These are edited highlights from a recent Clubhouse discussion among Hadrian founder and CEO Chris Power, a16z partners Katherine Boyle and Marc Andreessen, and Not Boring newsletter author Packy McCormick. The dialogue here is part of a broader discussion that touches on a number of issues related to launching a modern manufacturing business, including software design principles, finding the right balance between automation and human labor, and why we began offshoring so much to begin with.

You can listen to the entire discussion on a16z Live.


KATHERINE BOYLE: Why is it so hard to get precision parts made in aerospace and defense? What does the American industrial defense base look like in terms of how many shops there are and the types of people who run them?

CHRIS POWER: People don’t realize that every single advanced manufacturing company — whether they’re making rockets, satellites, jets, drones, or energy for climate change — outsources about 80% to 90% of their custom parts. This isn’t like automotive where you’re printing 1,000 widgets to go on a Ford. This is “Hey, we need five crazy complex-geometry parts that are going on a rocket.” It’s about $40 or $50 billion in spend that’s coming out of these space, defense, and other advanced manufacturers going into a domestic supply chain, but it’s being routed through 3,000 or so small, owner-operated high-precision machine shops. So, in aggregate, it’s a huge industry, but it’s incredibly fragmented. You’ve really got 98% of that base doing less than $10 million in revenue. 

And it’s real mom-and-pop — usually, there’s a machinist who’s the owner who runs the business, and maybe no more than 15 employees. This was a legacy holdover from the first space race in the Cold War, because this is how the whole space and defense industrial base got built up. And it worked very well for a very long time. Actually, I have a lot of respect for these business owners, and part of how I want to play out Hadrian is making sure that they’ve got a path to come along on the journey with us. 

But the reality is that as customers have started moving faster, as you’ve seen new space and new defense companies speed up to Silicon Valley pace, all the legacy primes have to come up to that speed with them. The supply chain is just built around 10- to 12-week lead times, with a very low quality bar, low customer service bar, and incredibly low NPS. And it just got to a breaking point over the last couple of years where new space companies that are trying to move super fast were just not getting the experience they wanted out of these machine shops. Really, the entire advanced manufacturing base of the U.S. is sitting on this house of cards of these 3,000 small businesses. 

The real problem for the country, though, is that the average age of an owner of one of these machine shops is about 60, and the average age of a worker is about 55. So over the next decade — as we butt heads with China, and as we’re trying to win “space race 2” — because of the demographics and people reaching retirement age, the supply base is just going to get worse. It’s a real hidden problem that no one really knows about. You look at all these advanced rocket companies and say, “Well, the supply chain must be really slick,” when, in fact, it’s a complete disaster.

We can do it — it’s America, it’s going to be fine — but it’s not going to be as simple as grabbing 500 people from TSMC, importing them, and training everybody. Because that’s just not how you do things at scale.

PACKY McCORMICK: Someone mentioned to me recently that we’re in this four-year window where there’s an explosion in demand that makes these types of companies economically possible and necessary for the first time, really. Can you explain the why now on the demand side? Why is this such an exciting time in space and defense?

KATHERINE BOYLE: On the defense side, it’s definitely true that the time to build is now. I don’t know that it’s a four-year window; I actually think in some ways there’s a little bit more of a time crunch, because a lot of the companies that have raised capital over the last five years to serve the Department of Defense — in this hardware-software hybrid mission that they’ve been pursuing with startups — are trying to get out of what we refer to and what the DoD refers to as the “valley of death.” That’s when a company goes through pilots with the DoD, they go through certain contracts with the DoD, and then they get to a point where they’re ready for production, where they’re ready to really ramp up — and the DoD doesn’t give them contracts. 

On the defense side, what makes it a very critical time is that there’s been a funding glut, in many ways, of a lot of defense-related startups that are hardware-software hybrids or drone companies that are trying to work with the DoD. And there’s a question of how large these companies can get to get really into being prime contractors, or selling as a program of record. I think it’s different in the commercial space. 

But I definitely think we’re in a unique time where a ton of capital, and a lot of talent, has gone to these companies. People are excited to work in defense — aerospace and defense has never been more exciting for people coming out of school, even for people who may not have wanted to work on the defense side, say, five or 10 years ago. So this is a unique moment in time where we have the opportunity to really invest in shoring up these capabilities. 

