如何打造下一个 SpaceX 或 Tesla

You don't have a defense industrial base without an industrial base. It's been a real wake-up call that we need to get building again.

1. Why AI Still Can't Turn a Wrench

AI will be a phenomenal tool that can be used in the physical world. The reality is, AI cannot turn a wrench yet.

The majority of the top 10 most valuable companies in the world are hardware companies.

2. Steel-Toed Boots and Giant American Flags

What's it like to visit one of these sites?

You're building a physical thing. You're not sitting at a desk on your computer; you're out in the world. It sounds kind of sexy and romantic, but when you're actually doing it, it's rough.

When you show up at Mariana, Hadrian, and Castelion, you meet people who are so passionate about what they're doing, and that produces really good results.

There's that very famous chart of U.S. power production versus China's new energy production. What would it take to close that gap?

I don't think that is a technology problem.

3. No Defense Base Without an Industrial Base

Connor and Erin, thank you so much for joining the podcast today. We're here to talk about a16z 50 and this year's list. Let's get straight into things. I know a big theme this year was the American industrial base. Can you talk about what we've learned this year and why that's such an important theme right now?

We've been saying for a while, you don't have a defense industrial base without an industrial base. I think this year really put that into perspective. With conflicts going on in the world, we've realized that the hollowing out of our industrial base over the last several decades has real consequences on the battlefield for soldiers, both American soldiers and our allies. I think it's been a real wake-up call in the United States that we need to get building again.

Yeah. I think it's also not just one factor, but multiple factors all happening at the same time. When you take a step back, you have world conflict and the continued evolution of the capital markets, with capital flowing into this ecosystem, which I'm sure we'll talk about later. But one of the bigger things is that you actually have an entrepreneurial base right now raising their hands and saying, "I'm going to build in this ecosystem." It's incredibly important, not just for our nation, but also because you can build really valuable companies in it. I think it's a confluence of geopolitical factors, capital markets, and entrepreneurs all coming together. It's a really interesting time to be building, and a good time for us to be investing.

When we looked at a16z 50 this year, we thought about where we are in the history of American dynamism. We're still early, but one theme that's evolved is that you really need to look at American dynamism companies as a totality, end to end: from the input that comes out of the ground, to the factory that makes the thing, to the end system. That's how we've set up a16z 50 this year, to think through those 3 lenses. Each of those pillars is uniquely important to the end product or system, but I don't think people usually think that way. A lot of the companies in our portfolio are addressing things from an inputs perspective, a factory perspective, and now a systems perspective.

One theory a person could develop after looking at the list is that we built out this industrial base for defense, for building missiles and drones, but it's now a tailwind helping a lot of different industries in America. There's energy, robotics, and AI. Can you talk about how all of these things fit together?

Yeah. It's funny because if you look back in history, even 70, 80, almost 100 years ago, a lot of the time, technology in America was developed by the government. If you look at how we went to space for the first time, how we developed the internet, or how we developed nuclear energy, it was the government saying, "We're going to do this for the purpose of the government." Then the long tail of that technology was super valuable for all kinds of things: the internet being created, energy being proliferated across the world.

What's different now, to some extent, is that private industry and private capital are saying, "We're actually going to be the impetus that develops something." To your point, many of these markets serve the federal government or the Department of War in some way. But the beauty of these things is that when you build an industrial base, either down the supply chain or in the end systems themselves, those things inherently have many use cases.

The term "dual use" gets talked about a lot in investing. To some extent, I think focusing on too many things at once is a folly for a startup. But it's absolutely true of what's happening with any factory being built today, any new factory, any energy production, or anything touching minerals or materials in the ground. Those materials aren't just going into a defense system; they're going to go into my uncle's car or my kid's future toy. I do think the tip of the spear may be focusing on the government, but in the long run, these systems will be used in the Waymo we took today or in all kinds of other things.

4. How DJI Went From Drones to Robots

I think a good example of this is DJI in China.

Yeah.

It's gone through multiple arcs in its lifetime and journey. It started out building consumer drones and recognized in the late 2010s that there was a large market for drones in both public safety and national defense. So it built out a high-scale drone production business for defense. Now the company behind DJI—or the folks behind DJI—are the largest producers of robots in the world, because a drone motor is very similar to a robot actuator, which is the hardest part of building a robot.

