AI & Defense Technology with Anduril CEO Brian Schimpf
50:36 Share

Welcome to No Priors. Today, we're speaking with Brian Schimpf, the co-founder and CEO of Anduril, a next-generation defense technology company. Anduril made early use of AI in its creation of low-cost drones, sensor networks, machine vision, and other systems. We're excited to explore what AI means for the future of defense. Brian, it's great to have you on the show.

Hey, thank you very much for having me on. Well, I would love to start with your personal story and background, how you got interested in technology, how you got interested in defense tech. I know you worked on a variety of robotics and other things. It'd just be great to hear about your, you know, what led you to start this great company.

Absolutely. So I've been someone who's been coding since I was 12, you know, love doing it forever. You know, if I could just have spare time, I'd actually just be coding. I love it. It's great. In college, I ended up working on self-driving cars when that was just getting started. So I worked on DARPA Grand Challenge, Urban Challenge,

We had kind of a small but mighty team at Cornell that was mostly undergrads, just a handful of us. And it was a pretty amazing time to see how fast robotics technology could progress. It went from literally nothing working in like, what was it, 2004, to multiple teams being able to drive through cities in 2007. So just the pace that this technology can move was really impressive to see.

After that, I went out to Silicon Valley, I joined Palantir. I started out as a forward deployed engineer, so I was working with customers all the time. That's where I ended up working a lot with the government in national security space. The thing that was really amazing to me was the degree to which

people in that space really cared about the mission, what they were doing, and the sense of purpose that they had around the problems they were trying to solve and how important they were. And that's just an incredibly motivating aspect of working in this area, combined with it being some of the most technically challenging and hard problems that you can get your hands on. It's a really interesting and unique space to work on. At Palantir, I ended up running product and engineering towards the end, built some really cool products

particularly in the data space around how do you do analytics, how do you make the stuff work at scale, and how do you build modern software in an intelligent way while also doing this in the government space. So we've got a lot of experience there. And then ended up founding Anduril with some really good friends, inclusive of Trey Stevens, Matt Graham. We had met up with Palmer Luckey, who started Oculus, totally interesting guy, like an incredible character, like an incredible human guy.

And he was incredibly passionate about the defense space as well. And so, you know, when he was getting out of Facebook or, you know, as he would insist, I say when he was fired from Facebook, we decided it was time to do this. So we got together and started Anduril around 2017.

And you met your co-founders besides Palmer at Palantir. Is that right? Did you all work together before? That's right. So, you know, Matt and I actually knew each other in college. And then Trey, we met at Palantir and we were all really good friends. And having folks with the background in the defense space is incredibly important. How you sell your

what that looks like, the types of problems they want to work on, how that whole mechanism works is so different than any other sort of enterprise sales and certainly different than any consumer space. So having people who really understand that

mechanism of what does that look like? How do you what is that sales motion? How does this really work? Because that's kind of probably the most important part of doing defense is understanding how you actually get your technology scale deployed and sold, which is like nothing else. It is a very unique and challenging problem. Yeah, I would love to talk more about that later. I think when I first met you, one of the things that really stood out was that because I think I met you all when you were just starting the company.

And I was really excited about defense tech at the time because Google had just shut down Maven, which was there working with the DoD, which I thought was really odd, given that traditionally the technology industry had spent a lot of time working with the defense industry. And so it felt like there was a real opportunity to do a lot in terms of new technologies for the defense world. But the only company in defense I ever ended up actually investing in is Anduril.

You know, there's other companies that are doing things that impact the defense industry, like scale for data labeling or applied intuition or companies like that. But you're the only pure play one that I thought had a shot at actually building something really big. And I think you've had, you know, some of the fastest velocity in terms of government adoption. How did you choose what to build initially? And how did you think about that go-to-market cycle or what the government actually needed relative to these big technology disruptions that were happening at the time of drones and AI? Yeah.

Yeah. So, you know, when we started out, we kind of recognized that the traditional players were incredibly good at making, you know, aircraft carriers and like, you know, exquisite fighter jets. So if you have, you know, sort of fairly unlimited time and unlimited budget, they can make you something very gold plated. And they're actually like, that's really hard stuff. Like they deserve some credit on that. But when it came to, you fast forward 20, 30 years in terms of like, what is the sort of obvious tech that's going to be critical for defense?

Our thesis at the time was this is going to be higher quantities of lower cost systems. And to make that work, you have to drive it with more intelligent software. Like the software component of this is the most critical dimension of it.

And the best people in software were, like you said, they're in Silicon Valley, and they were choosing not to, and explicitly and loudly choosing not to work on these problems with defense. The other part we recognized was to be really successful in moving the technology forward, you had to control the hardware and the software to really be able to move the needle.

The way defense buys at the end of the day is they want to buy a capability. They want to buy something that's like, I have to find ships. That requires me to have drones or satellites with all the software, with all the networking, with everything together that actually solves my problem. So first off, we understood that we kind of had to solve the holistic problem. Second, we know we needed to really bring kind of this modern software approach into the defense space to be able to solve it.

