Bosch Is Building a Robot That Cost 5% of What Its Competitors Spent

Humanoid is betting it can win the humanoid robotics race by spending far less than its rivals.

The UK company — founded in 2024 as SKL Robotics Ltd. — has raised under $100 million, roughly 5% of what Figure and Apptronik have raised, yet last week it announced a production partnership with Bosch, the German industrial conglomerate, to build its HMND 01 humanoid robot at scale for the European market. The gap between what it raised and who agreed to build its robot is the whole story.

The capital comparison is the hook: Forbes reported that Figure has raised around $2 billion and Apptronik roughly $1 billion. Both competitors are building extensively in-house — their public trajectories suggest a vertical integration approach where hardware, software, and manufacturing are controlled end-to-end. Humanoid's argument is different. Its argument is that the race to build a working humanoid may be as much about partnership architecture as it is about capital — and that its lean approach, if it works, changes what the build-out of physical AI actually looks like.

The proof point is a March proof-of-concept run at Bosch's intralogistics facility in Bühl, Germany. Over five days, a single HMND 01 unit moved boxes across five different sizes, weights, and footprints in a conveyor-to-trolley transfer task, coordinated by Humanoid's KinetIQ AI software, according to RoboticsTomorrow. It is a narrow test in a controlled setting. But for a company trying to get Bosch to commit factory floor space, it was the demonstration that turned a pitch into a partnership.

Bosch will serve as the contract manufacturer and provide what it calls a Design for Excellence framework — essentially a structured plan for taking a robot from prototype to mass production, covering hardware design, supply chain, and cost optimization, per RoboticsTomorrow. Peter Svejkovsky, Head of Corporate IP at Bosch, said in a statement that the company has the global production infrastructure and deep industrialization expertise to make that step. "With our global production infrastructure and deep industrialization expertise we are the perfect partner to take the step from prototype to volume manufacturing," he said. CEO Artem Sokolov said the agreement bridges the gap between proof-of-concept validation and large-scale deployment.

Why Bosch chose to partner rather than build its own is its own question. Per Thomasnet, Bosch and Humanoid also plan to explore integrating Bosch technologies — including sensors, actuators, and drive systems — into future versions of the HMND 01 platform. The detail that Bosch is planting its flag inside a new category while letting someone else carry the development cost is not a conclusion stated in the source. It is an inference the reporting supports. The arrangement lets Bosch deepen its industrial automation business while positioning itself inside a new category.

The timing of the two deals matters. Just six days before the Bosch announcement, Humanoid disclosed a separate Schaeffler partnership — a binding, phased agreement under which the German industrial components company will supply more than half of Humanoid's joint actuator demand through 2031, a volume The Robot Report estimates translates to a seven-digit number of actuators. Schaeffler also committed to deploying more than 1,000 Humanoid robots at its own facilities, scaling to support 100,000 robots by 2031. The first systems go live in Germany before the end of 2026, with a three-month capability demonstration and integration testing period at Schaeffler's Schweinfurt plant running through June 2027.

That 100,000-robot target is still a projection. Actuator supply chains, software reliability at fleet scale, and the actual cost of integrating humanoid workflows into existing factory operations are all open questions. Humanoid has given itself a narrower margin for error than competitors with ten times the runway.

The partnership model Humanoid is running is a legitimate alternative to the billion-dollar vertical integration bet, or it is a workaround that runs out of road when the real scaling problems arrive. The binding agreements are signed. Whether the model holds at execution is the next question.