You don't start a factory with a blueprint. You start with a single cell, a modular block of motion and logic, waiting for its first instruction. At Foundry Robotics, that instruction is likely a CAD file for a battery pack or a missile fin, uploaded to a system that calls itself an "Everything Factory" [Foundry Robotics, retrieved 2024]. The promise is not just a robot that welds or picks, but a software-defined cell that can be taught, overnight, to assemble something entirely new. It’s a vision of manufacturing that treats physical assembly like a software deployment, and it has convinced Khosla Ventures to write a $19 million check to make it real [RoboDaily, 2026].
The wedge: assembly as a software problem
Industrial robotics is an old field, dominated by giants like Fanuc and ABB that excel at high-volume, repetitive tasks. Foundry’s wedge is complexity. It is not targeting the ten-thousandth identical car door. It is going after the high-mix, low-volume assembly problems that have stubbornly resisted automation, the intricate sub-assemblies for satellites, drones, and specialized vehicles [FutureTEKnow, 2025-2026]. The company’s core bet is that AI, particularly computer vision and adaptive control software, can finally crack these tasks. The robot isn't just following a pre-programmed path; it's seeing the parts, understanding tolerances, and making micro-adjustments in real time.
A dual-use thesis from day one
Foundry’s positioning is explicitly dual-use. The same modular cell that assembles a commercial electric vehicle battery pack could, with different software and security protocols, assemble a critical component for a defense prime. The target customer list reads like a who’s who of American industrial and defense ambition: from legacy primes like Lockheed Martin and RTX to "neo-primes" like Anduril, SpaceX, and Shield AI [fwddeploy.com, retrieved 2026].
The team betting on AI-first hardware
Foundry is the vision of solo founder Adarsh Kulkarni, a robotics engineer who cut his teeth as the Head of Robotics & Automotive Solutions Engineering at Scale AI [SignalHire, retrieved 2026]. He is a practitioner from the world of AI software, now arguing that the factory floor needs the same kind of foundational software layer. The early team is small, supplemented by a Chief of Staff and a handful of colleagues [LinkedIn, retrieved 2026].
The crowded field and the scaling cliff
Foundry is entering a space alive with activity and capital. Competitors are approaching the same problem from different angles.
| Company | Primary Approach | Key Focus |
|---|---|---|
| Machina Labs | AI-driven robotic sheet metal forming | Rapid prototyping, aerospace panels |
| Hadrian | Automated precision machining | Defense & aerospace components |
| Divergent | Digital production system for automotive | Vehicle structures, adaptive manufacturing |
| SAEKI | Robotic composite manufacturing | Large-scale structures |
| nTop | Generative design software | Design for additive manufacturing |
Foundry’s differentiation rests on its narrow focus on assembly and its explicit dual-use software stack. The risks are pronounced:
- The systems integration trap. Deploying in a real factory means integrating with legacy machines, ERP systems, and human workflows.
- The pilot purgatory. Defense and aerospace sales cycles are famously long [fwddeploy.com, retrieved 2026].
- Founder bandwidth. As a solo founder, Kulkarni must simultaneously be the visionary technologist, the recruiter, the fundraiser, and the face to enterprise customers.
What to watch in the next 18 months
The seed round provides a long runway to hit technical milestones. The next phase will be defined by a shift from potential to proof. The key signals to watch will be less about new funding and more about tangible deployments. First, a named customer announcement beyond a pilot. Second, the expansion of the leadership team. Third, a clearer product roadmap. Foundry’s premise is that software can restore agency, that flexibility and speed, powered by AI, can compete with sheer volume.