So we

We built up the software technology we call Lattice from day one. It's really about a very defense-specific problem, which is how do I just track and understand everything that's going on in the environment? How do I make sense of the huge volume of sensor information that's being pulled in to find tanks, ships, airplanes, all these things that are very hard to locate? And then we've been able to apply this to a number of different kind of hardware platforms from there. Started out doing more

force protection and base protection and border security. So how do I find people that are on the ground or cars or animals and just classify what it is, what's going on over very wide areas and do this in a really efficient way. We migrated into autonomous, you know, platforms of all kinds. So, you know, we started out with more drones, particularly looking at smaller drones. So things that you can fit into like, you know, kind of briefcase sized drones.

helicopters that we've been able to deploy in a number of different scenarios and bring a lot of smart autonomy to that problem. And then we've been extending this to larger and larger systems as we've gone along. The largest thing we're working on right now is an autonomous underwater submarine that's about the size of a school bus. It's about going to be 60 feet long, 10 feet diameter. It's a very big boat.

And, you know, it'll be able to go huge distances for like weeks at a time. So really kind of complex problem where, you know, you have a real physics limit with things underwater. It turns out you can't communicate with them very well. So you really have to make the autonomy work to make these things go. So we've been able to apply a lot of technologies around how do we just automate a lot of the more mechanical pieces of what people in the defense world do and make it so that they can scale out

these sort of lower cost systems. And we believe that's kind of the critical dimension of what needs to be solved to enable this next generation of technologies to work. How do I make this more scalable, more autonomous, more

to enable me to have the volume of systems that I need. Everyone would like this to be lower cost and still have the same capability. It's just until you solve that problem, you're constrained by manpower. How many humans can I actually deploy to be able to do this? And we have to change that curve to be successful as a country in defense and for Anduril to succeed.

Ryan, it's not totally intuitive to somebody who comes from outside of defense why higher volume of systems is suddenly a need, or is it more the idea that you're placing something that would have been people before? So can you sort of explain that as a thesis?

Yeah. So when you look at how the U.S. has operated, we have very expensive, very exquisite systems that are largely manned. So, you know, you can think aircraft carriers, fighter jets, bombers. These are outrageously expensive platforms that cost, you know, just huge amounts of money to build, to maintain, to train pilots on.

On top of that, we have a pilot shortage, right? We don't have enough pilots to be able to fly all these things. We have recruiting shortfalls consistently. And even more so, we have an industrial capacity issue. Our ability to make...

you know, nuclear submarines is going down. We are producing less nuclear submarines per year, year over year than we have in the past. So we're like 1.7 per year we can produce these submarines. It's not getting better. And the workforce to make these is retiring. It's a very artisanal process. It's hard to train up people. We have very few suppliers in the industrial based into it. So we're not going to outbuild China. That's not an option on these large, you know, very expensive

It's just it's not going to work. It's a bad strategy. The other part of this is you look at what happened in Ukraine, where, you know, a lot of how we thought warfare would progress is kind of playing out. It's almost like surprising the degree to which you don't see even the Russians deploying fighter jets or any of these, you know, kind of exquisite platforms. Like, where is the air campaign? Where is the air war that is kind of in the hallmark of how the U.S. has fought?

The reality is that the risk from surface to air missiles, air defenses is so great at this point that you can't deploy these very expensive systems. The rate of attrition is too high. And so what you see is you want to shift all of the ability to collect intelligence to be able to operate much more autonomously to these lower echelons, to these lower units.

So, you know, the other piece of this is what we've seen and we've predicted is that the ability to know what is going on in the battlefield has gotten very good. Right. So the ability for, you know, the Russians or, you know, the Ukrainians with Western assistance to understand where, you know, large bases are, where the airfields are, where there's, you know, kind of.

setting up and staging to prepare in advance has gotten incredibly good. So you cannot hide in it to a certain extent. So what has happened? You've moved these troops to much more disaggregated small units that essentially become totally uneconomical to do any sort of targeted attack and see the huge battle changing result out of that.

And again, that pushes you down to I need more systems operating at a lower level with units that are operating more autonomously. So this becomes an incredibly hard thing to stop. Right. And you've seen this in Ukraine where it's just the fighting and the way it's played out is much smaller units operating at a much lower level combined with when the U.S. has deployed, you know, kind of the long range strike capabilities that has pushed the Russians back that removed any advantage they had from any of these traditional military strategies because they can't.

up and, you know, kind of close to the front line. It doesn't work anymore. And what was struck the first day was all the Ukrainian airfields. Those were all taken out. All the fixed infrastructure was taken out. So,

So this is sort of the inevitable result of the huge amount of intelligence and information that is now available, not just to militaries, but commercially. All the commercial satellite providers can give you all of this information. So you're kind of entering into this world where it's hard to hide. It's hard to kind of mass troops. And so what you have to do is break up into smaller units and have speed, mobility, and the ability to kind of hide in the noise as your huge advantage.