You have this company that, over the course of growing a large defense business that itself spun out of a consumer business, is now the world's leading expert on how to build and produce hardware for the next wave of physical AI, both consumer and industrial. This technology transfer and process-knowledge transfer between military and non-military use cases is really important, and it's always been how it works. It's great to have the federal government be a subsidizer for figuring out this process knowledge and the new technology. I think the United States' greatest gift is the discernment of the American consumer: we're going to push things to be more capable and cheaper than the government ever will. That symbiotic relationship in the industrial base is really important, and we're certainly seeing that on the energy side.

We're seeing that on the electronic side. We're seeing that across the entire industrial stack. To some extent, I don't think this is a super complex topic, but it's important to restate: when you build the first of a kind of anything, it's quite expensive. When you're building a bespoke military system or consumer product—

Zero to 100, it's going to cost a lot.

It's going to cost a lot. We'll get into this a bit later, I'm sure, but we talk about vertically integrating or using the supply chain base that you have. The beautiful thing about what's happening right now is that we have a federal government and a Department of War saying, "We're going to actively rebuild some of this industrial base." By doing that, they're changing the unit-economic decision from the start for these companies. Instead of building 100 things, maybe you build 100, but the government is willing to pay a cost the consumer or commercial market wouldn't. All of a sudden, that means more machines you can put in your factory, more capex that the government is helping you pay for at the start, and things that are financially viable and can operate at a scale that actually reaches consumers. It's important to identify that, and it is definitely happening right now.

Yeah. If you're building a new PCB fab in the United States, that's a pretty expensive operation. Those are pretty expensive pieces of machinery. If you have a defense buyer as your first buyer, helping you pay for or subsidize a lot of that capex investment, that's great. You know what else has PCB boards? Robots. You know what else has PCB boards? Hearing aids. The ability to pass on some of that capex to the defense industrial base will be really powerful as we rebuild industrial capacity in the United States.

5. The Elon Playbook: Vertical Integration

What specifically are you seeing startups do to drive down costs and move faster?

Yeah.

I mean, I think, like many things, we can look to Elon for a lot of answers here. It’s not a crazy statement to say that the few modern-day companies that have built anything at scale are probably SpaceX and Tesla. If you look at how Elon and his team have built those businesses, they’ve been very vertically integrated.

And look, I take the view that you shouldn’t vertically integrate everything. If you’re building a super-exquisite system like a missile that’s doing something very unique, you kind of have to, because you have to control the exquisiteness and the cost.

Well, SpaceX and Tesla were doing things for the first time. There didn’t exist an industrial base in the U.S. to support them, so they had to build it themselves. Where there are high-quality suppliers, you should use suppliers, for sure.

Yeah. I think the lesson is that if you don’t have an industrial base you can rely on quickly and reliably, then you have to do it yourself. That’s a hard journey. The idea that you can focus on the end system and build the most cost-effective system you can, while also figuring out the ball bearing and the solenoid valve, is a lot to do.

In the case of Tesla and SpaceX, I’d argue the auto industry was a little bit different because there wasn’t a battery supply chain. Elon had to recreate that.

The whole power electronics system, the whole drivetrain—they had to build that completely from scratch.

6. Mariana Minerals Bought a Copper Mine

That was based on papers published in the ’70s that had been collecting dust for 40 years. In terms of companies in the a16z portfolio, there’s Mariana Minerals, which bought a copper mine and is reinventing things end to end. They recognized that the existing players in the mining industry weren’t going to adopt technology quickly enough for it to make a difference or really move the needle. They also wanted to own the entire value chain so they could put their money where their mouth was and capture the end value of the new technology they were bringing.

They do everything from acquiring and developing the asset to mining and operating it, and refining the material themselves across multiple minerals. Then they sell a final commodity product in the market. They’ve also really leaned into AI and software as a huge component of their build. That’s another trend I’d point to.

Yeah, I think applying software and AI to the physical world will look different from what’s happened in legal, customer service, or any other category where AI is taking off today. But, to Erin’s point, many times you have to own the asset in order to change it. Not every time, but a lot of the time, seeing is believing in the physical world. When you’re purely a digital AI product, you can just see it, and it’s very easily transferable and understandable.

What Mariana is doing with this copper mine is impressive in a lot of ways. It’s not just saying, “Hey, we’re going to be better at extracting copper.” It’s rethinking everything. It’s—

The entire process of how you—

—plan, operate, mine, and refine materials. What’s cool is that once you own that process end to end and use software at every stage, you capture the data, and then it feeds back.

