The Exchanges

Q is like shock to the system. So it sounds like we haven't kept up, but I think there's, there's also something you've talked about, which is that the people who were involved several decades ago are also starting to retire. So what does that look like? How many companies are actually involved in this space of advanced manufacturing? And what are we seeing there in terms of that almost lapsing?

A Yeah. So there's something like 20 to 30,000 small businesses that make up the majority of the defense industrial base. And I think just to frame this before I go into the detail, basically, if you're making a fighter jet, you know, you're actually outsourcing most of the assemblies like a wing or an engine to another company like Pratt & Whitney. And then Pratt & Whitney has a network of, you know, 2000 suppliers that give them the components and they might do some engineering. And then, so what these companies are actually doing is taking a lot of, taking the Lego kits and assembling them and then throwing the assembly up another layer and then Lockheed finally stitches the whole thing together to, to give an example. So there's about Um, you know, we have, um, 30,000 or so of these small manufacturers that in aggregate make up the entire defense industrial base, but individually are like small businesses where maybe they have 10 to twenty million in revenue max, and there's maybe 15 to 20 employees. have, And the danger there is, um, because we outsource manufacturing and because we culturally decided that if you don't have a college degree, you're worthless to society. There's been no new entrance to the manufacturing workforce because it's not sexy and it's not rewarded in the culture. And because we've had no shocks to the system, no one's realized what a big problem it i…

AI assessment note: “something like 20 to 30,000 small businesses that make up the majority”

Q is a frivolous industry. Um, I'll just read you one tweet for fun, uh, from someone that I saw recently that said, no offense, but I a hundred percent think space colonization is a childish desire. So feel free to respond to that particular tweet. But really what I want to get at is within the sphere of things that you could pursue within Hadrian. Why specifically space to start?

A The really short version is like Stripe doesn't get to sell to AT&T until they sell to a bunch of white combination startups. There's just a bunch of huge amounts of capital flowing into commercial space, which means there's a bunch of net new spend and they're being run by thirty-year-olds, not fifty-year-olds, which You know, you want to sell the startups at first and you want to hit that early adult spectrum. So commercial space has all of the complexity of every other industry that we would want to serve in the future. So the automation we're building for commercial space is almost completely transferable, but there's a real need because these companies are trying to go super fast and they will pay for speed. You know, if they are sick of calling Bob's machine shop, but they're young enough from, you know, all these rocket and satellite companies that they're willing to take a shot on something new and really work with us to develop the system versus, so it's just your classic early adoptive problem. Then I think holistically is like, why go to space, uh, at all? Like why put humanity's resources out there? And I think, you know, we could go through this argument of like, you know, space is a war fighting domain. Like you need observability of the planet to stop nuclear launches. So you need satellites, you know, like GPS is a great example. We can go through all like the t…

AI assessment note: “There's just a bunch of huge amounts of capital flowing into commercial space”

Q I think it's, it's interesting that this term advanced manufacturing covers an industry which at least isn't always that advanced or maybe hasn't kept up with the time. So could you give a couple examples of ways that this space of advanced manufacturing is maybe not the reality that people might expect?

A On the defense side is a great example. Um, you know, everyone sees Defense primes coming out with drone programs or, you know, advanced fighter jets and stuff and, you know, really flashy products and marketing and things. Wow. Like everything, everything works perfectly. Um, under the hood, once you get below the assembly level, you know, someone making a fighter jet, really what they're doing is outsourcing all of the components, but the wing, the engine, you know, everything to a tiered layer of, you know, It's like, um, you know, we have, uh, 10 mini primes, then 20, you know, tier three suppliers, and then literally 10,000 small suppliers dotted across the country that are doing everything from a machine component to, I don't know, the circuit board that is one part of the fire control system or whatever. um, have, And that entire layer is basically, basically chaos. You know, the B two bomber, which is one of our most strategic assets in terms You know, the nuclear program about a year ago, the government had to issue an RFQ for some of the parts on the B to not just to replace the parts, but because no one, you know, the guy that designed it or engineered it retired and they didn't have it documented anywhere. Like there's literally no, you know, manual or document of how to make that part. So the government had to go out and say, Hey, we need someone to come up and lik…

AI assessment note: “On the defense side is a great example.”

Q don't I just acquire them? Why don't I do like a private equity play and just acquire these shops instead of doing what you've done now, which is more so build the technology, train individuals to, to service, um, them are almost like absorb that That intelligence. So can you tell us a little bit more about why you pivoted there and why you didn't go with that original approach?

A One of the things that, you know, you don't really learn until you're in the weeds is every, every piece of hardware variability, whether that's a different machine or a different cutting tool or a different vice or piece of work holding every new combination of those things that you add on to the kind of like the search space of the problem. Has a nonlinear impact on the software engineering complexity that you need to build to like, you know, get to that, get to a level of automation that actually solves the problem versus just, Hey, you're 20% better than your competitors. So you have a great private equity rollup. It wasn't until I was in the weeds of that problem that I realized that most legacy machine shops have one of every machine ever made. And even within, you know, they have 20 machines, five of them might be the same, but even those five are set up a completely different way. And that adds this In enormous nonlinear complexity in terms of the software automation that you have to build on top of those systems to the point where it's, it's close enough to impossible that you can call it impossible and you can squeeze out some margin improvement, but you would never ever, ever get it to the point where it was a scalable system or it was repeatable enough that it was, you know, going to solve the problem at the kind of multi-billion dollar level that I want to solve th…

AI assessment note: “every piece of hardware variability... has a nonlinear impact on the software engineering complexity”

Q place because, I mean, maybe I'm wrong about this, but I imagine they, they might feel like their jobs are being automated away or that they're being replaced in some way. And of course, some of them, it sounds like they're planning to retire, but how do you actually convince the person who has all this knowledge to, to want to share it and be part of this new system?

