Johny Fernandez: Most factories in America are built to make one thing. Change the product, and retooling the line can take months and millions of dollars. My guest wants to build a factory that can make almost anything, run by robots that learn on the job. His company is less than a year old, and it’s already shipping hardware for energy, telecom, self-driving, and defense customers. And last week it raised $100 million to scale up. But first, a look at the company. The road from an idea to the opening bell starts right now.
America wants to bring manufacturing back home, but there’s a problem. Traditional factories can take months to switch from building one product to another. What if robots could do it in weeks? That’s the bet Adarsh Kulkarni is making after building military robots at Ghost Robotics and leading robotics at Scale AI. He co-founded Foundry Robotics. The company uses AI-powered robots to assemble complex hardware, from drones to satellite components, without rebuilding an entire production line. And investors are betting big. Foundry just raised $100 million, expanding 20-fold into a 130,000-square-foot California factory. But building robots is one thing. Building a profitable robot-run factory is another. Can Foundry scale fast enough to transform American manufacturing from Startup to Stock Exchange? Foundry Robotics founder and CEO Adarsh Kulkarni is next.
Joining us now is Adarsh Kulkarni, co-founder and CEO of Foundry Robotics. Mr. Kulkarni, welcome. Thank you so much for being with us today.
Adarsh Kulkarni: Absolutely. Thanks for having me.
Johny Fernandez: All right, let’s go ahead and get started. You call it America’s first autonomous factory for assembly. Walk us through the floor that you have in California and what you are looking at.
Adarsh Kulkarni: Foundry is the first company really rethinking assembly and manufacturing from the ground up. We actually build factories, so we think a lot about what happens when you redo the building from the ground up. You walk into a Foundry facility and you see a mixture of robotic work cells doing a ton of different kinds of work, whether it’s fastening, putting things together, or cabling. You almost see this grid of robots, some of them building drones, some building batteries, some building self-driving car hardware. For the first time, we can build many things under the same roof. And it’s not that it’s all robotic. There are human technicians doing the really, really finely tuned things that robots can’t do yet. But you also have [unclear: “UVs”; likely AGVs] moving between the human technicians, the robots, and the others. When I describe it, it sounds sci-fi, but it’s real now. We can change very quickly. We build drones on Monday, batteries on Tuesday, and satellites on Thursday. That’s real, and it’s real because not only do we build factories, but we fine-tune robotics foundation models for manufacturing. There’s a lot of talk about next-generation robotics and how it’s going to do all this household work for you. We think that’s great, but we think there’s a lot more work it needs to be doing in the factory, and that’s what we focus on. We take best-in-class AI and foundation models, and we make them focus on manufacturing and we deploy them at scale. To my customers, I’m just the greatest manufacturing-as-a-service partner you could get, because as a result of all my technology, I’m faster, I’m cheaper, I’m higher quality. And you better believe I’m domestic.
Johny Fernandez: I love hearing that. I want to talk about a traditional assembly line. It’s built around one product. Why does changing it take months, and how do you get it down to weeks? Like you said, it sounds sci-fi, something you see in movies, but now it’s becoming a reality.
Adarsh Kulkarni: If you look at traditional manufacturing today, it’s very much automatic assembly. What I mean by that is there are machines, but they are programmed very rigidly to do the same thing again and again. Historically, that’s what automation in manufacturing has meant: machines programmed to follow very, very specific instructions millions of times over. Or you have manual assembly, which is all people who have been trained to do a specific kind of wrench turning again and again. Both of them are very rigid. My co-founder, [unclear: “Immanuel”], ran the Model Y program at Tesla. It’s the best-selling car in the world. And even at Tesla, it costs hundreds of millions, if not billions, of dollars to retool when you change a car. What allows us to be so fast and generalized at Foundry is that we don’t focus on the products, we focus on manufacturing itself. The factory is our product. I don’t sell a drone or a battery or a satellite. By doing that, we take the AI and focus it on different manufacturing primitives. I have an AI that’s going to get really, really good at fastening bolts. I have another one that’s going to get really good at cabling, and the good news is it can do cabling for many different products. The way I summarize it is we are specialized in the process, but we are generalizing the product, because manufacturing is so broad. It’s so big. Most of America doesn’t have a good feeling for what happens when you step inside a factory anymore. It’s so complicated. There’s so much going on. By focusing not on this product or that product, but on this process or that process, for everybody, it allows us to rethink how we do it. And so we have this army of robots and people that are able to put multiple different kinds of products together.