CHRIS POWER: I think as a startup, you can back out of the problem and think about how you build massive amounts of market share in a fragmented industry. If you look at the best venture capital-backed companies, the pattern is always the same, where the first batch of customers that gets them to half a billion in revenue is new and fast-growing. You want to generate as many tailwinds as possible. In the space and defense market four years ago, for every program of record the contracts for those parts would have been locked in for the last 20 years. (They’re not even contractual — it’s just so hard to resource a part from one factory to another because of the quality standards, that there’s a real barrier to that revenue shifting.) Whereas four years ago you start seeing all these new space companies, and they are trying to move so fast and they’re generating a bunch of net new spend that, as a startup, you can go grab while you wait for the long tail of spend to unlock.

You see something similar with, say, HR software. How do you build a PeopleSoft competitor? You have to start with the batch of customers that need onboarding software and are growing fast, and then grab onto those. And then you eventually build into all these enterprise sales cycles.

So if you started Hadrian five years ago, there would not have been enough net new spend and customer pain to really get us accelerated to the point where we could build a huge, really exciting business and attract the talent and the capital. The reason it makes sense now is because the long tail of the spend, maybe an extra $25 billion, is inevitably going to unlock for the first time in 30 years, because of the demographics of the retirement age and all these competitive pressures on our customers. That’s why now is the right time to do it. 

But if you started this probably two years later, you’re dead. If you started it two years earlier, probably there’s not enough space companies to fund the initial revenue. 

MARC ANDREESSEN: There’s a whole bunch to it. It is flat-out easier to build companies that are just software — it just is. I think that’s even more the case today with all the supply chain issues. Anything hardware-related has lots of potential issues that, for a startup, can be very bad. It’s very easy for hardware startups to go under, so building anything with hardware is definitely playing in hard mode

The other big issue we think about a lot is that incumbents have a lot of power. Incumbents in many complex industries are organized into either a monopoly or some form of oligopoly cartel; they’re frequently intertangled with the government; and they’ve achieved what’s called regulatory capture. So they basically have taken over their regulatory bodies at the government level or at the industry level, and they have a fair amount of power to keep new entrants out.

On the other side, the pattern is just very clear: The industries that we’re talking about are both really important and there needs to be significant advances. Aerospace and defense are industries that really matter to the industrial base of many countries, to the economy, and to the future standard of living for people all over the planet. And these markets are big. When you get out of the pure software world and into these other markets, what you find is they’re generally much bigger slices of GDP. The numbers involved are just staggering, so the prize is very large. 

I always think about it in terms of whether there’s a macro strategy to be able to go take one of these big markets? And then, is there a very specific entry strategy? Is there a tip of the spear — some kind of technological change or methodological change or economic change — where there’s reason to believe there’s a seam that you can inject into and get started? That seam might be a new technology, a new approach, or a new approach to an underserved part of the market. But you need to have some very specific theory on how to enter. 

Another question is the people question, which is ultimately the most important question of all. What kind of founding team, and what kind of executive team and customer service team, do you need to build to be able to go into these markets? When you put a stake in the ground and you say, “I’m going to build a new car company” or “I’m going to build a new rocket company,” if you really push that hard and you start with the right team, you can recruit the best and the brightest from all around the world. And you can basically get all of the fired-up, sharp, and energetic people out of the incumbents, who really want to do new things to come work with you. You can basically take the talent. There’s a new kind of entrepreneur emerging that has the complete skill set, and then there are these new kinds of teams forming that didn’t really exist in the past, that I think are really exciting. We’re bullish on the idea. 

These startups will continue to be the startups in hard mode: they will have more drama, they will raise more money, they will spend more money, they will run higher octane, they will have greater risk, some of them won’t work. But the opportunities are so big that the ones that do work and the teams that can pull it off are going to build really amazing things.

CHRIS POWER: We’re seeing this decline in people — software engineers, especially — who are sick of working on optimizing ads or building dog-walking apps. This is a real moment in history where people actually want to work at Hadrian because of the mission, but also they get to build software that affects the physical world and they can see it the next day, versus seeing it through 100 layers of abstraction. If we were just a SaaS company, my God, would I be sleeping more. 

But another interesting thing if you have this wedge strategy that leads into a big market and you time it right, is you solve the Lindy problem. And the Lindy problem, as I see it, is functionally that as a startup you want to grow very fast on an exponential curve, but the Lindy effect is such that the faster you grow, the faster you fall. If you nail this strategy correctly, you have this wedge with which you get the startup growth, but if you lead into a larger sector where the average company that you’re disrupting has been around for 30 to 40 years, you should get this massive tailwind that leads into a 100-year company. Because it might take you a couple of years to get a billion in revenue, but maybe 20 years to get to the full 40 billion. 