And these very expensive platforms get lost very quickly. You cannot sustain them. You can't operate that way. It doesn't work anymore. So we've seen this massive shift in terms of how you have to operate. And I think the U.S. is starting to take those lessons away and looking at it for what's going to happen in the Pacific, where to withstand an invading force, to withstand...

And, you know, someone who is trying to take over your territory, it actually really favors the defender now if you can operate in this new way. It would be immensely hard for any sort of successful invasion of Taiwan to actually play out because of what we've learned and what we've seen and how the U.S. has responded and how allies have responded in terms of the types of technology you need.

If a layman thinks about sort of the premise of using sensor data to understand the environment in high volume, low cost systems, that sounds like a very general capability, right? And so did you guys start with a sort of rank ordered list of where you wanted to apply that, you know, defending bases to submarines and sort of how do you begin to attack something as large a market as like, you know, American defense?

Yeah, absolutely. So the interesting thing is like, you know, there's still physics limits to all of this and you only have like five types of sensors. You're really only going to be talking about it. Turns out, you know, you're going to take a picture of something. It's either going to be thermal or it's going to be, you know, visible range. You're going to be able to sense radio emissions. You're going to be able to use radar in a couple of different ways and everything.

That's kind of the list. That's really it. And so the, like, you know, the ways you actually process this information is, is actually quite general, right? Like you can apply, you can build up a set of technologies and techniques, machine learning techniques around how do you detect and recognize different target types in, you know, forest environments, desert environments, all these things. So in a lot of ways, this is a fairly general purpose problem that

technology is very good at solving right so we've kind of built up a toolkit of saying hey given you know this type of target we're looking at we can tailor the specific sensors for like the ranges and types of you know things you need to detect but the algorithms the software and how you process this turned out to be an incredibly general problem set um and and i think this is you know kind of the lesson of modern software is you can build these general purpose platforms which are

incredibly powerful and I think wildly more expensive and hard to build than people anticipate in terms of the amount of infrastructure, tooling, support you need to be able to bring to bear on this. But this is the modern way you build software. Very general purpose technologies that you can apply to a lot of different problem sets.

And we've been able to kind of like weave in all the different aspects of machine learning and how you actually apply those techniques. It's kind of a tool, right? Not really an end in and of itself. We don't do AI research. We're not interested in any of that. We're looking at how can we kind of best take what is working in the state of the art and apply it rapidly into these defense problem sets.

Then from there, the way we think about it is we need to build specific tailored kind of hardware capabilities that then solve specific problem sets. In terms of how we actually work through the defense space, the model we've had that's worked well is I think the thing that's probably not obvious is defense in a lot of ways is quite analyzable.

You can really model out and simulate how different capabilities would play out in a battlefield, given your understanding of your adversary's capabilities, the other capabilities that you have to bring to bear.

And you can really anticipate like, hey, if I had a system that can work at this range against targets of this type, you know, would have to have this sort of resilience on communication, then this would be effective in this way, right? And so you can really start to analyze out where are these gaps? What would the system need to do and predict how well your systems will perform?

The other part of it is really understanding, like, what are those urgent problems that need to be solved? So one of the areas we've worked on and had a lot of success in is counter drone systems. So this was a problem that sort of very rapidly emerged as, you know, you can kind of think of it as historically the U.S. has been one of the few countries that has been able to make like cruise missiles.

But with the low cost of drones now, effectively, Iran and nearly every country has the equivalent of low cost cruise missiles for like $100,000 or less per drone compared to a Tomahawk missile, like over $1 million. They're less capable, they're less good, but they still cause a lot of problems.

And so essentially this air defense problem moved from you're defending against state actors that are sort of like, you know, throwing very high end threats at you to this is a problem that you're going to have everywhere in every conflict. And so the pervasiveness of the problem created a lot of urgency and the pace at which the, you know, kind of people bringing these drones to the fight could innovate was through the roof. This was not like 10 year development programs. They're able to turn new drone technologies in months, right?

And so that requires an entirely different way of building and deploying technology. And one that I think lends itself very well to kind of the software first approach, where

where we can deploy a set of hardware out, you know, radars, cameras, things like that, and then very rapidly iterate on the algorithms and models that you use to detect and classify these threats and rapidly pull in new ways to defeat these drones, be it jamming, be it shooting them down, be it missiles. We can pull all those things together into kind of a combined package to just solve the problem, right? The customers just want to solve the problem.

So we've had a lot of success where there's these areas of just sort of urgent need, particularly where speed is the determinative factor. How fast you can respond and adapt to changes in your adversary is where I think we've done the best so far. So that's kind of how we've guided, you know, where we've invested and where we've put our efforts is, you know, areas where, you know, nobody's building the right tech. It requires this really, you know, software first approach of how do you make these systems cheap and smart and at scale?

And how do you find these problems within there? There's urgency. They need to solve it. And the speed really matters. And that formula has worked incredibly well for us. And I think everyone's realizing you look at SpaceX, you look at all these companies,

What seemed to be only doable on a 20-year, multi-billion-dollar investment is now doable in years. And so the space of problems you can start to attack with this modern approach is really wide open. So we've seen just a huge amount of sort of headroom to grow into on the types of problems we can now go after using this very faceted approach. Absolutely.