And this is an important point. This isn’t just Mariana. If you look at another business like Hadrian, it’s a perfect example of how you can use software. More importantly, it’s not one silver bullet that suddenly makes a factory or a mine perfect. It’s actually 15 or 20 different things.

When you look at what Chris has done at Hadrian, it’s everything from quoting and taking an order for a part to actually making the part, inspecting it, and transferring it. There are all these different intermediate steps within a factory. The beautiful thing about starting from a fresh slate, like Chris has in building his factories, or taking over something, as Mariana has, is that you can look at every single step from a first-principles perspective and ask, “Was this process that’s been done for the last 100 years the right process?”

That’s what’s happening across every layer of the physical world right now. It’s not just mines or factories. It exists in many different forms. My personal belief, and what we see in a lot of our portfolio companies, is that AI will be a phenomenal tool in the physical world. The reality is that AI can’t turn a wrench yet. It will at some point, but we’re far off. We can have a discussion about the robotic models and autonomy we see in the physical world; we’re getting closer.

The nuance is that these portfolio companies—and many non-portfolio companies—understand the physical world isn’t about putting forward a silver bullet, some shiny AI product. It takes a deep understanding of what a process looks like physically and how to manipulate or change something that’s happening in the world. That’s going to look different depending on the process.

Well, maybe a specific example to talk through is low-mix, high-volume versus high-mix, low-volume, and how AI can change the game there. Historically, if you’re spending a lot to set up a line with a lot of automation—

—you really need to be producing a lot of one single thing.

Mhm.

That’s how you get the cost down and justify the investment in a pretty bespoke manufacturing line. In the U.S., we do a lot of high-mix, low-volume manufacturing, which is very hard to automate because it’s expensive when you have to change the line all the time.

What we’re seeing across our portfolio is that manufacturing technicians have the intuition, and now they have the software tools, to change the economics of producing low-volume, high-mix work on the line. A manufacturing technician—someone without a college degree, or maybe they have one, but not necessarily someone who studied software engineering—can use these tools to reprogram and reconfigure their manufacturing lines on the fly. They understand the manufacturing process and know how to do the work in the physical world, and now they suddenly have all these tools.

You can retool a line without calling a system integrator and waiting three months and spending hundreds of thousands of dollars to reconfigure it for a new setup. A manufacturing engineer can use Claude and do it in 20 minutes. That’s been a real game changer across lots of companies we’ve seen in this space.

And why that’s important is, when you look back at how we’ve articulated the AD 50 this year, there are kind of three categories. A lot of the time, what I and many people in the sector focus on is the systems companies. There are a lot of really interesting systems companies—Anduril, Saronic, Castelion. The list goes on, and there are a lot of people building really good end systems.

But if you can’t iterate on the design and the prototype, you not only don’t get to an end system that’s different and better, you also move at a slower pace. By the time you finally fix a production-level missile or plane or whatever, you have to go back and rejigger and build up all these subsystems to meet the scale. If you have a more transferable, adjustable line, you can build the prototype faster, and then set up each component so it can get to rate production quicker.

Yeah. You’re designing for manufacturing from the outset.

Exactly. That’s a lesson I keep coming back to from Tesla and SpaceX. In a lot of ways, they had to make those choices because they didn’t have any other opportunity. But in the end, that mantra of designing for manufacturing now exists within all these systems companies.

To some extent, we're now in the age of seeing them produce. Look at what Anduril has built in Ohio and now on the East Coast in Baltimore. Look at what Saronic has built with its shipyard. These are massive facilities that, on a very near timeline, are going to be producing thousands, tens of thousands, and hopefully hundreds of thousands of systems. We haven't done that in America in a long time.

7. Building It Before the Government Asks

Yeah, this is no longer an on-paper theoretical exercise. This is happening in the real world, and real stuff is coming off the lines. Something special about AD founders is that they make any problem an Elon-sized problem. I'd love to hear how that applies to the different industries we're talking about here.

Yeah, I think when you look at defense specifically, this has historically been one of the hardest sectors in which to build the thing you think the government needs. Unfortunately—and I experienced this in my past when I served in the Army—the buyer is not the user, and the person appropriating the dollars for the buyer is completely removed. So you have all these disjointed decision-makers or groups that matter.

