way smaller than the theater? And that's where capabilities like Echelon, the simulation environment come into play. That's

Speaker 2: your new acquisition?

Speaker 6: That's the new acquisition. And so there's a there's a program called the Joint Simulation Environment that started to get after the problem of we can't rehearse the way that we used to because the domains are so large. Once people had this realization, they're like, great, well, maybe we'll try a simulation. And what they found in simulation is that there was stove piping across vendors. Every vendors had their own simulation. Every vehicle platform had its own simulator. And then if you were in space or on the sea or in the air, you had a different simulator. And then by the way, across countries, you had different simulation systems. And so Echelon became the leader in the joint simulation environment, which is one of the reasons we're so excited about it because it is the only multinational, multi domain, multi vendor, and there's one more multi in there, in the world. And it creates this unbelievable foundation to enable not just what it does today, which is training and rehearsal against like high end threats and complex scenarios, but it creates the opportunity to introduce, well, do autonomous systems do in that same sort of environment? So the things like collaborative combat aircraft or advanced weapons, those things get pulled into the environment. And now you can see, well, this is how humans and AI are able to work together in these scenarios to inform the designs of those things. And if you continue to play that forward and you see you know, it's belief that in every vertical, whether it's A and D, medicine, like pick pick your thing, vertical leaders will show up when they figure out how to close the feedback loops between, you know, the data aggregation and generation through training, through deployment, and close that loop in their domain. And one of things we're very excited to do with Echelon and Shield AI together, we're the leader in in AI pilots. You might have seen the United States Air Force announcement that Shield is one of the two companies selected to build AI pilots with the collaborative combat aircraft. And so we have some great work on that front. And now by taking the simulation leader across services and domains, we can start substantially scaling that simulation environment, augmenting all of the real world data and start closing that loop to build extremely high performance AI pilots that we're able to test against the authoritative threat models with the authoritative blue force models and also train US and allied forces to fight effectively with those AI pilots as part of their formation. So we're very excited about it in the short term and on a decadal basis.

Speaker 1: Decadal basis. I like that term. Yeah. Decade thinking in decades literally. I've never heard

Speaker 2: that before. What what is the actual interface what are the like the interface of Echelon Yeah. Look like? Is it is this like am I thinking like flight simulator type of thing or is it more, you know, you're just kind of running a simulation and seeing seeing

Speaker 6: Yeah. So so so you can think of it a very simple view would be like a flight simulator. Mhmm. But, you know, with a proper as you've seen, like, a flight simulation dome. Yeah.

Speaker 2: Yeah. Yeah.

Speaker 6: Very very realistic physics, very realistic sensor models to include radar, night vision, EOIR, or regular light cameras, appropriate physics, appropriate red forces with, like, not just, like video game style, but these are the actual capabilities across like the electromagnetic and visible spectrum that really enables very advanced training and understanding of performance.

Speaker 2: Yep. Fascinating.

Speaker 1: Well, thank you so much for taking the time on a busy launch day. Congrats on the progress and have a great rest of your

Speaker 2: I'm really glad you guys are building this company.

Speaker 1: Yeah. This feels extremely Yeah.

Speaker 6: I appreciate it. Thank you. Love the Gong, by the way. Thanks to the the two Gong strikes.

Speaker 1: Of course.

Speaker 2: Of

Speaker 1: It's massive progress. Yeah.

Speaker 2: Great to have you on, Ryan.

Speaker 1: We're we're we're glad and and thankful for everything that you're doing with the with the rest of the team. So thank

Speaker 2: you. Alright.

Speaker 1: Day. Let me tell you about Figma. Agents, meet the canvas. Your AI agents can now create and modify your Figma files with design system context in beta starting today. Go check it out.

Speaker 2: Bill Ackman is saying he's likely down 20% year to date now. He was, I believe, took a big position in Meta not too long ago.

Speaker 1: He had some Or takes. I don't know. It might might pencil out. I don't know.

Speaker 2: Yeah. I I think it's it's still I'm still like

Speaker 1: people are.

Speaker 2: It'd be silly to be bearish on Meta over the long run, but certainly

Speaker 1: Capital bloke gave some more context on what's happening in the market. He said the market is just a guy staring at two screens. One has Truth Social. The other is Anthropics blog. In front of him are five buttons that say software, semis, European defense, energy, and gold. Trump or Anthropic post and he hits a button to make those stocks move 5% up or 5% down. It is it does feel like that's the nature of the market these days. Signal had a funny screenshot here. Someone built a or assumingly vibe coded an app that turns any TV into a retro flip flap split flap display. These are very cool. You've seen these. They flip around. You see them in old airport terminals for when the flights are leaving. Wayne chimes in and says, I'm very likely gonna build this with Claude Co. This afternoon and post a link to a free download to this thread because this is absolutely ridiculous to to to suggest that someone should pay a $199 for something that probably took about eighteen minutes to make. And Yash, the creator says, I will look for your post. And exactly one minute later, open source my tool to render your efforts a total waste. Try me. And so there's a little bit of a of a standoff in the vibe coding world. I thought that was very very funny. Is there anything else you wanna

Speaker 2: Lastly, some text from q one of last year. Mark Zuckerberg texts Elon. Looks like Doge is making progress. I've got our teams on alert to take down content doxxing or threatening the people on your team. Let me know if there's anything else I can do to help. Elon reacted heart.

Speaker 1: Mhmm.

Speaker 2: And he says, are you open to the idea of bidding on the OpenAI IP with me and some others? Zach says, wanna discuss live?

Speaker 1: Oh, interesting.

Speaker 2: And Matthew Zeitlin, former guest says, you can tell who is used to antitrust investigations.

Speaker 1: Well, stay safe out there. If you're planning on bidding for some IP and you're two of them, you're chatting with the richest man in the world, yeah. Maybe stick to a phone call. Who knows? Well, good luck out there

Speaker 2: and It will be an interesting weekend Yeah. Of news. The markets are now closed which means

Speaker 1: We can go to where do we go on the weekends? To the beach.

Speaker 2: Well, was gonna say it means that you wanna be refreshing Truth Social. Yes. You wanna stay up to date.

Speaker 1: And it's it's red. We closed way down. The S and P closed down 1.67.

Speaker 2: Good opportunity to touch grass. Yes. But next Monday, we've got Take Him, Logan Bartlett. We got Cody Ko joining. We got going Cody Ko going back to back with Brett Adcock.

Speaker 1: Let's go. The You wanted

Speaker 2: Never before on a podcast Never before on a have we had Cody Ko and Brett Adcock back to

Speaker 1: I'm very excited. I've been a huge fan of Cody Ko for a long time. I watched a ton of his YouTube videos.

Speaker 2: Yep. It

Speaker 1: will be very fun. Anyways Leave us five hope you have. Podcast, Spotify, sign up for newsletter, tbpn.com. We will see see Monday. Monday.

Speaker 4: Flashback. Goodbye.

Speaker 1: Flashback.