How do you think about the degree of autonomy that you build into some of the systems? So for example, you mentioned sort of the bus size autonomous submarines that you folks are building or things like that. How truly autonomous are they? And what do you need to develop from a machine learning or technology basis perspective to be able to drive these systems for the periods of time where there is no operator actively engaged with them?

So the the autonomy in the defense space. So, you know, it's probably best to talk a little bit about the principles of what's actually going to work here. The U.S. is not going to adopt systems that have autonomous robots going out and making sort of lethal decisions about, you know, what to attack or not. That's not really how the U.S. works.

We assign accountability to the person that employed a weapon system. That's how it works, right? We have a real accountable system around employment and use of force. That is a good thing. I like that. I want that to be the case. So when we think about autonomy for these things, it's very much how do we have a very predictable system that the operators can trust and they know what the effect is going to be when they deploy it.

And so in a lot of ways, we've advocated for kind of a crawl, walk, run strategy for this because the challenge with it is the adoption, because the issue is that the operators have to trust it. They have to know what it's going to do when they hit a button. And so a lot of these areas that are more non-deterministic, so things like reinforcement learning or potential applications of LLMs into the space, they are inherently non-deterministic, and that is a risk. And so kind of quantifying that, knowing how to bound it, knowing that it is kind of...

even if non-deterministic, very predictable within a certain bound. That is very, very key to getting the adoption that you want. So what we thought about for this is, you know, how do we have systems that, you know, one, you kind of know what they're going to do within a reasonable degree, like they meet your intent. They can be a little bit unpredictable in how they do it, but within a certain bound. And they...

really are about increasing human agency and control over the problem. So presenting the right information, not too much information, not too little, synthesizing the information in the right ways to aid human decision-making and sort of cognizance over the problem space. And that's sort of the principles we have kind of guiding how we think about applications of autonomy into this space. You know, in terms of, you know, then, you know, take something like an underwater submarine that's going out for weeks, you

You know, that's like a little bit different, right, in terms of where you're going to have human agency in this. And in that case, you need to be able to go out and give it kind of like a mission intent, right? So a lot of the challenge we have is how do you take what is inherently, you know, kind of a human concept of agency?

I want to go out, look for mines that are potentially in this harbor. When you identify the mine, get an image of it and then be able to continue on your search and cover this area and find the most likely places for mines. I was like, that's the type of mission you might want to do. That's kind of a human description. How do you translate that into a machine language that enables it to actually go out and carry out that mission? That's a lot of the challenges that we're sort of wrestling with. It's this really

interesting problem of combining a user interface and a way to specify it with a way to encode that with a way to then execute on it and capture all the contingencies and what-ifs and what could happen and all the types of behaviors you would want to do to enable this to go and then have it work reliably 100% of the time for weeks on end with no hiccups and no issues. And getting that reliability where you need it to be is also a huge challenge for anyone who's worked in this space of

When we looked at self-driving cars, when we looked at any of these things, it's a question of how many nines of reliability are you getting to? And that is where the time, the effort, the investment comes in, in terms of making this go from a research project to it's deployable, it's trustworthy, and something that we could stand by in terms of an operational capability. Yeah, that makes a lot of sense. I mean, you're taking a very thoughtful and structured approach to this to make sure that the primary focus is

which is to augment and help and protect people in the field is really sort of mad. And I guess when you look at things like LLMs,

which is sort of, as you know, sort of the current exciting thing in certain parts of AI. Where do you view the biggest applications for it from a defense perspective? I guess there's all the work that you folks are doing in terms of drones and machine vision and sensor networks and Lattice and sort of understanding situational awareness in the field. And then separate from that, there are things that are perhaps a little bit less related directly to Andro, like certain forms of cyber defense.

And it seems like there may be applications there, for example, directly for LLMs or maybe not. So I'm just sort of curious how you think about where these technologies fit. Yeah. So when I'm kind of looking at LLMs, I see kind of two dimensions of where they're kind of quite useful. So one is kind of a synthesis dimension. So how do I take large amounts of information, kind of capture it into this network and be able to then interrogate it to ask questions about that huge volume of information?

you know, that is a hugely beneficial area, right? Like there's a ton of, let's even just say like current state of the art, just text, right? Like there's a huge amount of text information that the US has access to in terms of intelligence reports, collection of information, all of those things. And so being able to, you know, rapidly synthesize that, be able to ask, you know, kind of questions of, you know, that knowledge base is hugely valuable, right? Now, I think to make that again, really,

operationally deployed, you can't sort of hallucinate about anything. That would be very bad if our, you know, we're hallucinating intelligence about, you know, something that happened in China. That would not be a good situation, right? So kind of proving through all those problems is really key. There's a lot of applications where, you know, kind of synthesis of open source reporting and what's happening on this chat forum, right? You know, those types of things are

incredibly valuable. So the other dimension of this is more creative, right? So the generative part of this, where you are creating content given patterns you've seen in the past. And that's where I think there's a lot of application. The reality in DoD is there isn't like, you know, in a lot of ways,

It's a very operationally oriented community. They are kind of doing things in a very predictable way. You train people to operate a predictable way. You train people to do things in a way that you give an input, you know what the output's probably going to be. And so in a lot of ways, that creative aspect of it isn't really necessarily part of the job. Areas like you've called out, like cyber, are definitely an area where it is.