Back then, we didn't really have neo-primes or AD companies to serve. What you had were primes that would sit there and say, “Tell me what you want. I'll develop the thing after you give me the money.” The age-old problem is that you get a system that's not technology-forward. You get a system that was developed based on what was needed 10 years ago, and the end soldier gets something that's not super valuable to them.

What's different now, going back to the Elon-sized problem, is that you can think through what the military actually needs. The beautiful thing is that venture capital, what we do at American Dynamism, and the rest of the capital markets will fund that in advance. You can build in advance of the government saying, “Here's the thing for you to build.” The outcome is that the soldier gets a better product, something that fundamentally changes the battlefield and helps with deterrence, security, or safety. More importantly, you get a system built on a timeline aligned with the market's requirements and what technology can do today.

Instead of waiting for the government to tell you what it wants, the government signals, you see it, and you build in advance using private capital. That just hasn't happened in the history of defense. Take Castelion, building a hypersonic missile that isn't a super-exquisite thing. It's really hard to build a missile that goes that fast and is reliable. The government has now given them contracts and shown up and bought missiles. But if you look at the last 3 years of that business, it was private venture capital and entrepreneurs taking the risk and building the thing in advance. If you see that again and again in defense, you'll get better products, and the neo-primes and new companies will start to take a lot of that market share.

But it's not just defense. Erin sees it in energy as well. Erin would know better than I would, but we're at the one-yard line. We're just starting that process. I think another part of it is a little more intangible. When I talk to Turner, who's the CEO of Mariana Minerals, he's flying to middle-of-nowhere places all over the world to build his company. He's out where there's no cell phone signal, where there's no electricity, touching dirt, and spending weeks and months away from his family.

It has to be worth it. You're building a physical thing; you're not sitting at a desk on your computer. You're out in the world. It sounds kind of sexy and romantic, but when you're actually doing it, it's rough. The travel is really hard. The people you're engaging with aren't like other startups that operate on the same timescale and culture as you. These are really tough people to negotiate and work with, and they have a very different culture.

And if you're going to be putting yourself through that, it better be damn worth it. People really gravitate toward these problems that feel huge and enormous because these are real missionary founders that we see building in this space, even as it's heated up, there's more capital, and it's gotten cool.

8. When Hardware Became the Most Valuable Game

It's still brutal. I think the biggest thing that has changed in American Dynamism in the last few years is that we always knew the companies being built in this category were really important to the fabric of America. When we talk about defense systems, energy, and manufacturing widgets that go into everything, that stuff is really important. It affects our safety and how we live our lives. But the biggest change is that those companies weren't just the most important thing to us. Now they can be the most valuable companies, and that's a big distinction. It's not just solving a problem for safety or security; it's solving a problem for safety and security, and those can be some of the largest companies in the world. That's what technology has changed in these industries. Mining has always been a big industry. Defense and energy have, too.

It hasn't been economically worth it for a young, smart person with a lot of ambition to go after these categories, but I think—

Companies like Palantir, Anduril, and SpaceX have proved that you can build a really big business operating in these categories. If you look at the top 10 most valuable companies in the world right now, the majority are hardware companies. They're building a physical product. We've come full circle.

You get the most ambitious founders of our generation, who want to run toward that hard problem and know it's going to be worth it.

And the beautiful thing about many of these industries, like mining, is that Rio Tinto has been around for a long time. Boeing's been around for a long time. These are good companies building good products. But when you take a first-principles technology approach to what audacious thing needs to be built, and you get entrepreneurs like the folks on this AD 50—

Like, “Yeah, I'm going to rebuild the American energy grid. That sounds fun.”

And that just didn't exist 50 years ago. It didn't exist 5 years ago. It's something fundamentally different: America's best and brightest are building toward this category for the first time. Importantly, like Erin said, the opportunity is there. You see companies that have redefined space launch, redefined military systems, and redefined all these categories of American Dynamism that we care about. It's exciting. As someone who served in the military, you always think of what you do as something greater than yourself. The capital markets exist to make money. We're not a charity; this is a capitalist society. But in the end, a lot of the companies on this AD 50 list are doing something that's both going to be large and valuable, and inherently important to the fabric of America. That's pretty exciting.

What's it like to visit one of these sites, like a manufacturing site, a mining site, or a drone manufacturer?

