BMW doubles down on humanoids

BMW is widening two separate AI programs inside its manufacturing and engineering operations: humanoid robots on the factory floor and domain-specific AI in vehicle development. The company’s latest move follows an 11-month pilot at its Spartanburg, South Carolina, plant, where two humanoid robots from Figure AI worked on a live assembly line, according to a June 2 report by Yahoo Finance. In parallel, BMW Group and Mistral AI disclosed a partnership on June 1 to apply AI to crash simulation and related engineering work, using BMW’s internal data and Mistral’s model-training capabilities. Together, the announcements show BMW moving beyond one-off demos and into programs with named teams, datasets and production targets. ### What exactly did BMW change on the humanoid side? BMW has created a “Center of Competence for Physical AI in Production” to coordinate humanoid and other learning-system work across its manufacturing network, according to BMW materials and related industry coverage. The center is intended to pool robotics and AI expertise and to evaluate technology partners under industrial production requirements rather than lab conditions. (finance.yahoo.com) The Spartanburg pilot was the proof point behind that step. Yahoo Finance reported on June 2 that two Figure AI humanoids were used over an 11-month test at BMW’s South Carolina plant, where they worked on a live assembly line. Figure said in a November 2025 post that its Figure 02 robots were deployed at Plant Spartanburg and ran on an active assembly line every working day during the later phase of the project. (press.bmwgroup.com) ### What did the robots actually do in Spartanburg? Figure said on November 19, 2025 that its robots handled sheet-metal loading tasks at BMW’s Spartanburg plant, picking parts from racks and placing them onto welding fixtures. The company said the deployment contributed to production linked to 30,000 vehicles, with more than 90,000 parts loaded during the project. Those figures came from Figure, not BMW, and they describe a narrow production task rather than general-purpose factory work. (finance.yahoo.com) Independent industry coverage of the pilot said the program focused on a single repetitive material-handling application and tracked cycle time, placement accuracy and human intervention as key measures. ### Why is BMW pairing this with a Mistral AI deal? (figure.ai) BMW Group said on May 28 that it is working with Mistral AI to build industry-specific AI models for crash simulation. The company said it runs thousands of virtual crash simulations each week and has accumulated more than one petabyte of historical crash-simulation data for training and analysis. (rockingrobots.com) Dr. Franz Decker, BMW Group CIO and senior vice president, said the company is combining its engineering datasets with Mistral’s model-training capabilities to build specialized AI for complex development work. Marjorie Janiewicz, Mistral AI’s chief revenue officer, said the project is aimed at “Industrial AI,” with crash simulation as the first use case. (press.bmwgroup.com) ### What does BMW say the simulation work is supposed to improve? BMW said the Mistral collaboration is meant to improve the quality, accuracy and speed of crash-simulation tasks. The company described the work as a first step toward scaling domain-specific AI into other parts of vehicle development and the broader BMW value chain. (press.bmwgroup.com) That matters because the simulation program is not separate from manufacturing strategy. BMW’s public materials describe digitalization and AI as core parts of its production model, and the company’s physical-AI center is set up to move tested systems into broader plant planning. ### Where does BMW go from here? BMW has already started a European humanoid pilot at Plant Leipzig in Germany, working with Hexagon and Hexagon Robotics, according to company materials and trade coverage from March and May. (press.bmwgroup.com) The Leipzig project is focused on battery assembly and component-manufacturing applications, while the new Physical AI center is intended to support wider deployment decisions across BMW plants. (visit-bmwgroup.com) On the engineering side, BMW said crash simulation is the first application in its work with Mistral AI, with additional use cases planned across vehicle development and the company’s value chain. The next visible milestones are likely to come from BMW’s Leipzig pilot updates and any new disclosures on how the Mistral models are being used inside engineering teams. (press.bmwgroup.com) (visit-bmwgroup.com)