Uh, we're both on an offensive and defensive case. There's a lot of potential for how do you automate, you know, kind of attack patterns or in the, in the reverse direction, how do you defend against, you know, potential tech patterns, classify different things, you know, that are coming in. I, I think there's a lot of really interesting applicability there. Um,

I think there's a lot of applicability into the autonomy space as well. So what I just described of, I have a human intent of the type of mission I want to create. How do I then translate that into activating the machine to do it? And there's a lot of very cool research I've seen on how do you start integrating these LLMs with APIs and train them to call and invoke these APIs and external systems is, I think, a really interesting space of how do you start to get these to interact with

the real world in an interesting way. And again, there's our philosophy and this is going to be put a human in the loop at the right spot. So you could use the LLM to do a lot of the legwork of crafting this mission plan, structuring it for you and giving you a 95% good baseline to then edit and modify from, but saving you a huge amount of time, training and need to, you know, kind of have that creativity exist in the hands of the warfighters.

You can really amplify their ability to interact with these systems. So we're really interested in areas like that. Again, it's going to come down to a question of like,

Can you get this to a level of reliability and precision that meets what we would expect and what, you know, everyone's going to hold the Department of Defense to in terms of it works at a high enough percentage that we trust it, right? Or how do I put the right human controls in place to get that reliability up where I need it to be? So I think there's a huge amount of exploration on it. And, you know, like everyone's seeing in the space, right?

You've got kind of these foundational models and the infrastructure intact, but actually then the applications, there's just so much, you know, and I think we're going to see a massive tidal wave of new applications coming over the next 18, 24 months. It's going to really open up people to like, what is possible with these? How good can it be? How can you really think about expansively what these technologies can do? We've started doing, you know, kind of a little bit of exploration on that, but it's something we're really excited to drill into more.

Ryan, I want to go to talking about Andral as a business. Can you just give our listeners a sort of sense of, you know, you're about six years in, sort of how many people are you? What can you say about, you know, who you serve and deployment scale?

Yeah. So we are, yep, we're about six years in, we're about 1600 people. We are based in Orange County in Southern California, but we have locations kind of throughout the US and a little bit internationally. In particular, we're doing a lot of work in Australia where, you know, they've been a phenomenal partner and we're really scaling up quickly. And they feel a lot of urgency because, you know, the threats that we're seeing around the world are largely in their backyard. It's a real problem. Yeah.

So, you know, we've been able to kind of scale up both in the U.S. and internationally. We're working with every military service, you know, special operations, Army, Marine Corps, Air Force, Navy. We're working kind of holistically across the whole defense spectrum and working on a wide array of problems. The areas that we've seen, you know, kind of the most mature deployment are, you know, sort of obviously the areas we've started first.

But areas around border security, base protection, we have a pretty extensive footprint there. We have an extensive footprint in counter drone systems, and that's continuing to expand and get more broadly deployed. And then some of the areas that we've invested in more recently, so some of these underwater vehicles, some more of the autonomous drone technologies are kind of just now getting to a point of scale and getting fielded and deployed. The reality with working with a defense customer is

you kind of, you know, you have very few buyers, right? It's like, there's not that many customers you're working with. And so you're kind of subject to their timing and desire for what they're going to fund at what time. So it's this very kind of episodic, you know, where you hit scale points. So you might go from, I've sold 10 a thing a year to all of a sudden I've sold 5,000. And it's this sort of very large jump that you hit, you know, kind of in a...

not easy, clean way, which makes it very hard to scale the company, right? Because you're like, well, you've got to be prepared to manufacture and build these things at scale and support it at scale.

but I don't know which one's going to hit when. And so, you know, you're constantly doing this balancing act of being prepared to hit the scale button, but not overdoing it early, which is a really tricky thing to thread. And I think we've had a huge advantage and, you know, are very appreciative of the amount of funding we've been able to attract because I think we've shown success. We've been able to show we can solve these problems and get it deployed with customers. We can get adoption, but also a recognition of like, you know,

Turns out it's expensive to build and scale these types of hardware technologies. And so we've had a lot of success in attracting funding and getting phenomenal investors who really back the vision of what we're trying to do. And I think that's enabled us to grow and scale with customers where we're able to kind of jump this valley of death of, well, you're a small company with interesting technology, but can you actually produce this at a scale that is meaningful to move the needle for the department on any sort of real conflict? Yeah.