I think this is what you’re going to see with companies like FlexPort. Are they the fastest growing company of all time? No. But will that company potentially be around for 100 years and be very hard to disrupt once they reach maturity? Probably yes. You’ve got to be really intentional about starting that sort of a company. There’s three phases: There’s the first five years, then the next 30, and then there’s thinking about how you hand it over to the next generation in year 50. That’s how I think about timing and building massive companies, as well.

You could argue that the biggest thing that Elon Musk has done with SpaceX, for the country, is not SpaceX. It’s the fact that they trained thousands of aerospace engineers who were 25 and now 30, and now are starting all these companies.

PACKY McCORMICK: Why is this the thing to work on for the next 35 years? That’s a huge commitment to make. What’s at stake, and what’s important for people to realize?

CHRIS POWER: The Planet Labs co-founders have this concept of an “Archimedes company,” which is one that, by being as financially successful as possible, also has knock-on effects to the world. The mission is perfectly aligned with the balance sheet. Because, often, what you get are mission-driven companies that try to do their mission, but it’s in conflict with them being financially successful. So I was obsessed with finding, in the American industrial space, an Archimedes’ lever of a company, where if you build it and pull it off, you can have an outsized impact on something that you really care about.

And as an Australian, I was seeing the rise of Chinese influence across the world. I’m a big student of history, and I’m a firm believer in reserve currency and that the world changes every roughly 100 years. We’re clearly in that trap of the global order potentially changing. And if you look back at history, the thing that wins wars, or prevents wars from starting, is basically peace through strength. But it’s not just who has the best fighter jet. It’s how many fighter jets can we make and what’s the replenishment strategy?

If you just look at the dynamics in space and defense in America, you realize that the supply base is on the decline. You think about why 50% of the F16s are grounded because they don’t have parts; they’re cannibalizing other planes to get parts. And if you dig five layers deep, it’s not turtles all the way down — it’s Bob’s Machine Shop all the way down. In this decade, as we settle the solar system, I would like to help ensure it is largely under a free, peaceful world order.

MARC ANDREESSEN: There was this narrative for 20 years or so, between the mid to late ‘90s to maybe five years ago, which was basically the “cloudification” of manufacturing. It’s this idea that manufacturing is a thing that can be abstracted away and offshored. A lot of it moved to China, specifically. A lot of people, I think, have this mental image of this “China manufacturing cloud” — stuff happens in China, they just build this stuff and send back finished products, and there’s not much more to it than that — and, of course, those of us who have worked in or invested in hardware businesses know that that’s not quite how it works.

And then, of course, there’s political and economic tension and controversy over all this offshoring in the last 20 years. Calling it “globalization” is the positive framing, and then the “offshoring or hollowing out of American manufacturing” is the negative framing. This is a topic that always comes up in politics. In the last few years, for a variety of reasons, it feels like that conversation in the U.S. has really shifted. I alluded to this in my Build essay, which is that maybe some of that made sense for a while, but maybe it’s gone too far. Both President Trump and then President Biden, in remarkably similar language, have basically said it’s time to reverse that process and it’s time to bring manufacturing back to the U.S. for a variety of reasons — economic reasons, but also national security reasons.

There’s an easy narrative there, which is “Oh, we can just bring it all back.” You hear this a lot now with the Russia situation. It’s like we’re going to move into this bipolar/multipolar world. We’re no longer going to be reliant on Russia and its natural resources. We’re not going to be reliant on Chinese manufacturing. We’re just going to be able to do everything domestically, and you’ll have a more segmented or isolated world. But maybe that picture is also too abstract — the global supply chains are so complex, and so integrated, and there are so many parts that constitute any finished product, that reshoring manufacturing is actually going to be much harder than people think. 

How does one analyze all that, and where will this all land, say, 10 years from now?

CHRIS POWER: In terms of reshoring manufacturing, it is significantly harder than people think it is. It’s less of a supply chain problem and more of a culture and human problem. The first problem is obvious. In the way you train people and the way you learn skills, you cannot skip any of the ladders. You cannot take a workforce that’s used to making widgets for Toyota and get them to run a semiconductor factory. There is no training. You have to go up these complexity rungs in skills in a human population. The only thing you can optimize for is how fast you can climb those rungs. You absolutely just can’t skip to the top. 

So reshoring manufacturing is not as simple as, “Hey, let’s bring pharma back and let’s bring semiconductors back.” That’s only successful if you can either import the talent wholesale, or if you have sub-industries of manufacturing complexity that generate the talent that can actually do the highly complex stuff. We lost that population of people and lost our starting base when we offshored all the simple stuff. 