I mean, Connor and I have the best job of all time. I do own steel-toed boots now because of the number of site visits I've been to with portfolio companies where they're like, “Yeah, you've got to show up in steel-toed boots.”

Yeah.

I mean, it's not all sexy, to be honest. These are locations that are often hard to get to, in the middle of nowhere. We're talking about physical things that require earthmoving. In my time in the military, I had a senior officer who would always tell me, “The physical world doesn't care about you, so just be careful in the physical world.” These are tons of dirt being moved, massive factories being built out—Hadrian, or what AMA will do.

These things are huge, and it's impressive. One of my favorite things is this running competition between companies over who can have the bigger American flag at their site. Maybe you don't see it at the remote sites; maybe you do. I actually don't know who holds that today. Hadrian's was pretty big last time I saw.

I think Valar is pretty big.

Valar's big, too. Okay, good. And again, that's awesome—the fact that you have that kind of passion for something greater than just the product or the economic opportunity you're building. It drives purpose behind a lot of these teams. When you show up at Mariana, Hadrian, or Castelion, you meet people who are so passionate about what they're doing, and that produces really good results. When you have to do something really hard in the physical world, if you have people willing to run through walls—sometimes literally—to solve that problem, you get really audacious, great outcomes.

One of my favorite things to do is visit the same company multiple times within a 12-month period, whether they're in the portfolio or not, because you learn a lot by how fast—

A factory buildout happens.

And so, one thing we do is, you don't just go once. You go once, and then you go 3 months later, and you go 3 months later, and you go 3 months later. I'm not a manufacturing expert. I've spent enough time on a factory floor between my time at Palantir and this job that I know some of the machines and roughly how a line works. But there's something about seeing things accumulate in a factory.

You watch it go from an empty warehouse with epoxy floors, a big American flag, maybe a couple of pallets, and founders excited to show you their rented forklift. They have a basketball hoop or gym set up in the corner. Fast-forward 6 months, and they're freaking out about how to expand. They're saying, “We need to get the city out to give us more power because we have too many machines plugged in.” It's kind of chaos: technicians and operators everywhere, people building stuff. That's really satisfying and fun. We're lucky to get to see that journey and how quickly it can happen.

Yeah. Has anyone ever done a time-lapse of that?

Oh, there's a lot. It's a whole category of filmmaking, and everyone compares theirs.

I do like the gym thing. I remember when I first visited many years ago, Isaiah at Anduril made me do a rep.

Yeah, they all have their bench press setups.

9. How Defense Unlocks the Robot Wave

Which is cool. Is there a contagion effect to things being made in America? Because the actuator is made here, can we make this and then that? When you're able to compress one timeline, can you compress multiple timelines?

I don't even think it's just about timelines. I also think it's cost and quality. It's interesting because when we look at defense autonomous systems in particular, Saronic is making a lot of boats, Covenant is making a lot of missiles, Castelion is making a lot of missiles, and Anduril is making a lot of things. When you build many systems, you also require many subsystems below them.

There are really interesting military systems being built today. The corollary, back to the comment we had earlier, is that by building for defense, you also build for a different industry. We don't have a lot of humanoids or robotics in the world today. There are thousands, maybe; I don't know the exact number. But when you think about the subsystems that go into some of our military systems, they're the exact same subsystems that will go into future robotic systems: actuators, motors, valves, switches, and those kinds of things.

When you can build those to quality and at the right cost, all of a sudden you make other markets that maybe weren't viable or real make sense. I think that's going to happen in the robotics world. We're going to spend all this time developing those subcomponents and build for a lot of these military systems, which we need for security, safety, and deterrence. Then all of a sudden, a robotic wave will come to America. I said it: AI can't turn a wrench yet. They can. They will eventually. That will be enabled by all the actuators, switches, and subsystems we've built.

It's going to serve a different market and make it financially viable, and we're not going to have to rely on factories in China. As you build an industrial base, it can be used for many things. Down to the component level, I just think we don't have enough of it in America. When you talk to a robotics company today that wants to build a robotic deployment, there's a 99% chance that the majority of what they're buying is coming from China, and some—

For many reasons—not just geopolitical ones—they want to bring that purchasing onshore: quality, cost, speed of iteration, and geopolitical reasons. One thing that's remarkable is how much of a supplier ecosystem has developed over just the last 3 years.