A I think there's a couple of things. The first one is the smartest and best people know that the industry is due for a transformation and that it's inevitable that someone's going to come along and do this and they want to be part of the winning team, not, not left behind. So that's a big part of it. Yeah. Incentives really matter. And I think. Culturally in manufacturing, you don't get many opportunities for growth. And there's a bunch of really smart people out there who've been kind of siloed and not given the opportunity to stretch and grow. And I think Culturally, they understand that if we're automating one part of the factory, we're not going to suddenly like fire 20 people. Like, no, no, no. Go pick whatever else you want to do. You know, manufacturing, engineering, or CMM, or a higher skill level of CAM programming. Um, and I think the other thing is, yeah, incentives really matter, and I think we're probably the only factory on the planet outside of SpaceX that is giving everyone in these roles, you know, the same Uh, tier of equity as the best software engineers have, you know, uh, every technician in Hadrian, whether you're packing boxes or whether you're an absolute master of your trade and maybe you, you know, you're one of the only 20 people left that can do it is being given the opportunity to have, you know, probably the only opportunity they've ever had to crea…

AI assessment note: “you're an owner of the company, and then you're highly incentivized to contribute”

Q I was going to say how many people fit that bill? Three?

A Handfuls. Yeah. Handfuls. Um, the factory talent is, yes, we always need more people, but there's no like one critical talent base where we're really, really struggling. I think in terms of automation, yeah, we just need all hands on deck. And I think what people don't realize that we struggle with sometimes in recruiting or at least have to Over message, you know, to make sure people understand is that you don't need a hardware or manufacturing background to come into a business like Hadrian or SpaceX or Anduril to be able to, you know, be a competitive candidate or create value, you know, with your software skills. Like a lot of what we're building looks and smells like enterprise SAS, except our customer is internal users, not external users. Um, and even the deep tech side is very close to like, you know, if you're a developer that's been working on like unity, And building like video game, you know, geometry engines and stuff like that. Uh, that's, that is very close to the software engineering we're doing internally and even integrating with the hardware. If you're the type of software engineer that is okay with like, Reverse engineering, some crappy API, you know, which anyone from like a rippling, for example, is done with like some crappy payroll API from the one vendor that like doesn't have it fully documented. That's the exact same problem space. And I would say tha…

AI assessment note: “Handfuls. Yeah. Handfuls. Um, the factory talent is, yes, we always need more”

Q Why are we so much better at building data centers than, than factories? Colossus, the biggest data center in the world, was built in under five months. How do we explain this?

A I think it's very easy, which is the financial markets understand data center offtake agreements. Because what that enables you to do is Brian says to me, Chris, please build data center, and I'll give you this kilowatt hour of compute at X cost unless you screw it up. And then all of the big capital allocators say, great, I'm going to let you put ten billion in capex in the ground. And as long as you can credibly execute, this is how we build power plants as well. It's like there's a price per minute, and then you put the capex in the ground. There's like a 30 year payoff period. Because manufacturing is high mix, low volume, and defense spends money every year, uh, you, you, you financially have a one year agreement. You literally can't get the capital markets to underwrite it unless you're at such a sufficient scale, which is what we're doing now, where you can functionally underwrite across many different contract vehicles, and then you can actually have this 10 year capex spread, and then you can build more data centers to be good at it. I am fully convinced it is literally just because there's like Amazon wants data center and it's a 30 year Off-take agreement, and then you could put the capex in the ground. Whereas because defense spends contracts once a year, you can't have that long-term revenue agreement, and therefore you can't, you know, have this structured finance…

AI assessment note: “the financial markets understand data center offtake agreements.”

Q When people talk about China's success, some, some people point to their industrial policy, their, their protectionism, their subsidization of certain industries. We've, you know, recently experimented with, with that a little bit. Um, what can we learn or, or not learn about what, what they've done, or how, how would, how should we think about it here? Maybe, Chris, you wanna take it first, Steph?