One of the reasons so many people were excited to back you, and you have great funders like Founders Fund and Andreessen Horowitz, General Catalyst, et cetera, is because I believe you were the fastest company since the Korean War to land what's known as a program of record, which often when you look at defense tech companies, to your point, they're doing these small sort of pilots programs.

which in some cases is small, maybe in the millions of dollars, and it looks like real revenue, but then it never really scales or grows. And a program of record is where you become a line item in the DoD budget effectively. And, you know, often those are on the scale of hundreds of millions or billions of dollars. And many of these contracts end up being public, right? Or in the public record. Are there any contracts that you can talk about or just so people get a sense of the scale at which, you know, some of the contracts that you've closed or programs of records that you've won have been? Yeah, so that's...

I think that's kind of like how we view success is this sort of, can we get to scalable programs? And we sort of measure our success by are we getting things to field in volume in a way that this is actually getting out? And so the, you know, the first one we had was with Customs and Border Protection doing border security. I think it was at like two and a half years into the company we had, I think it was like a

million dollar kind of contract to start to deploy that. And then I think maybe about a year or two later, we had landed a counter drone contract with SOCOM for a billion dollars to, you know, kind of scale out and get these technologies deployed. So, you know, these are pretty meaty contracts. You know, the nature of the defense business is you have these sort of very concentrated, large captures you're going after, which is

a bit high stress at times, but it's sort of the nature of what it is. But that's what you have to do to succeed. Now, to actually be able to do those, to convince the government you are a good partner and trustworthy to do this, we've established a lobbying group day one of the company. We were on the hill talking about what we do

why it's beneficial to national security and why it's beneficial to the taxpayer, why this is a more efficient model of how to build and deploy technology into defense. We've had folks who are kind of prior government service come in and be able to know how to thread those conversations, know what this very Byzantine budgeting process actually looks like, how to get these things through the knothole. The budget was being written down

But the budget for 2026 is what's being written today. Right. So it's like that is the crazy reality of the U.S. Defense Department is they do not operate on Silicon Valley time schedules for how to think about budgeting and scaling these things. So you have to kind of really understand how the system works and understand how do you

you know, kind of convince the government that you are a trustworthy partner. And then a big part of that is, you know, I think a lot of people in the tech world think that building the tech is enough, right? Like that, if I have a compelling product, that is what is, what matters. But,

But again, what the Department of Defense buys is a capability and the capability to them is not just the hardware and the software that does the thing. It's the support. It's the training. It's the logistics. It's the manufacturing. It's the, you know, it's all the integrations you got to do with everything else. It's the security, the infrastructure. And so you end up needing to build this like pretty complex organization that has, you know, probably like 50 skill sets to be able to address the holistic set of problems that you need to solve in

for this very unique customer base. And that is a real challenge. These people are hard to find. It's hard to even know what to ask for. It's hard to know who's going to be successful or not and get through all those wickets of compliance, security, all these things. It just takes years. That's just kind of how it goes. And having that knowledge in advance really enabled us

to hit the go button very quickly around being able to solve these problems in a way they can buy it, right? Like you have to make this easy for them to say yes and make it so that they're going to be successful if they take a bet on you. So we're very fortunate to have customers who are willing to take a bet on an unknown company, but we also kind of knew what was necessary to be able to scale and we're very aggressive in getting on top of that early. What do you think is the next sort of big area of technology shift or technology adoption?

in the defense world that's going to be a big sea change. Because I felt like when you came out, there was a very clear why now statement between the rise of drones and the rise of machine vision and ML and AI. And that created, I think, a really big opening for a great company, right? And do you think that there are any similar openings today? Or do you think that was really the big shift and everything going forward is sort of predicated on that change?

So I think that shift is very consequential and we're continuing to play that out. Right. In terms of the, you know, how do I have more autonomous systems kind of writ large, be able to operate more intelligently? I think, you know, there's probably two other major shifts that I think are pretty interesting on the space front. The.

I don't think the U.S. government has fully appreciated the monumental shift that has happened with low cost of launch, how cheap it is to get satellites, like very large constellations of satellites, into low Earth orbit and operate them efficiently at scale. There's a lot of technology overlap with the types of, you know, sort of command control and communications technologies that we've worked on. But just, you know, the...

to launch any given amount of satellite with Starship is going to be so cheap. It changes things so fundamentally that I don't think people have fully realized how monumental a shift that is going to be in terms of the ability to have these mega constellations for the cost of what it was to get like four satellites up before. So that is a monumental shift that I think is maybe 5% of the way through playing out.

The second piece of this is kind of, you know, it's a little unclear if there's sort of a fundamental technology shift or just a kind of

maturing of the, you know, sort of the startup community where things like hypersonics have a lot of actual really meaningful impact to defense. Like it's, you know, got to tailor the applications correctly, but there are multiple companies doing very credible things on, you know, hypersonic air vehicles, hypersonic rockets and missiles that will be wildly cheaper than the state of the art defense. And it's something that, you know, historically you would have said, that's actually like a really good Lockheed problem. They're really good at this high end, very expensive stuff.