One of the companies in our portfolio, 3 years ago, would order twice the number of PCBs they needed because they assumed they would have to throw 50% away for quality issues. They'd get a batch from their supplier, do the testing, and 50% wouldn't pass a quality test. Then startups emerged in this category and said, “We're going to fix this problem. We're going to use software and modern technology to redo the way PCBs are manufactured in the U.S. We're going to do it faster, cheaper, and better.”

10. The Rise of the Manufacturing Middle

Now that same company buys PCBs from new suppliers, and 70%, 80%, 90% of them are passing quality tests. I think we underestimate how much the co-location and co-development of this technology with the end customer and their suppliers really matters for driving quality higher.

Something really interesting is happening. I believe in the industrial base. Like we talked about earlier, you get this top-down demand signal from these new systems companies. We're building new military systems, but, by the way, Joby and Archer are building new planes, too. We're building new end systems. That's a fact: cars are being built differently, planes are being built differently, and military systems are being built differently. When you build these new systems, the inputs to those systems inherently need to change as well.

When you look at the industrial base in America today, historically, if you go look at how Boeing makes a 737, Boeing is a company important to the fabric of America, but they just pick up the phone and call Honeywell, and Honeywell helps them build all the different subcomponents. It's not Honeywell's fault. Honeywell says, “I don't want to build a new landing gear again. We have this landing gear that meets the spec and does the thing, so let's keep doing that same thing again and again and again.” It's rational. They're going to build more 737s, and they'll get paid the same way whether it's a 50-year-old landing gear or a new landing gear.

The problem is that when you keep iterating on new requirements for end systems, at some point the 50-year-old subsystem just doesn't work. Maybe it physically meets the requirements, but it's been designed and developed in a way that we haven't innovated on. Why does that matter? We've developed new materials. Physics hasn't changed, but the inputs to the physics and the inputs to the design have changed.

I think the opportunity for a lot of these manufacturing companies is what I call co-engineering. We talked about the long tail of these sub-tier suppliers: they've been in receive mode for the last 40 years. Honeywell gets a call saying, “Hey, I need another landing gear. Go make that landing gear,” and they've executed on the specs with that IP. When you want a new requirement, that middle layer can't think on behalf of Honeywell and change the requirement, because they're not in a position to do it.

In the future, you'll have this middle layer of supply companies effectively co-engineering with the end OEM to rethink not just what the end system can do in terms of speed or weight or mass, but the individual subcomponents, too. They'll be rethought, re-engineered, and rebuilt. The end system will change at the top level, and the individual parts and components will change as well. I think that just hasn't happened in the last 40 or 50 years.

So you're going to see a completely new wave of what I believe will be called the manufacturing middle. That will be co-engineering and pushing design and development upward in the supply chain, as opposed to just being in receive mode and saying, “Okay, I can build the same thing I built 40 years ago.” I think that's an opportunity. I defined it previously as creating the American manufacturing asset class right now.

Private equity has been playing in this ecosystem for a long time, looking at an asset, buying it, improving the cost, and selling it off. Again, that's not fundamentally changing the things that we build. Right now, companies will change the different things that we'll build.

A good example of this is the grid. When you think about the electronic stack—

We're about to go through a huge step change in how we manage and transfer power on the grid, moving from static and mechanical systems that were deployed at large scale across the U.S. in the '70s and '80s to modern power electronics and battery storage architectures, which are much more software-native and flexible. When you have such a complete step change in technology, where the technology replacing the previous thing is completely different, you have an opportunity to rethink the entire supply chain behind it and the way all of that design and engineering work happens.

And in the energy world, it's the same thing that's happening in the defense or commercial systems world: you have this massive amount of demand for data centers, new factories, and new builds. At this top-down level, you have the ability to innovate, like Erin's talking about, and rethink the entire architecture.

Because it's a completely new architecture, you can't rely on components that worked in the previous world and just reconfigure them. You actually have the opportunity to rethink that from the ground up.

Another important point here is that you could be pessimistic and say, “Okay, Connor, not all these new military or end systems are going to be adopted. Or, Erin, we're not actually going to increase energy by 4 times; there's not going to be that much demand.” I don't think it matters, because once you start the snowball of the demand cycle moving in one direction, even if you deliver half of that requirement—

You're not going back.

