Key Takeaways

  • Medius emerged from stealth with a $30 million seed round led by Kleiner Perkins and Costanoa to automate defense wire-harness production.
  • The company hit seven figures in revenue within six months by deploying off-the-shelf industrial robotic arms instead of designing proprietary mechanical hardware.
  • Deterministic automation lines from suppliers like ABB work when volume exceeds 10,000 units, but defense programs like the F-35 require high-mix, low-volume flexibility.
  • In defense manufacturing, 80 percent of the operational hurdle is clearing IT and security requirements before the robotics can even touch a part.

The High-Volume Trap in Defense Hardware

If you want to assemble 50,000 electric vehicles a year, your manufacturing playbook is solved. You buy deterministic machinery from legacy industrial vendors. As Justin Seeley put it: “If you're going to build like more than 10 to 20,000 of something like you should just call ABB and build a factory line for it, right? Like then you get deterministic every time. You get that weld perfect every time.” (play from 1:33:51) Fixed tooling works when the geometry never shifts.

Defense does not operate on consumer scale. The defense industrial base rarely builds hundreds of thousands of any complex platform. Instead, programs like the F-35 fighter jet demand low-volume, high-SKU configurations. Wire harnesses, the nervous systems of modern aircraft and vehicles, still rely on manual labor. Hundreds of thousands of assembly workers sit at pegboards running wires by hand because traditional automated machines jam the moment a flexible cable shifts two millimeters.

Hardware founders often respond by trying to invent custom robotic actuators to force physical flexibility. Medius took the opposite route. They left the hardware alone and attacked the problem with probabilistic vision-action models running on standard industrial arms.

Data Outlasts Custom Steel

Building proprietary hardware burns capital on tooling, maintenance, and supply chain delays. Standard industrial arms already possess mechanical reliability and precision. What they lack is hand-eye coordination for non-rigid objects.

Medius directs its capital into collecting proprietary technical training data rather than bending custom metal. When an arm perceives flexible wire bundles through vision models, it adjusts its trajectory in real time instead of expecting sub-millimeter repeatable placement. That adaptability allows the same cell to assemble different wire configurations without weeks of mechanical changeover.

This approach altered their capital structure. Seeley noted that the company can “probably pay for all of our capex through revenue,” (play from 1:36:23) reaching seven-figure revenue in its first six months of operation. If your capital expenditure is self-funding, venture dollars fund pure model development and data collection.

There is a second operational moat: clearance. Seeley points out that “getting technology in defense tech is like 80% IT and security requirements and then you got to solve the problem as well.” (play from 1:31:15) Hardware teams that ignore defense compliance will find their robots sitting outside the factory door, regardless of how fast they manipulate wire.

What to Do With This

Audit your product development roadmap this week and list every piece of custom hardware you plan to build. If a standard commercial component can physically perform the motion, cancel the custom hardware design and reallocate those engineering hours toward proprietary computer vision and training data. Selling software that controls standard machines gets you into production environments months faster than waiting for custom metal to arrive.