A Yes, I think the most important thing for people to realize is that China's stuff is not cheap because they have low cost of labor. It's because they subsidize CapEx, CapEx energy, which is the main cost of manufacturing, like, 95% of the cost of aluminum is just the kilowatt hour of the power price. And then they also do export subsidies, which is kind of like a reverse tariff. So if I send a million, if I sell a million dollars worth of stuff to ban, I get a refund from the government. They did that for the last 30 years in part to drag the capability off the US because they knew they're very long term thinkers. Unlike the US, we tend to operate only when we get punched in the face. Um, and it was a strategic subsidy. So the kind of byline is It's not economic protectionism. It's just an unfair playing field because right now it's companies like Brian and me versus the CCP, not versus Chinese innovated companies. I think we're happy to go toe to toe on an even playing field, but it's not an even playing field and it's not by 10%. It's, it's, you know, 70, 80% in some cases. So if you want to have the US as an exporter, you kind of have to do tariffs or some economic policy that just corrects for that structural imbalance. And then we'll fight over how efficient we can get in America and how many products we can produce. And I think that narrative is really wrong. Like the lev…

AI assessment note: “if you want to have the US as an exporter, you kind of have to do tariffs”

Q don't I just acquire them? Why don't I do like a private equity play and just acquire these shops instead of doing what you've done now, which is more so build the technology, train individuals to, to service, um, them are almost like absorb that That intelligence. So can you tell us a little bit more about why you pivoted there and why you didn't go with that original approach?

A One of the things that, you know, you don't really learn until you're in the weeds is every, every piece of hardware variability, whether that's a different machine or a different cutting tool or a different vice or piece of work holding every new combination of those things that you add on to the kind of like the search space of the problem. Has a nonlinear impact on the software engineering complexity that you need to build to like, you know, get to that, get to a level of automation that actually solves the problem versus just, Hey, you're 20% better than your competitors. So you have a great private equity rollup. It wasn't until I was in the weeds of that problem that I realized that most legacy machine shops have one of every machine ever made. And even within, you know, they have 20 machines, five of them might be the same, but even those five are set up a completely different way. And that adds this In enormous nonlinear complexity in terms of the software automation that you have to build on top of those systems to the point where it's, it's close enough to impossible that you can call it impossible and you can squeeze out some margin improvement, but you would never ever, ever get it to the point where it was a scalable system or it was repeatable enough that it was, you know, going to solve the problem at the kind of multi-billion dollar level that I want to solve th…

AI assessment note: “that was the core reason why you have to build”

Q is like shock to the system. So it sounds like we haven't kept up, but I think there's, there's also something you've talked about, which is that the people who were involved several decades ago are also starting to retire. So what does that look like? How many companies are actually involved in this space of advanced manufacturing? And what are we seeing there in terms of that almost lapsing?

A Yeah. So there's something like 20 to 30,000 small businesses that make up the majority of the defense industrial base. And I think just to frame this before I go into the detail, basically, if you're making a fighter jet, you know, you're actually outsourcing most of the assemblies like a wing or an engine to another company like Pratt & Whitney. And then Pratt & Whitney has a network of, you know, 2000 suppliers that give them the components and they might do some engineering. And then, so what these companies are actually doing is taking a lot of, taking the Lego kits and assembling them and then throwing the assembly up another layer and then Lockheed finally stitches the whole thing together to, to give an example. So there's about Um, you know, we have, um, 30,000 or so of these small manufacturers that in aggregate make up the entire defense industrial base, but individually are like small businesses where maybe they have 10 to twenty million in revenue max, and there's maybe 15 to 20 employees. have, And the danger there is, um, because we outsource manufacturing and because we culturally decided that if you don't have a college degree, you're worthless to society. There's been no new entrance to the manufacturing workforce because it's not sexy and it's not rewarded in the culture. And because we've had no shocks to the system, no one's realized what a big problem it i…

AI assessment note: “there's something like 20 to 30,000 small businesses that make up the majority”

Q Or even is it, is it just a matter of time? For us to catch up. Obviously it's a, you know, multivariate equation, but what would you say are the most important drivers or maybe the things that are underrated by people? Because I, I would imagine that some folks might just say, okay, let's just pour a trillion dollars into this. Well, that'll fix the problem, but will it?

A It won't. And the most important bit is, uh, talent. So if you, if you, if you think about software engineering, um, And you want to, you want to solve the problem of how can we produce more data scientists or people that are capable of working at open AI or something like that, you know, which is arguably like the top tier of software engineering. The way that you get more of those people is having this base layer of, you know, application developers and then front and backend engineers and then more hardcore backend engineers and then SREs and the data scientists. And, you know, maybe you have a million people you produce like a thousand Really incredible people in these like domain spaces and knowledge, like science, knowledge stacks on each other. You have this talent competition and you produce amazing people in these fields. Like, you know, us doesn't have the, uh, track team that goes and wins the Olympics unless we have, you know, thousands of colleges that are really competitive for track and field, right? That's, that's how it goes. And to take that sports example, you can't just start from a cold start and say, well, we want to be competitive. Uh, we want to go blow a trillion dollars and with the Olympics, it doesn't work like that. You know, that, that takes years and years of training. You need to get people when they're 14, you need to train them up. Like it's a …

AI assessment note: “It won't. And the most important bit is, uh, talent.”

Q place because, I mean, maybe I'm wrong about this, but I imagine they, they might feel like their jobs are being automated away or that they're being replaced in some way. And of course, some of them, it sounds like they're planning to retire, but how do you actually convince the person who has all this knowledge to, to want to share it and be part of this new system?