But there's companies now that are doing this for a fraction of the cost at five times the pace that will be able to fundamentally shift how the U.S. thinks about how those technologies can be applied, scaled, and built in a way now where it's like, maybe we can make one or two bets on hypersonic missiles, and that's kind of it. It's all we can afford. But that's going to change dramatically. And the intersection of all these things is actually very wild, right? Where you start to say...

proliferated satellites and drones, an ability to consume mass amounts of information, make sense of the whole battle space, and the ability to engage with targets at outrageously long ranges on a very quick timeline, that fundamentally alters the characteristic of how the US fights and operates, where instead of having to get really close, put a carrier group into position and put that huge asset and the thousands of lives at risk,

You start to be able to do this stuff at outrageous distances. Now, it's also very hard to counter where how do you actually then say, well, our adversaries will be able to do the same thing. So what does this actually even look like? In a lot of ways, you know, what we kind of think about with how the future of, you know, sort of warfare plays out from sort of a policy perspective is,

You are trying to create a deterrent effect where you are in a lot of ways preserving the status quo. You are saying like an invasion of Taiwan will not work. It's just not going to happen. Like it will not be successful. So you make it so that it's so clear that the outcome of these conflicts is sort of unwinnable. And you do that through both, you know, kind of a U.S. strategy.

threat of force, but probably more so an allied threat of force. So giving them the ability to withstand and, you know, kind of have the ability to prevent these, you know, sort of invasions and ingress on their sovereignty, that becomes really key. And so we think all those technologies kind of together really enable you to do this in an incredibly resilient way. It's hard to counter is the scary part. And then it pushes countering to a very different, you know, kind of strategy, which is quite complicated. Yeah.

So I guess Tekkens traditionally had a really strong relationship with the DoD. So Hewlett Packard, I think, was started initially on some defense-related projects, or at least they came very early in the life of the company. Cisco, Microsoft, Amazon, all these companies really deeply engaged with the government, with the Defense Department. The internet actually started as a DARPA project, right? And so there's a long history of these things being intertwined.

And more recently, a number of new companies in AI like Applied Intuition, Scale, obviously Anduril, are strongly engaged in defense work. And again, I think that's in sharp contrast to even five, six years ago, where it was quite controversial or at least unpopular thing to do. How do you think this renaissance came about? Like what changed or what shifted? Well, I think Vladimir Putin has a way of changing people's minds about the necessity of defense. And I think the

uh, invasion of Ukraine was, was a pretty large sea change in people's view of this, where prior to that, I think there was a belief that state on state conflict wasn't real. This wasn't going to happen. Nobody was crazy enough to do it. But I think we've seen that if a dictator thinks that force can be successful as a means of getting to their political ends, that is on the table. They will try to use it. Right. And so I think the

recognition that we need to provide the best technology to the US and to US allies, that we need to be able to, you know, kind of solve these problems in a more ethical way, be able to, you know, kind of deter this sort of aggression. That has been a massive shift from what we've seen, where this has become a very clear-cut issue.

I think the other side of it is, you know, you kind of look at U.S. policy and views towards China. And I think we're still in the middle of the shift. But, you know,

has kind of made clear in a lot of ways his intentions. Like, just listen to what he says, right? It is a very aggressive posture. He does not view the U.S. as like we're going to have some, you know, highly friendly relationship. Like, you know, it is a very tricky situation. And he has said repeatedly that use of force is on the table in Taiwan. And so I think, you know, you just do need to listen to what he says, you know, kind of domestically especially.

where they kind of give a different message outward. And so I do think that has been a pretty significant shift in terms of people's view of

we're not at this sort of end of history moment where you know sort of conflict is done you know people say this all the you know they said this before world war one before world war two right you know sort of like economic ties will make this so that force is no longer interesting to anybody it's like it's just it's not true there there are people who will use force to accomplish their political ends if that works and so we have to make it not work uh and so i i think the i

I think that realization has really shifted people in terms of their belief of why this space is important. And I think the other part of this is starting to see, you know, more willingness and engagement from the defense department to work with these new companies. So, you know,

if you had tried to do this 10, 15 years ago, you would have been crazy. Like, it would have had to work. The department was not ready to accept that new ways of operating were necessary. They didn't have the urgency. They didn't feel the pressure to reform. They didn't feel like they were falling behind. And I think that has changed largely because of

The pace at which China has been able to modernize and build new technologies and sort of like their pretty decisive advantage in countering the way that U.S. fights the longer range of their missiles, like all of these things are pretty substantial.

So there was a degree of realization that the U.S. was in fact behind. And so the department has responded and does adapt to these things. To give the Department of Defense credit, they have been operating for over a century and they have substantially changed the way they operate and organize multiple times over. There are very few commercial companies that can say that is true. That is just not actually common. So they do have an ability to change. It is a monster bureaucracy. It's millions of people.

But they do pivot. It is actually incredible that they can retool their organization against these, you know, kind of big shifts. It's not Silicon Valley pace. We wish it was faster, but they do move. And so, you know, we're kind of in the middle of that right now.

And so I think there's also this sort of degree to which you can move the needle now, where 10, 15 years ago, you could not. So I think the combination of those things is a pretty monumental sea change in terms of how the U.S. starts to think about defense. But I really do think that pivotal moment was the invasion of Ukraine and a realization that taking a passive stance is not, it's not, you know, it's not effective anymore. It's not going to work.

