Latest / Minds, Bodies, and Terawatts / Humanoid Announces Bosch Manufacturing Deal
Transcript
- 0:00[Host] This is Minds, Bodies, and Terawatts, where the assembly line assembles itself. Forbes reported today that Bosch, the world's largest automotive parts supplier, will manufacture humanoid robots for a UK startup called Humanoid, with Schaeffler supplying key components.
- 0:17[Guest] This follows a proof-of-concept at a Bosch logistics facility where Humanoid's HMND 01 robots autonomously moved boxes across five different sizes, weights, and footprints. Bosch tested them, then committed to building them. That sequence matters.
- 0:33[Host] Why is Bosch the name to watch here?
- 0:37[Guest] Bosch is embedded in every major automaker's supply chain globally. When they allocate manufacturing capacity to humanoid robots, they're signaling to their entire customer network that this product category has arrived. No startup announcement carries that kind of credibility.
- 0:55[Host] Humanoid's founder Artem Sokolov told Forbes this bridges the gap between proof-of-concept and large-scale deployment. But the company was founded in 2024. Two years old, a handful of prototypes.
- 1:08[Guest] Which is why the partners matter. Schaeffler signed a five-year actuator supply deal through 2031 and will deploy thousands of HMND units at two German factories starting this December. One deal gave a two-year-old startup two Tier-1 partners and a production timeline.
- 1:28[Host] Where does this sit globally? TrendForce projects 50,000 humanoid shipments in 2026, and Chinese manufacturers hold roughly 90 percent of that market.
- 1:38[Guest] This is Europe's most concrete move to build a domestic humanoid supply chain. Unitree's G1 sells for around thirteen thousand dollars, and Bosch has decades of cost discipline from automotive. But price is secondary. The real question is whether these robots can generalize beyond a controlled demo.
- 2:00[Host] Most humanoids today run thirty to ninety minutes before they need human help. That is an eight-hour warehouse shift where the robot taps out before the first coffee break.
- 2:13[Guest] Exactly. The demo was five box types in one facility, and every box is a fresh physics problem — grip pressure, center of gravity, surface friction all shift with each pick. But the deeper gap is training data. Tesla built its driving AI on 35 billion miles of fleet data from millions of cars. Humanoid has one logistics facility.
- 2:41[Host] Physical Intelligence raised a billion dollars to build robot foundation models and they are still data-starved.
- 2:48[Guest] You can simulate a warehouse in software, but simulation misses how a dented box handles differently from a pristine one, or how humidity changes grip dynamics. The real world has to teach the robot, and right now there is almost no classroom.
- 3:07[Host] Bosch operates logistics across hundreds of its own facilities worldwide. If HMND 01 deploys inside that network...
- 3:15[Guest] Every shift becomes a training run. The manufacturer doubles as the first customer, and the first customer generates the data that makes the next generation smarter. Hardware partner, component supplier, deployment site, and data pipeline — one deal, four roles.
- 3:34[Host] What Patrick Deglon calls the Body in Unscarcity — when the company building it also trains it with every dented box and humid shift, the gap between prototype and product collapses.
- 3:48[Guest] For more on humanoid commercialization and the training data bottleneck, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.