The other thing, which is a bit more esoteric, is that when we outsourced manufacturing we lost this industry that naturally generates serious people. And what’s interesting is that we build up society in these layers of abstraction, but manufacturing is one of those industries where because you make contact with reality every single day, you can’t hide from the realities of how the real world works. So having a big manufacturing population actually generates a ton of serious people who, when they are 50 and in Congress, are making really rational decisions that are grounded in real-world experience. Whereas if you remove manufacturing or other hardcore industries from the equation, you have someone who’s now 50 and who’s grown up having lived a very easy life, and they’re making decisions that pattern match off that easy life — not off of making contact with reality every single day. 

We can do it — it’s America, it’s going to be fine — but it’s not going to be as simple as grabbing 500 people from TSMC, importing them, and training everybody. Because that’s just not how you do things at scale. It’s just not how the world works.

You could argue that the biggest thing that Elon Musk has done with SpaceX, for the country, is not SpaceX. It’s the fact that they trained thousands of aerospace engineers who were 25 and now 30, and now are starting all these companies. One of my hopes with Hadrian is that by building the company big enough, we teach people how to do this, and they go off and start all these other advanced manufacturing companies, because we’re teaching them the meta of how to build things in the real world.

These startups will continue to be the startups in hard mode . . . But the opportunities are so big that the ones that do work and the teams that can pull it off are going to build really amazing things.

PACKY McCORMICK: You want to make it as simple as possible for people to be able to do these complex jobs, without removing from the process autonomy, agency, and the feeling that they’re learning something and achieving something. How do you think about creating all of these jobs and making it possible for people to learn how to do these high-tech jobs in a scalable manner?

CHRIS POWER: I think a lot of the complexity is primarily tribal knowledge. If you can instill tribal knowledge down to some checklist processes, that’s the first big step. Because then training is, “Hey, you’re learning how to cut metal on a saw. Here are the 10 steps. The first seven are probably easy, but if you get stuck on the last three, please ask for help.” Whereas most manufacturing training is, “Hey, apprentice under someone for three years, and maybe they’ll spend an hour vaguely teaching you about saw-cutting metal at night. But who knows.” So we structure it a lot. 

Within the factory, you start at shipping and receiving, then you go into tool building, then you go into repeat jobs and machining, and then you go into prototype machining. Then, you can be a quality inspector or a CAM programmer. So within each of those rungs of hourly pay bands and seniority, there’s enough short cliffs to jump where you can have people learning on the job. And building tools actually teaches you a lot about machining. So by the time you’re ready for a promotion, you’ve already learned half of it through the context of doing the previous rung in complexity and skill.

The only way to rebuild America’s manufacturing workforce is to flood the first part of that funnel with thousands of people who are smart, young, maybe fix their car on the weekend, maybe play video games — and then have this self-learning process throughout the funnel. Because if you try to skip someone to the eighth stage of complexity, it takes three years to train them and they kind of die. We’re also aware that we can’t have our top 10 machinists spending six years training someone. 

But also that comes from great operating leaders from places like SpaceX and GE, who’ve learned these lessons of how to train people, and we’ve baked that philosophy into the software by listening to the people who have done this sort of en masse training before. 

PACKY McCORMICK: How do you build a culture where you bring all of these top people in — from a bunch of different backgrounds, levels of expertise, and areas of expertise — and get them working together in a cohesive manner?

CHRIS POWER: I think the first thing is you have to hire the first batch of people who are as close to understanding the other world as possible. And then once you’ve got that core 10 or 12 people who really get it, you can build cultural reinforcements that set the stage so that anyone new coming in gets the play. People kind of laugh at me for this, but literally everyone in the company eats lunch together at the same table. So when people are hanging out, you’ve got maybe a top principal software engineer who was at Google having lunch or playing foosball or whatever with someone who’s coming from Chick-Fil-A, and is on an hourly wage working for us in the factory.

You have to start with a core group of people that really get the vision and understand the root cause of why the culture is so bad on both sides. And then you really have to proactively fight the reversion to the mean of class dynamics, or whatever, inside the company. We really think intentionally about that, and it’s one of the more powerful things about Hadrian that we need to nurture over time. We have rejected top software engineering candidates purely because there’s a 5% to 10% risk that they’ll come in and walk all over the machinists. And now that we’ve got the core, we can show people what we’re doing and people can adopt our culture.

Join the Newsletter

Technology, innovation, and the future, as told by those building it.

Thanks for signing up.

Check your inbox for a welcome note.