A I think there's a couple of things. The first one is the smartest and best people know that the industry is due for a transformation and that it's inevitable that someone's going to come along and do this and they want to be part of the winning team, not, not left behind. So that's a big part of it. Yeah. Incentives really matter. And I think. Culturally in manufacturing, you don't get many opportunities for growth. And there's a bunch of really smart people out there who've been kind of siloed and not given the opportunity to stretch and grow. And I think Culturally, they understand that if we're automating one part of the factory, we're not going to suddenly like fire 20 people. Like, no, no, no. Go pick whatever else you want to do. You know, manufacturing, engineering, or CMM, or a higher skill level of CAM programming. Um, and I think the other thing is, yeah, incentives really matter, and I think we're probably the only factory on the planet outside of SpaceX that is giving everyone in these roles, you know, the same Uh, tier of equity as the best software engineers have, you know, uh, every technician in Hadrian, whether you're packing boxes or whether you're an absolute master of your trade and maybe you, you know, you're one of the only 20 people left that can do it is being given the opportunity to have, you know, probably the only opportunity they've ever had to crea…

AI assessment note: “you're an owner of the company, and then you're highly incentivized to contribute”

Q Hadrian itself, what would you say the key risks are? So, so this paints a really, really interesting picture of what we can do to solve the problem. If Hadrian were to fail, why would that be? Would it be financing? Would it be, you know, a miscalculation on, on how complex some of these systems are? What do you think, you know, some of the, the key risks are?

A Yeah, I, I think the broad risk of Hadrian is that it's just unbelievably complicated, and if one piece doesn't work, then the whole thing doesn't work. So the execution You know, it's a complex coordination problem to take a, to take a teal, uh, meme. I think the second problem is as the business grows, the reality that the truth seeking of like, is this automation possible within X timeframe and how we're planning and kind of around that is, is, is really, really hard. Cause you have to make sure that people are being realistic around. Are we actually saving time with this or did we just build scaffolding for this process, which is totally fine. Like you can't really build automation until you have the scaffolding in the process, but Recruiting cycle of someone highly skilled, it might take three months. And if you're growing revenue at an exponential monthly rate, and you're booking customer sales on your ability to deliver that, and you're expecting the automation curve to go here and intersect your hiring plan here. And the automation curve turns out to be here and you have this gap. It's not as simple as, Hey, we're breaking even on this part of the project. Let's just put more labor into it for another 12 months because you can't hire into that gap for Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. And then I think, I think the third big risk is in manufacturing. If you have …

AI assessment note: “the broad risk of Hadrian is that it's just unbelievably complicated”

Q I was going to say how many people fit that bill? Three?

A Handfuls. Yeah. Handfuls. Um, the factory talent is, yes, we always need more people, but there's no like one critical talent base where we're really, really struggling. I think in terms of automation, yeah, we just need all hands on deck. And I think what people don't realize that we struggle with sometimes in recruiting or at least have to Over message, you know, to make sure people understand is that you don't need a hardware or manufacturing background to come into a business like Hadrian or SpaceX or Anduril to be able to, you know, be a competitive candidate or create value, you know, with your software skills. Like a lot of what we're building looks and smells like enterprise SAS, except our customer is internal users, not external users. Um, and even the deep tech side is very close to like, you know, if you're a developer that's been working on like unity, And building like video game, you know, geometry engines and stuff like that. Uh, that's, that is very close to the software engineering we're doing internally and even integrating with the hardware. If you're the type of software engineer that is okay with like, Reverse engineering, some crappy API, you know, which anyone from like a rippling, for example, is done with like some crappy payroll API from the one vendor that like doesn't have it fully documented. That's the exact same problem space. And I would say tha…

AI assessment note: “Handfuls. Yeah. Handfuls. Um, the factory talent is, yes, we always need more people”

Q not unique. I heard in a couple other interviews you mentioning that, like, 50% of F-sixteens are grounded because they don't have the requisite parts, or I think you gave another example of the International Space Station, something happening to it, a hatch getting jammed, and then a similar phenomena happening where they had to, like, go and hunt down a piece of paper, right? Can you tell that story?

A The F-sixteen one is, is this curious, but basically If you look at the DOD stats, more than, more than 50% of the F-sixteans are grounded, and it's mostly because they can't maintain them, and it's mostly because they can't get parts, and what ends up happening is they end up cannibalizing other planes for parts, and that makes the problem worse and worse and worse and worse and worse because of this, you know, advanced manufacturing supply chain issue, which is not a COVID issue. It's not, you know, an inflation issue. It's been going on for years. It's just now that it's coming to the forefront as, you know, more and more people are realizing how important Defense wasn't is from the space station one is hilarious. So there's a, I only get some of the details from here. Um, so if anyone from NASA is listening, I apologize, but basically, you know, the, the hatch at the space station gets jammed and astronauts, astronauts want to know how much like force they can apply it on jam the hatch. Cause if they apply too much force and they snap it, like, you know, we're screwed. Um, So, you know, they ask NASA, what's the spec for this hatch component? And it takes them several days to figure it out because there's no, there's no record of that part. And eventually they find that, you know, the guy that made it 20 years ago or whatever is retired. And as the story goes, you know, the…

AI assessment note: “the hatch at the space station gets jammed and astronauts, astronauts want to know”

Q Or even is it, is it just a matter of time? For us to catch up. Obviously it's a, you know, multivariate equation, but what would you say are the most important drivers or maybe the things that are underrated by people? Because I, I would imagine that some folks might just say, okay, let's just pour a trillion dollars into this. Well, that'll fix the problem, but will it?