Brian, one last question for you. If we're, you know, thinking now about the 2026 budget, like where do we still need more urgency? Like where are we under invested as a nation from a defense perspective or what adversary capabilities do we need to react to that are coming online?

I think the biggest areas that I get nervous about are, it turns out the Pacific is incredibly large. If you overlay the size of the U.S. onto the South China Sea, it's, you know, the U.S. looks quite small. It's an amazingly huge amount of area. And the ability to know what is going on, to sense what's happening over that wide of an area, and to be able to act and respond is a very challenging problem. So that sort of ability to kind of

conduct the military operations we need to do at those ranges while keeping sort of troops safe and presenting that credible deterrent, you know, instead of forcing the administration to make a choice of do you put troops in harm's way, instead saying you have options that

China would take seriously and not have to guess whether it's politically expedient for you to put troops in harm's way, that changes the deterrence calculus tremendously. So making that clear apparent and knowing that we have that ability to reach out and act is huge. And then the second area that I think is undervalued is the degree to which we need to provide capabilities for our allies and that those capabilities are not the same as what the U.S. needs to conduct warfare at very long range.

You don't want necessarily to give Taiwan the ability to strike deep into mainland China. That

Ukrainians would 100% be doing that if we gave them tools to do that into Russia. And I'm not clear that's good, but you want to give them the capabilities to be able to withstand an invasion, to deter it, to make it incredibly hard to make that successful, to protect their coastal areas, protect the fiber optic cables, the shipping lanes, all of those things. So you need them to have those abilities. And that often is a little bit at odds with the ranges and capabilities that the U.S. wants to have.

So figuring out kind of a sophisticated strategy of how do we enable those allied forces, give them capabilities they need, invest in those capabilities while simultaneously having the U.S. have a credible deterrent that, you know, is sort of more politically effective.

believable, is really, really key to, I think, presenting that credible deterrence that, you know, the U.S. will respond either through allies or directly in a way that is not going to be so politically tenuous that they are sort of forced out of, you know, acting. Where else can the tech industry engage or support the DOD or other branches of the government further?

I think a big part of this is things you started to see with Amazon, Microsoft, Google, and being able to provide infrastructure, cloud capability, all of those things into classified spaces and working in the ways the DoD needs to work. I think there's a lot of areas on...

Talking with Flexport about how do you start to help the Defense Department on logistics and be able to do this more efficiently. There's a lot of areas where I think Silicon Valley has figured out efficient ways to operate in a lot of these conventional industries and modernize a lot of these traditional ways of doing business that have direct application into the DoD. It's not easy. It's not straightforward. You've got to really figure out how to...

adapt your products to what they need. But I think that is a huge part of it. The other component is I really believe Silicon Valley does itself a disservice by not engaging in Washington. There's sort of a belief that we should leave us alone. We're going to figure out the tech. Don't regulate us. Let the tech mature, and then we'll figure out the consequences later.

I think we're seeing blowback on that, right? We've seen blowback on crypto. We're seeing blowback potentially on AI, on social media. These things...

Washington does have a vote and rightfully so, right, in terms of how this technology is used, how it impacts people's lives and engaging, telling the story and helping them think, helping Washington think through thoughtful regulation, thoughtful ways to actually manage this and how it impacts the country is incredibly critical. So I think that's the other dimension that people need to take seriously is

spend time in Washington, actually engage with the legislators, educate them, help them understand they want the engagement. And there's a huge amount of benefit people can derive from that in both sides. You know, when I first invested in Anduril, which was at the seed round, it was the first time that I'd ever in my life gotten like hate texts. So it was initially very controversial. And I can't even imagine what you folks dealt with as part of starting a defense company right as Trump was coming in.

And so I was just curious, like, what kept you going during that period? How did you deal with the criticism and unpopularity? And to your point, the second Ukraine happened, the same people started sending me texts about how amazing of a company Andrel is, which was, you know, incredibly disconcerting for me to see how people just flipped, you know, like you should have conviction in what you do, I think. What kept you going?

So I've, for better or for worse, only worked in places that are incredibly controversial. I've worked on working at Anderle. You're working on border security. You're working on defense. You need to have clear conviction about why you are doing what you are doing. And the thing I've learned with this is leadership and conviction matters. We've been clear. We work on weapons. We believe that's important. We believe you can do it ethically. We believe it's necessary. And we're not shy about that. And in fact, the most controversial articles...

have been some of the best ones where they sort of say like, Andrel is doing all this advanced tech to solve these problems. And this is scary stuff. And we're like, this is scary stuff. You need to take it seriously. And that is why we are here. And that is why you want the best people working on this. You just have conviction in what you're doing and why. And if you do it in a responsible, ethical way, that's all you can really ask for.

Yeah, I think that's a great point. I think people really confuse popularity with leadership. So, you know, that point really resonates. Thanks so much for coming on, Brian. This was really awesome. And thank you for your work with Anduril. Thank you. I really appreciate it.