Johny Fernandez: That’s truly fascinating. Something you run everything on is software you call Foundry OS. What does the factory learn every time it builds something?
Adarsh Kulkarni: Great question. Foundry OS is one of our secret sauces at the company. We’ve talked a lot about robotics at the level of two robot arms building a thing. When you zoom out and you’ve got 100 of those, or a thousand, or a million of those across 100 sites, just getting them to work together and talk to each other is a pretty complex problem that nobody has really solved before. When you think about it, these robots are good at fastening bolts, these robots are good at cabling, and those robots are good at welding. How do you schedule the jobs? I need this welding robot to work on this problem. I need this human technician to work on that problem. That’s a pretty complex set of constraints to work around. On top of that, do I have the parts yet? Do I have the supply chain? There are all these questions that traditional manufacturing throws a lot of people at. At any large-scale manufacturer today, there are thousands of people who manage all of this orchestration, deployment, and complexity. We think you can do that all much, much more efficiently. Foundry OS is the single hub where all the orchestration of the different jobs happens. It’s where all of our AI training happens. It’s where all of our data collection happens. It’s where all of our interactions with the customer happen, where they can log into our dashboard and say, okay, here are my ten units, five were built yesterday, five were built the day before, and they’re arriving at my facility tomorrow. It’s a completely different way of experiencing a manufacturing partner, and it all runs on that one system. What all of this allows us is, for the first time in history, and people talk about exponential returns, it’s never happened in manufacturing, the more products we build at Foundry, the better we become at building all products. That’s wild. That’s new. That is what is really, really exciting.
Johny Fernandez: Now that is exciting. The other thing is that you don’t sell robots into someone else’s factory. You own and run the factory yourself. Why did you build it that way?
Adarsh Kulkarni: Great question. There is a lot of complexity in running a couple tens of thousands of robots in a building. On the one hand, you have a warehouse, which is four walls and a roof. On the other end of the spectrum, you’ve got a data center, which is one of the most complex facilities in the world. Traditionally, factories are pretty much advanced warehouses, but the way we’re seeing robotics deployment happen, we need fiber lines hanging from the ceiling, we need multi-100-gig drops, we need GPUs under each robot and then a mini data center off to the side. We need electrical charging stations for all the [unclear: “see of UVs”] we have. It’s a very different amount of infrastructure you need, so the complexity of the site is pretty significant, and expecting a customer to do that doesn’t really make sense. The second thing is we just need more capacity in the United States. Foundry does critical work in defense and energy and telecom. Even if I was selling robots, we still need to build more capacity in the United States. My thesis for the company from the very beginning was, if we need more capacity anyway, if we’re going to need to build more factories from the ground up anyway, why don’t we do it in a tech-forward way? AI is a thing now. Robots are starting to work. How would we redo this in a way that rewrites the laws of physics? All of that complexity is something we get to work on because we run the building. If I was trying to sell you a robot, I’d have to figure out how to productize it. I’d have to make it lighter. I’d have to make it cheaper. Instead, I can have a much, much larger R&D department solving so many interface problems on how you run a factory that you wouldn’t be able to otherwise. I’ve been in robotics for a long time. You’ll see this in the industry: the business of providing an end service on your robotic fleet typically is a much better business than selling a robot. The best example is Waymo. You don’t buy a Waymo. Even if you’re a billionaire, you can’t buy a Waymo, but you can take a ride in one. It’s the same thought process here at Foundry.
Johny Fernandez: For more than a century, the factory has been built around one idea: keep the product moving. In 1913, Henry Ford put the Model T on a moving assembly line at his Highland Park plant in Michigan, and within a year, a car took about 93 minutes of labor to assemble. Then in 1961, the first industrial robot, the Unimate, went to work on the General Motors line in New Jersey, handling red-hot metal parts. Since then, robots have gotten faster and stronger, but most still do one job on one line for one product. But now the question is different: can a factory learn a new product as fast as engineers can design one? And this is the bet Adarsh Kulkarni is making in California. Welcome back, Adarsh Kulkarni, co-founder and CEO of Foundry Robotics. Before the break, we were diving into what you’re doing. The next question is $100 million, oversubscribed, led by a [unclear: “Marlinspike partner”]. Where does that money go first?