A It won't. And the most important bit is, uh, talent. So if you, if you, if you think about software engineering, um, And you want to, you want to solve the problem of how can we produce more data scientists or people that are capable of working at open AI or something like that, you know, which is arguably like the top tier of software engineering. The way that you get more of those people is having this base layer of, you know, application developers and then front and backend engineers and then more hardcore backend engineers and then SREs and the data scientists. And, you know, maybe you have a million people you produce like a thousand Really incredible people in these like domain spaces and knowledge, like science, knowledge stacks on each other. You have this talent competition and you produce amazing people in these fields. Like, you know, us doesn't have the, uh, track team that goes and wins the Olympics unless we have, you know, thousands of colleges that are really competitive for track and field, right? That's, that's how it goes. And to take that sports example, you can't just start from a cold start and say, well, we want to be competitive. Uh, we want to go blow a trillion dollars and with the Olympics, it doesn't work like that. You know, that, that takes years and years of training. You need to get people when they're 14, you need to train them up. Like it's a …

AI assessment note: “It won't. And the most important bit is, uh, talent.”

Q all the time. And so if you really don't have that data layer, it's so hard to tell what's truly going on. And I'm curious to know how that impacts the end customer, right? So if you're a SpaceX or an Android buying parts, uh, through Hadrian, do they get access to that data? And if so, how does that impact that relationship compared to what they're used to experiencing?

A What customers want is effectively like DevOps level visibility that's not muddied by human aggressiveness or, Hey, we'll get the job done. Don't worry. We'll, you know, we'll pull it out of the fire. So, you know, like the world's best Gantt chart, basically. Uh, to build that though, you need the entire factory data system to get that data. Because if you just have, you know, the Flexport strategy in the early days was, you know, build the customer portal first and then have a bunch of like analysts just like typing in shipping and receiving data. You can't do that in manufacturing because it's so complicated and you have all these like cultural issues of people are always going to say, I'm going to get the job done. And that's not because they're lying. It's because like, you know, they're going to, they got white line fever. They're going to get the job done. We're a couple of weeks ish away from launching this to customers because it took so long to build that internal scaffolding to give people the data. But yes, it's like, even now we're sending like just Excel spreadsheet summaries of like, Hey, here's where we're at in this production run. And people are literally responding like, this is amazing. Like no one's, you know, no one's ever given this data or whatever, which is from, from like software land is completely insane. It would be like, you know, AWS not having a …

AI assessment note: “people are literally responding like, this is amazing. Like no one's, you know, no one's ever given this data”

Q Hadrian itself, what would you say the key risks are? So, so this paints a really, really interesting picture of what we can do to solve the problem. If Hadrian were to fail, why would that be? Would it be financing? Would it be, you know, a miscalculation on, on how complex some of these systems are? What do you think, you know, some of the, the key risks are?

A Yeah, I, I think the broad risk of Hadrian is that it's just unbelievably complicated, and if one piece doesn't work, then the whole thing doesn't work. So the execution You know, it's a complex coordination problem to take a, to take a teal, uh, meme. I think the second problem is as the business grows, the reality that the truth seeking of like, is this automation possible within X timeframe and how we're planning and kind of around that is, is, is really, really hard. Cause you have to make sure that people are being realistic around. Are we actually saving time with this or did we just build scaffolding for this process, which is totally fine. Like you can't really build automation until you have the scaffolding in the process, but Recruiting cycle of someone highly skilled, it might take three months. And if you're growing revenue at an exponential monthly rate, and you're booking customer sales on your ability to deliver that, and you're expecting the automation curve to go here and intersect your hiring plan here. And the automation curve turns out to be here and you have this gap. It's not as simple as, Hey, we're breaking even on this part of the project. Let's just put more labor into it for another 12 months because you can't hire into that gap for Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. And then I think, I think the third big risk is in manufacturing. If you have …

AI assessment note: “the broad risk of Hadrian is that it's just unbelievably complicated”

Q is a frivolous industry. Um, I'll just read you one tweet for fun, uh, from someone that I saw recently that said, no offense, but I a hundred percent think space colonization is a childish desire. So feel free to respond to that particular tweet. But really what I want to get at is within the sphere of things that you could pursue within Hadrian. Why specifically space to start?

A The really short version is like Stripe doesn't get to sell to AT&T until they sell to a bunch of white combination startups. There's just a bunch of huge amounts of capital flowing into commercial space, which means there's a bunch of net new spend and they're being run by thirty-year-olds, not fifty-year-olds, which You know, you want to sell the startups at first and you want to hit that early adult spectrum. So commercial space has all of the complexity of every other industry that we would want to serve in the future. So the automation we're building for commercial space is almost completely transferable, but there's a real need because these companies are trying to go super fast and they will pay for speed. You know, if they are sick of calling Bob's machine shop, but they're young enough from, you know, all these rocket and satellite companies that they're willing to take a shot on something new and really work with us to develop the system versus, so it's just your classic early adoptive problem. Then I think holistically is like, why go to space, uh, at all? Like why put humanity's resources out there? And I think, you know, we could go through this argument of like, you know, space is a war fighting domain. Like you need observability of the planet to stop nuclear launches. So you need satellites, you know, like GPS is a great example. We can go through all like the t…

AI assessment note: “commercial space has all of the complexity of every other industry that we would want”

Q I think it's, it's interesting that this term advanced manufacturing covers an industry which at least isn't always that advanced or maybe hasn't kept up with the time. So could you give a couple examples of ways that this space of advanced manufacturing is maybe not the reality that people might expect?

