Defense robotics startup Medias raises $30M seed from Kleiner Perkins to automate wire harness manufacturing for the defense industrial base

Medias raised a $30 million seed round led by Kleiner Perkins, with participation from Khosla Ventures, coming out of stealth roughly nine months after founding in mid-February 2026. The company builds AI-powered robots for wire harness manufacturing inside defense factories.

What they build

Wire harnesses — the bundled cabling assemblies found in aircraft, drones, and military vehicles — are still built almost entirely by hand, on large flat tables called layup boards, by workers who may spend fifty hours constructing a single unit. Medias uses off-the-shelf robot arms rather than custom hardware, which Seeley says accelerates the security and safety certification process that he estimates consumes about 80% of the effort in getting any technology into defense environments.

The robots don't replace the skilled assemblers outright. The model is closer to capacity augmentation: automate the sub-tasks that can be automated so that existing artisans can build more, and run the line during nights and weekends when workers aren't on shift.

“Just announced today a $30,000,000 seed round with Kleiner Perkins and Kasinoa... We did with 7 figures in revenue within the first six months and basically scaled very fast. So the idea is like, let's go do 8 figures in the next six months... most of that's for scaling data — hyper specific data where you're actually doing technical tasks.”

Why probabilistic robots, not assembly lines

Seeley argues the defense industrial base is structurally mismatched with deterministic factory automation. If you're producing more than 10,000 to 20,000 identical units, a fixed ABB line makes sense. But the U.S. built roughly 100 F-35s last year. At those volumes, a general-purpose robot arm that can switch between ten or twenty different SKUs without tool changes is more valuable than a highly optimized single-purpose line.

The Ukraine argument cuts the same way: designs change fast in active conflict, and a probabilistic system lets an operator retrain the robot in roughly ten hours rather than retool an entire production line.

Seeley acknowledges the possibility of a higher-volume commoditized drone future but frames it as a later problem, with 20,000 to 50,000 units as a nearer-term target that still doesn't reach mass-production scale.

Revenue and use of funds

Medias reached seven figures in revenue within its first six months. Seeley's target is eight figures in the following six months. That pace, he says, means the $30 million seed won't go toward capital equipment — robot arm purchases can be covered by revenue. The capital is allocated almost entirely to AI talent and data collection. Training robots for precise technical assembly tasks requires specialized operators producing high-quality, domain-specific data, which Seeley describes as significantly harder to generate than general-purpose robotics training data.