Adarsh Kulkarni: Great question. There are really two big buckets where the money goes. The first is we need to scale up capacity. We have so many customers banging down our door right now because people need a manufacturing partner in the United States that is tech-native and easy to work with, and who actually understands their products. Almost a third of Foundry used to work at Tesla. We understand next-generation modern hardware design. We understand how to do design for manufacturing. We understand how to do manufacturing, whether it’s autonomous or not. We’re just a great partner. So that’s one side of the business: we scale up capacity, we get to build more things, we work with more customers. The other side is the robotics R&D. We have so much more to build. We have many different kinds of models to train. We have new versions of ourselves coming out, some bigger, some smaller. We design a ton of our own end-of-arm tooling, the stuff you put at the end of the robot arm, whether it’s a gripper or a screwdriver or whatever it is. Historically, that’s not something anybody is really focused on. We have our own fastening end effector. It’s basically a really, really fancy screwdriver that is torque-sensitive, manufacturing-grade, and tied in directly to Foundry OS. We have a ton of custom hardware that goes into that. As we scale up from our R&D office in San Francisco that was about 10,000 square feet, we’re now across the bridge in Emeryville at 130,000 square feet. That’s 15 times the size, and it allows us to really invest in R&D, in what the robots can do, what we can train the models to do, and what we can start deploying in production. It allows us to solve our customers’ problems immediately. The other thing we’ve seen very quickly is that we can start to invest in other businesses. Foundry’s bread and butter is electromechanical assembly. We put stuff together. We manage your supply chain for you. All your parts come to us, we QA them, we test them for quality, we make sure it’s the right count, we grade your suppliers for you and give it back to you. We make your life really easy that way. But very quickly we can see that there are certain parts of the supply chain that really need help. We were at the Pentagon last week and we saw the same thing. Wire harnessing, for example, is a super large focus area. We started doing wire harnessing six weeks ago, and now I have seven customers for it. These things snowball quickly because the industry wants to design the next greatest drone, the next greatest satellite, the next greatest battery, but then they’re stuck with this problem that manufacturing is really hard and it handicaps design. What we’re going to do for manufacturing is what AWS did for compute. No software engineer coming out of YC, or Y Combinator, thinks, oh my God, how am I going to scale my data center? That’s not a question anymore. Five years from now, or three years from now, I want that to be how hardware founders think. [unclear: “Foundries just to get go”], right? That’s how we scale manufacturing, and it’s going to magically work the same way AWS does.
Johny Fernandez: That’s definitely interesting. The next question: a $19 million seed in April, a $100 million Series A in October, and a factory 20 times bigger. The big question a lot of you want to know is, how do you scale a hardware company that fast?
Adarsh Kulkarni: That’s a great question. For me it comes down to talent. I am 100% sure we have the greatest combination of manufacturing expertise and robotics applied research. That is not normal. You normally have one or the other. My co-founder ran the best-selling car in the world. I invented the robotics data industry at Scale AI and built that business with Alex Wang. I used to build robot dogs for the Defense Department. We are such a unique combination of robotics and manufacturing, and our company reflects that. We’ve hired some absolutely legendary people from manufacturing and the overall hardware industry. We have people who built satellites. We have the founding software leader of all of Tesla’s manufacturing operating system. He runs our software team today. We have a world-class team in manufacturing. We also have a world-class robotics team that’s focused on deployment. What I mean by that is not making the models fold another T-shirt better, but how do I build a satellite with four nines at superhuman speed with 24/7 uptime. Our research focuses on deployment, and those people sit next to each other, they eat lunch together. Traditionally, that’s not what happens. By harnessing expertise in all of these different domains while being very low-ego and high-humility in the domains that are not yours, we really get to put our first-principles hat on and ask, what should a factory look like in 2026? When I build a building in 2030 from the ground up, it really should look pretty different from the buildings we built in the 70s. And yet today, there’s not a huge difference. When you come down to it, our company is a combination of our people, and our people are absolutely world-class and extremely diverse in where they come from. When you have people who are highly talented, highly technical, and highly experienced, you tend to make pretty good decisions that allow you to go really, really fast.
Johny Fernandez: [unclear: “Mo”] Adarsh, definitely interesting to see how you’re growing, what you’re doing, and what the future holds for you. I want to thank you so much for joining us this morning here on the show.
Adarsh Kulkarni: Absolutely. Thanks for having me.
