Most American hardware startups hit the same wall. The prototype works, a customer wants a few thousand units, and nobody at home will build them at that volume without a year of tooling and a big check. So the order goes to Shenzhen. Foundry Robotics has raised $100 million to be the place that says yes instead.
The Series A was oversubscribed and led by Marlinspike Partners. Early backers Khosla Ventures and Red Glass Ventures came back, SE Ventures, Wing VC and Axiom Ventures joined, and existing investors Hanabi and Zero Shot Fund put in more.
A factory you rent, run by robot cells
Foundry owns and runs the factory. Customers bring a design and get assembled product out, as a service, without paying for a production line of their own. The company says the jump from design to production takes weeks.
The plant is a 130,000-square-foot building in Emeryville, California, twenty times the space Foundry started in. Inside it runs grids of robotic cells. The AI models in those cells are trained on the basic moves of assembly, the fastening, routing and fitting that show up in almost every product, so a cell can switch from one job to the next without being rebuilt. Software called FoundryOS ties the robots, the floor and the cloud together, and every production run feeds data back into the models.
The list of what it already handles is long for a young company: drones, autonomous vehicle subassemblies, wire harnesses, underwater vehicles, telecom and data center gear, ground robots, rugged computers, robot compute boxes and satellites. Wire harnesses stand out. They’re among the most stubbornly manual jobs in electronics, still mostly done by hand, which is exactly why so many of them come from Mexico and Asia.
Defense is the anchor customer type
The backers and the quotes point the same way. Marlinspike’s Mislav Tolusic talked about defense and industrial makers who need assembly capacity at home. Duane Fotheringham, president of HII’s unmanned systems unit, said production capacity is “a national constraint on getting unmanned systems to the fleet.” The release doesn’t say HII is a customer. But a shipbuilder’s drone division lending its name to a startup’s funding announcement says something.
The people fit too. Founder and CEO Adarsh Kulkarni was the fourth employee at Ghost Robotics, which builds legged robots for the US military, and later Scale AI’s first head of robotics. Co-founder Emmanuel Lamacchia ran the Tesla Model Y program, a car Tesla builds at about 1.5 million a year. The rest of the leadership includes a former Tesla software lead and veterans of Raytheon and Northrop Grumman.
Where the money goes
The funds go into finishing the Emeryville build-out, putting the AI models to work on real production and hiring robotics researchers, software engineers and factory staff.
The hard part is the economics. Low-volume, high-mix assembly is expensive precisely because every product is different, and robots have historically been good at doing one thing a million times. Foundry’s bet is that foundation models for physical work have changed that. Those models need a lot of real-world data to get there, which is why companies like Maxinsights are raising money to film people doing manual work. Foundry gets its data the other way, from its own factory floor, one paid job at a time.
If the cells really adapt across products, the factory gets smarter with every order. If they don’t, it’s a contract manufacturer with expensive robots.