You're not going back, and it doesn't make financial sense. When you rethink something from first principles and make a new design and engineering decision, a lot of times that comes with better cost and better efficiency, so you're just going to buy that thing now—

Better serviceability, easier to manufacture.

Yeah, yeah, you're never going back.

11. Spinning Out of Anduril and SpaceX

I'm curious what you're seeing in terms of founders being second-generation Anduril or second-generation Palantir. Maybe they see some of these learnings and the opportunity to be the middle-layer supplier in something that hasn't been touched in decades. What are you seeing in terms of those opportunities and those kinds of founders?

Yeah, for sure. A great way to light a fire under an entrepreneur is for them to be banging their head against the wall on an engineering or supply-chain problem. They're like, “Okay, I really care about this. This is a pain in my ass. I know it's bothering other people. I know it's a problem that I don't just have here at Anduril or SpaceX or Tesla. It's a problem more broadly. I'm going to go solve that more generally.”

The lineage of founders who've lived and breathed this on the manufacturing floor—and not just founders, but even engineering teams who are like, “I saw this problem. I solved this within Anduril or Saronic, and I'm going to go out and build this as a company”—is something we're certainly seeing. I know everyone in the ecosystem, so I have automatic access to a captive customer base. This is how ecosystems get built. People make fun of YC for all YC companies kind of selling to each other, but that's how it works. They're your friends. They're tolerant and understanding; they're willing to work with you, help you, and put up with something breaking because they're working on their own really hard problem. If the thing you're doing isn't the most existential thing to them, they're very willing to outsource it to you. That's beautiful. That's how Silicon Valley was built. What's really making me optimistic is the extent to which we're starting to see that in these physical-world ecosystems and supply chains.

Yeah. I think the few companies you mentioned, SpaceX and Palantir, weren't overnight successes. Those companies took a long time to reach both the product scale and the size and success they have today.

Overnight successes that took 20 years.

Exactly. I think the learning point—and we invest, and a lot of the entrepreneurs on this list come from those ecosystems—is that you learn this relentless focus to do one thing as well as you can, down a path until you've succeeded. I've seen this with the Castelion guys, for example. They focused for 2.5, 3 years on the first hypersonic missile, which was really hard to build. Other people showed up and said, “Oh, be multi-product. Go build this small drone here.” And they're like, “No, we need to do this one thing really well.”

I think that comes from the fabric of SpaceX. At SpaceX, the concept of being a responsible engineer is, “If this thing doesn't work, it's my ass. I'm going to obsess over this thing again and again.” That's what it takes to build really hard things. There will be different derivations of success. Not everyone will think just like the SpaceX mindset, but if you look at a lot of these businesses in the physical world, you kind of need to do that to build something really hard.

What you do in doing that is you learn process knowledge. You learn how to do it. So thing two, thing three, thing four are a lot easier. You've developed a supply chain. You have your relationships. You've developed design-for-manufacturing processes. You've learned how to build a factory to a high quality, high spec. You've learned how to build that trust with customers. If you can do that one thing really well, that sets you up to do other things really well, but it doesn't work unless you deliver on that one core thing.

12. America's Real Edge: Its Capital Markets

So we talk a lot about competition with China. What does the United States do uniquely when it comes to our capital markets that really enables these startups?

Well, I think this may arguably be one of the things Brian Schimpf, the founder of Anduril, was talking about a couple of days ago: We have literally trillions of dollars of private capital in our capital markets that inherently follow success—both technology success and, to some extent, production success as well.

If you look at the United States, number one, we do not have a top-down-directed industrial policy that says the state will put this amount of money into this. I think to some extent that's a good thing, because the money will follow the entrepreneur and the money will follow the success. We've always had good capital markets in the United States. I think the distinction is that the capital markets have all of a sudden said, “You know what? The category of A&D is really interesting.” Erin and I have been living and breathing this for a long time.

If you look not just up and down Silicon Valley and Sand Hill Road, but at the major banks, JPMorgan has committed trillions of dollars toward investing in the physical world—not just defense, but factories and other things in the physical world. You see that again and again across not just our venture capital firms, but private equity is now starting to play in this category in a big—

Big credit firms.