A On the defense side is a great example. Um, you know, everyone sees Defense primes coming out with drone programs or, you know, advanced fighter jets and stuff and, you know, really flashy products and marketing and things. Wow. Like everything, everything works perfectly. Um, under the hood, once you get below the assembly level, you know, someone making a fighter jet, really what they're doing is outsourcing all of the components, but the wing, the engine, you know, everything to a tiered layer of, you know, It's like, um, you know, we have, uh, 10 mini primes, then 20, you know, tier three suppliers, and then literally 10,000 small suppliers dotted across the country that are doing everything from a machine component to, I don't know, the circuit board that is one part of the fire control system or whatever. um, have, And that entire layer is basically, basically chaos. You know, the B two bomber, which is one of our most strategic assets in terms You know, the nuclear program about a year ago, the government had to issue an RFQ for some of the parts on the B to not just to replace the parts, but because no one, you know, the guy that designed it or engineered it retired and they didn't have it documented anywhere. Like there's literally no, you know, manual or document of how to make that part. So the government had to go out and say, Hey, we need someone to come up and lik…

AI assessment note: “On the defense side is a great example.”

Q When people talk about China's success, some, some people point to their industrial policy, their, their protectionism, their subsidization of certain industries. We've, you know, recently experimented with, with that a little bit. Um, what can we learn or, or not learn about what, what they've done, or how, how would, how should we think about it here? Maybe, Chris, you wanna take it first, Steph?

A Yes, I think the most important thing for people to realize is that China's stuff is not cheap because they have low cost of labor. It's because they subsidize CapEx, CapEx energy, which is the main cost of manufacturing, like, 95% of the cost of aluminum is just the kilowatt hour of the power price. And then they also do export subsidies, which is kind of like a reverse tariff. So if I send a million, if I sell a million dollars worth of stuff to ban, I get a refund from the government. They did that for the last 30 years in part to drag the capability off the US because they knew they're very long term thinkers. Unlike the US, we tend to operate only when we get punched in the face. Um, and it was a strategic subsidy. So the kind of byline is It's not economic protectionism. It's just an unfair playing field because right now it's companies like Brian and me versus the CCP, not versus Chinese innovated companies. I think we're happy to go toe to toe on an even playing field, but it's not an even playing field and it's not by 10%. It's, it's, you know, 70, 80% in some cases. So if you want to have the US as an exporter, you kind of have to do tariffs or some economic policy that just corrects for that structural imbalance. And then we'll fight over how efficient we can get in America and how many products we can produce. And I think that narrative is really wrong. Like the lev…

AI assessment note: “you kind of have to do tariffs or some economic policy that just corrects”

Q before is you have all of those different roles sit together at lunch or actually meld together in a way so that there isn't like the software engineers here and then the New machine is here and the old machine is elsewhere. And so can you speak a little bit to that and what you're seeing in terms of what aspects of each culture are being brought together and highlighted?

A Yeah, I think, and again, we're not perfect. We're probably the only ones doing it at all, but many, many, many challenges. Um, I think the first one is making sure that the software engineering team understands that, uh, the person that's operating is the, is the golden goose and they are the customer and changing that mindset from Oh, there's this internal team that does some function. You know, it's like the customer service team at a SAS startup, like who cares, you know, whatever. But if the customers, if a SAS customer screams and says, we need this feature, like everyone knows to go and jump. So making sure we understand that our internal team is the customer of what we're building. And, you know, if they have one hour of downtime, it's a really serious problem. And also their happiness as a user is like a really, a really serious problem. And that creates this Interesting dynamic where, yeah, you've got like technicians that are, you know, hourly wage or whatever, and they're the software engineer from Stripe running around being like, oh shit, this guy thinks my product is terrible. Like, you know, this is a really, this is a really serious issue. And that, that means a lot. The big problem to solve is continually reinforcing that everyone's working hard and they're, you know, there are different challenges. And I think people not from the software domain are like, why…

AI assessment note: “continually like melding those communication streams together and letting people, you know, understand each other's worlds”

Q not unique. I heard in a couple other interviews you mentioning that, like, 50% of F-sixteens are grounded because they don't have the requisite parts, or I think you gave another example of the International Space Station, something happening to it, a hatch getting jammed, and then a similar phenomena happening where they had to, like, go and hunt down a piece of paper, right? Can you tell that story?

