Figure AI's partnership with BMW — announced with considerable fanfare in early 2024 — placed humanoid robots on the assembly line at BMW's Spartanburg, South Carolina plant, the company's largest U.S. facility. The initial deployment involved Figure 01 robots performing bin-picking and parts-transfer tasks.
The Figure 02 robot, unveiled in August 2024, represented a significant hardware upgrade: OpenAI-powered end-to-end neural networks replaced the task-specific programming used in Figure 01, allowing the robot to understand and respond to natural language instructions. A widely circulated demonstration video showed the robot answering questions about its surroundings and performing object manipulation tasks with apparent fluency.
The factory deployment results, as reported in trade publications through early 2026, tell a more cautious story. BMW has continued the pilot but has not disclosed plans for large-scale deployment across the plant. Tasks that Figure robots handle reliably in the factory environment are mostly structured material handling — moving components between fixed positions in a known layout. Tasks requiring adaptation to unexpected situations, like a misaligned part or a bin filled at an unusual angle, still require human intervention.
Figure AI has since released Figure 03, with improved manipulation capability and better uptime metrics. The company's CEO Brett Adcock said in April 2026 that the BMW partnership remains active and that Figure plans to expand to additional manufacturing customers in 2026. He declined to name them or provide deployment scale numbers.
The broader picture in humanoid robotics is a gap between demonstration capability and production reliability. Robots that perform impressively in structured demos with controlled lighting, known objects, and rehearsed tasks often encounter significant failure rates in the variable conditions of a real factory. The companies solving that gap — Figure, 1X Technologies, Agility Robotics — are doing so more slowly than 2024 timelines implied.