Big credit firms. That's so important because many of these companies in the A&D world require a lot of capital. When you build physical systems—a Falcon 9 rocket, a CCA, or a Blackbeard missile—those are expensive to build up as production lines and physical systems. Part of this is that the capital markets speak to where success happens. The reality is that our capital markets aren't top-down driven by our government, unlike China's. That allows us to be far more flexible and, again, to follow success. More importantly, it allows us to fund things before it's obvious. We talked about that earlier: the concept of building in advance of the requirement, the need. Maybe we're a little bit late on the energy side, maybe a little bit late on some of the defense systems.

But look, the capital markets have woken up, and we’re putting hundreds of billions of dollars, if not trillions of dollars, toward this category for the next 10-plus years.

Yeah. If you look at just the first half of 2026, more risk dollars went into the defense ecosystem than all of last year combined, which was many times what went into the defense ecosystem in all of the previous years combined. Historically, in defense specifically, there’s been a narrative around the valley of death: How do you get a government contract to propel you through the period when you need to build out capex?

What’s really exciting is that the valley of death is starting to disappear with the support of capital markets and risk capital. That’s extending beyond the defense ecosystem to energy, robotics, critical minerals, and some of these other really important sectors, which is super exciting to see.

13. The US vs China Power Chart

Yeah, I had one question. I’m not sure if it’s too much of a tangent, though. There’s that very famous chart of U.S. power production versus China’s new energy production. America’s line is basically flat, and China’s is up and to the right. I guess my question is: What would it take to close that gap? How fast can we build new power?

Yeah, I don’t think that’s a technology problem. I think we need to do some soul-searching about our will as an American people. Do we want conforming projects to be built? Do we want to grow? If we do, I think the technology and the capital will follow.

Yeah. And it’s funny, because this isn’t just an energy problem. It’s an everything problem.

It’s an everything problem. And look, we’re optimists inherently. I do think that snowball has started to roll. The question is, back to our discussion on capital markets earlier, will there be emergent winners in these categories that then fuel more capital to flow into the ecosystem? It’s this kind of self-fulfilling prophecy: If you build it, they will come. But at some point, there needs to be a pot at the end of the rainbow. If there is one for a few companies, I think capital will flow, like I mentioned.

The last thing I’ll say, just real briefly, is that anything in the physical world takes time. To set up a factory at a scale where you really affect our power generation, our military systems, or whatever, you can’t take a couple of engineers, throw them in front of a couple of servers and computer screens, and come up with an answer. It’s moving dirt. It’s physically understanding a supply chain. These things take years. There’s a little bit of patience we need to have on the capital markets side, but again, I think I’m optimistic. I think it’s happening, and to some extent, it’s ours to lose. It’s ours as a government and a nation to mess up, to get in the way of this. That’s a risk, don’t get me wrong, but—

I do think it’s another long-term benefit of capital markets, even if it’s short-term painful. People need to feel—I mean, you mentioned these are physical things that take time. They’re also physical things in that they have a physical manifestation for the people and communities that live around them. The most important thing is that American society at large feels like they’re participating in the economic growth that’s coming as a result of this development.

Software had the ability to be siloed because it existed in just a few cities. It was kind of this virtual thing. You experienced it in the way your interaction with digital services got a lot better over the last 20 years, but it was a little more abstract. The physical world is different: Someone’s building a data center, a power plant, or a mine in your backyard. You have a say and a stake in that. You physically see it; it affects your life. If you don’t feel like you have a stake in that economic upside, it’s hard. I do think this is one thing that capital markets will solve really well in the long term, because companies will figure out how to bring communities along for the journey and make it financially and economically worth it for them.

And I think there are some good examples of that. A few companies on the AD50 list build real things in their communities and show up the right way. There are also bad examples from the last few years. If you look at what AMA does, they show up to these really small machine shops that exist. They partner with them; they don’t cut the management and fire people. They say, “How can we build new products?” and grow the kind of business that you have today.

It’s that type of concept—working with communities as opposed to working against them—that I think we still need to wrestle with a little in the tech world. Just because we sit in Silicon Valley doesn’t mean we understand everything that’s happening in the world. Fundamentally, I think a lot of these AD50 companies will build stuff that touches Middle America. It’ll physically exist in Middle America, and I think that’s net good. I think that means more jobs, more prosperity, and a world of more energy and abundance, where we have a safer America and—

Less concentration of wealth in these pockets, these elite pockets on the coast. It’s a good thing.

Yeah.

But it will take time.

Yeah. Well, Erin, Connor, thank you so much for joining.

Awesome. Thanks.

Awesome. Thanks so much.