A The F-sixteen one is, is this curious, but basically If you look at the DOD stats, more than, more than 50% of the F-sixteans are grounded, and it's mostly because they can't maintain them, and it's mostly because they can't get parts, and what ends up happening is they end up cannibalizing other planes for parts, and that makes the problem worse and worse and worse and worse and worse because of this, you know, advanced manufacturing supply chain issue, which is not a COVID issue. It's not, you know, an inflation issue. It's been going on for years. It's just now that it's coming to the forefront as, you know, more and more people are realizing how important Defense wasn't is from the space station one is hilarious. So there's a, I only get some of the details from here. Um, so if anyone from NASA is listening, I apologize, but basically, you know, the, the hatch at the space station gets jammed and astronauts, astronauts want to know how much like force they can apply it on jam the hatch. Cause if they apply too much force and they snap it, like, you know, we're screwed. Um, So, you know, they ask NASA, what's the spec for this hatch component? And it takes them several days to figure it out because there's no, there's no record of that part. And eventually they find that, you know, the guy that made it 20 years ago or whatever is retired. And as the story goes, you know, the…

AI assessment note: “they find the drawing in his desk drawer, you know, from the paperwork”

Q before is you have all of those different roles sit together at lunch or actually meld together in a way so that there isn't like the software engineers here and then the New machine is here and the old machine is elsewhere. And so can you speak a little bit to that and what you're seeing in terms of what aspects of each culture are being brought together and highlighted?

A Yeah, I think, and again, we're not perfect. We're probably the only ones doing it at all, but many, many, many challenges. Um, I think the first one is making sure that the software engineering team understands that, uh, the person that's operating is the, is the golden goose and they are the customer and changing that mindset from Oh, there's this internal team that does some function. You know, it's like the customer service team at a SAS startup, like who cares, you know, whatever. But if the customers, if a SAS customer screams and says, we need this feature, like everyone knows to go and jump. So making sure we understand that our internal team is the customer of what we're building. And, you know, if they have one hour of downtime, it's a really serious problem. And also their happiness as a user is like a really, a really serious problem. And that creates this Interesting dynamic where, yeah, you've got like technicians that are, you know, hourly wage or whatever, and they're the software engineer from Stripe running around being like, oh shit, this guy thinks my product is terrible. Like, you know, this is a really, this is a really serious issue. And that, that means a lot. The big problem to solve is continually reinforcing that everyone's working hard and they're, you know, there are different challenges. And I think people not from the software domain are like, why…

AI assessment note: “continually like melding those communication streams together and letting people, you know, understand each other's worlds”

Q superhuman saw someone else in a similar light. And so I'm curious to know if there's someone that you were like, wow, I can't believe this person is I was building something. No one's ever heard of them because in that case, no one had heard of this guy, Marc-Andre Leclerc at the time. So does that help, uh, kind of paint a picture of, of what we're looking for?

A Yeah, yeah, yeah. What really inspires me is musicians that tool away by themselves in, you know, bedrooms or caves for like five years and then produce these like works of art, uh, that are orders of magnitude better than anything that like, You know, publishing studios get out and whether that was like Boston in the seventies where literally everyone thought they were in a real rock band. It was like one guy that not only like wrote, wrote the music, wrote the albums, played all the instruments, but then half of the recording equipment, he custom built himself purely, purely just to like get his work of art out there. Stuff like that is inspiring to me, not because it's like a single human, like, you know, I am nothing without the team, but, um, It's because it's a reminder that, you know, these mythical heroes exist and like, you know, people from non-college educated backgrounds or like completely outside the system is still there, are still there toiling away. And like, you know, humanity has still got it, you know, like it's, it's not, it's not a like cattle mill of people coming out of education and going and like clicking buttons on computers at Google or Goldman Sachs or whatever. Like we've still got this, like, Incredible ability to go off and do random things. I mean, like other examples is like, I'm obsessed with, uh, these random, like YouTube building videos wher…

AI assessment note: “whether that was like Boston in the seventies where literally everyone thought they were”

Q Why are we so much better at building data centers than, than factories? Colossus, the biggest data center in the world, was built in under five months. How do we explain this?

A I think it's very easy, which is the financial markets understand data center offtake agreements. Because what that enables you to do is Brian says to me, Chris, please build data center, and I'll give you this kilowatt hour of compute at X cost unless you screw it up. And then all of the big capital allocators say, great, I'm going to let you put ten billion in capex in the ground. And as long as you can credibly execute, this is how we build power plants as well. It's like there's a price per minute, and then you put the capex in the ground. There's like a 30 year payoff period. Because manufacturing is high mix, low volume, and defense spends money every year, uh, you, you, you financially have a one year agreement. You literally can't get the capital markets to underwrite it unless you're at such a sufficient scale, which is what we're doing now, where you can functionally underwrite across many different contract vehicles, and then you can actually have this 10 year capex spread, and then you can build more data centers to be good at it. I am fully convinced it is literally just because there's like Amazon wants data center and it's a 30 year Off-take agreement, and then you could put the capex in the ground. Whereas because defense spends contracts once a year, you can't have that long-term revenue agreement, and therefore you can't, you know, have this structured finance…

AI assessment note: “financial markets understand data center offtake agreements”

Q superhuman saw someone else in a similar light. And so I'm curious to know if there's someone that you were like, wow, I can't believe this person is I was building something. No one's ever heard of them because in that case, no one had heard of this guy, Marc-Andre Leclerc at the time. So does that help, uh, kind of paint a picture of, of what we're looking for?

