.jpg)
Boston Dynamics says Atlas humanoid robots are now training inside Hyundai Motor Group Metaplant America in Georgia, moving a deployment roadmap announced in January into factory preparation. The first work covers logistics and parts sequencing: placing components in the order an automotive assembly line needs them before planned production use from 2028.
The site is called the Robotics Metaplant Application Center, or RMAC. Hyundai said on August 26 that the center had opened in June; Boston Dynamics described its September 21 announcement as the launch of RMAC and said the first phase is operational. The meaningful change is therefore the start of Atlas training on defined manufacturing tasks, rather than the construction of a previously unknown facility.
RMAC is designed to sit between a robot demonstration and a production line. Hyundai says the center reproduces operating conditions, collects real-world data, and runs testing and verification before deployment. Boston Dynamics says Atlas is now preparing parts for logistics and sequencing, while Hyundai's earlier technical plan describes training around actions such as lifting, turning and recovering from an interrupted movement.
The companies intend to feed training data from RMAC together with operational data from Hyundai's software-defined factories, where production equipment and processes are managed through shared software and data. That loop is meant to expose the robot to factory variation, update its behavior, and repeat validation before a task reaches normal operations. Neither company disclosed the learning algorithm, supervision level, number of Atlas units in the center, or whether any training runs occur on an active production line.
The timetable remains staged. Hyundai plans to introduce Atlas to repetitive parts-sequencing work in 2028, then expand to component assembly by 2030 if process-by-process safety and quality checks pass. Boston Dynamics says RMAC will move into a building about ten times larger in 2027 and begin exploring tasks with existing Spot and Stretch customers in manufacturing, aerospace, semiconductors, logistics, food and life sciences.
The announcement does not report task-completion rates, cycle time, uptime, safety incidents or a comparison with human workers and fixed industrial robots. It therefore establishes an operating training program, not production readiness or economic advantage.