A Yeah, yeah, yeah. What really inspires me is musicians that tool away by themselves in, you know, bedrooms or caves for like five years and then produce these like works of art, uh, that are orders of magnitude better than anything that like, You know, publishing studios get out and whether that was like Boston in the seventies where literally everyone thought they were in a real rock band. It was like one guy that not only like wrote, wrote the music, wrote the albums, played all the instruments, but then half of the recording equipment, he custom built himself purely, purely just to like get his work of art out there. Stuff like that is inspiring to me, not because it's like a single human, like, you know, I am nothing without the team, but, um, It's because it's a reminder that, you know, these mythical heroes exist and like, you know, people from non-college educated backgrounds or like completely outside the system is still there, are still there toiling away. And like, you know, humanity has still got it, you know, like it's, it's not, it's not a like cattle mill of people coming out of education and going and like clicking buttons on computers at Google or Goldman Sachs or whatever. Like we've still got this, like, Incredible ability to go off and do random things. I mean, like other examples is like, I'm obsessed with, uh, these random, like YouTube building videos wher…

AI assessment note: “whether that was like Boston in the seventies where literally everyone thought”

Q all the time. And so if you really don't have that data layer, it's so hard to tell what's truly going on. And I'm curious to know how that impacts the end customer, right? So if you're a SpaceX or an Android buying parts, uh, through Hadrian, do they get access to that data? And if so, how does that impact that relationship compared to what they're used to experiencing?

A What customers want is effectively like DevOps level visibility that's not muddied by human aggressiveness or, Hey, we'll get the job done. Don't worry. We'll, you know, we'll pull it out of the fire. So, you know, like the world's best Gantt chart, basically. Uh, to build that though, you need the entire factory data system to get that data. Because if you just have, you know, the Flexport strategy in the early days was, you know, build the customer portal first and then have a bunch of like analysts just like typing in shipping and receiving data. You can't do that in manufacturing because it's so complicated and you have all these like cultural issues of people are always going to say, I'm going to get the job done. And that's not because they're lying. It's because like, you know, they're going to, they got white line fever. They're going to get the job done. We're a couple of weeks ish away from launching this to customers because it took so long to build that internal scaffolding to give people the data. But yes, it's like, even now we're sending like just Excel spreadsheet summaries of like, Hey, here's where we're at in this production run. And people are literally responding like, this is amazing. Like no one's, you know, no one's ever given this data or whatever, which is from, from like software land is completely insane. It would be like, you know, AWS not having a …

AI assessment note: “We're a couple of weeks ish away from launching this to customers”

Q manufacturing. And so, you know, if someone's listening, they might be like, okay, a bunch of people are retiring. Some people are, are not very keen on the idea of us continuing to pursue space. But what would you say, you know, to, to the average person, like what's at stake here? What are we going to lose if, if these people retire and we don't have these things documented?

A Unless we solve this problem, I think the country and, you know, our way of life is, is a six existential risk. And the analogy I give to people is if you're living in a small town or something like that, you know, we've built up 200 layers of abstractions in society so that like you can be an artist or you can be a painter or like, you know, you can be in finance or in crypto or, you know, making video games or whatever happens to be. And the reality of the world is that You know, 200 years ago, you know, we were killing each other over food and like, it's a, it's a miracle in the first place that we're here and that, you know, society is relatively stable and like the roads get back. People forget because they grew up in America that it's so successful that like, you know, they don't have to worry about having a bulletproof car. Otherwise, if you have more than like a 100,000 dollar net worth, someone in the gang might try and steal your daughter, which for the rest of the world is like a reality, right? Um, but because You know, America is so isolated culturally, you know, most Americans view of geopolitics is Russia bad. You don't learn those lessons. And as a younger person, you don't realize that, you know, unless we quote unquote keeps the, keep the roads paved, like this can all fall over, you know, very, very quickly. So if you run the scenario of saying, okay, right n…

AI assessment note: “Unless we solve this problem, I think the country and, you know, our way of life is, is a six existential risk.”

Q advanced manufacturing. And so, you know, if someone's listening, they might be like, okay, a bunch of people are retiring. Some people are not very keen. On the idea of us continuing to pursue space, but what would you say, you know, to, to the average person, like what's at stake here? What are we going to lose if, if these people retire and we don't have these things documented?

A Unless we solve this problem, I think the country and, you know, our way of life is, is at a existential risk. And the analogy I give to people is if you're living in a small town or something like that, you know, we've built up 200 layers of abstractions in society so that like you can be an artist Or you can be a painter or like, you know, you can be in finance or in crypto or, you know, making video games or whatever happens to be. And the reality of the world is that, you know, 200 years ago, you know, we were killing each other over food and like, it's a, it's a miracle in the first place that we're here and that, you know, society is relatively stable and like the roads get back. People forget because they grew up in America that it's so successful that like, you know, they don't have to worry about having a bulletproof car. Otherwise, if you have more than like a 100,000 dollar net worth, Someone in the gang might try and steal your daughter, which for the rest of the world is like a reality, right? Um, but because, you know, America is so isolated culturally, you know, most Americans view of geopolitics is Russia bad. You don't learn those lessons. And as a younger person, you don't realize that, you know, unless we quote unquote keeps the, keep the rose page, like this can all fall over very, very quickly. So if you run the scenario of saying, okay, right now, basicall…

AI assessment note: “Unless we solve this problem, I think the country and, you know, our way”