Boston Dynamics Opens Atlas Training Center at Hyundai Metaplant in Georgia

Boston Dynamics has opened a training and testing center for its Atlas humanoid robot at Hyundai Motor Group Metaplant America near Savannah, Georgia, putting the machine into hands-on work inside an active automotive manufacturing environment. The Robotics Metaplant Application Center, known as RMAC, is designed to prepare Atlas for real production tasks rather than controlled laboratory demonstrations. The company announced the opening on September 21, 2026, marking a step from research prototypes toward factory-floor deployment at scale.

The center represents a deepening of the relationship between Boston Dynamics and Hyundai Motor Group, which wholly owns the robotics company. Hyundai Motor Group Metaplant America, or HMGMA, is a large manufacturing complex in Georgia. By placing Atlas inside that plant, Boston Dynamics is testing whether humanoid robots can handle the complexity, variability and pace of automotive production. The effort is part of a broader industry push often described as physical AI, where machines learn from real-world data and physical interaction.

Atlas is not starting with the most difficult assembly jobs. According to the companies, the robots are initially learning logistics tasks, such as preparing and sequencing automotive parts before placing them in the correct order for assembly. That work involves moving parts, understanding packaging and respecting the rules of a working factory. Over time, the scope is expected to broaden into component assembly and other tasks.

RMAC functions as both a testbed and a training center. It is meant to operate in strong coordination with the auto plant, so that lessons from the factory floor feed directly into robot behavior. Boston Dynamics has described the facility as a Robot Behavior Factory, a place that takes real parts, fixtures and workcells, adds Atlas robots and skilled pilots, and produces diverse, high-quality data for training. That framing matters because it suggests the center is not a one-off pilot but an ongoing data engine.

Key Facts

Boston Dynamics announced on September 21, 2026 that it opened the Robotics Metaplant Application Center at Hyundai Motor Group Metaplant America near Savannah, Georgia. The first phase of RMAC is now operational at HMGMA. The launch follows the unveiling of Atlas at HMG Media Day at CES 2026. The center is training Atlas robots in real-world environments, including the logistics and sequencing of automotive parts before they are placed in the correct order for assembly.

The deployment numbers are substantial. Hyundai Motor Group plans to deploy 25,000 Atlas units across Hyundai Motor and Kia plants globally over the next few years. To support that, the Group intends to establish a new production facility in the United States capable of producing 30,000 robots annually. The first phase of RMAC is already running, and Boston Dynamics plans to move operations into a new building at HMGMA next year, expanding RMAC to approximately ten times its current size.

Training methods combine multiple approaches. The Robot Report reported on September 21 that Atlas learns from teleoperation, reinforcement learning in simulation, and Universal Manipulation Interface devices. The tasks require high dexterity. Car parts must be handled directly because they have complex surfaces. They must respect complex grasping rules for cosmetic or fragile elements. They often need bimanual manipulation because they are large, awkwardly weighted or fragile. The work also involves extraction from shipping packaging and insertion into high-density cubbies for transport to the line.

Zack Jackowski, Chief Product and Technology Officer at Boston Dynamics, framed the effort in broad terms. "Together with Hyundai Motor Group, we are pioneering physical AI at enterprise scale," he said. "We have established a clear strategy to optimize our supply chain, production, testing, and deployment capabilities for Atlas." Jackowski also told The Robot Report that the earlier trials were time-boxed deployments focused on lessons needed to design a scaled-up presence. RMAC, by contrast, is a permanent facility focused on both data collection and application testing. He added that the design assumptions made for the production version of Atlas have been very close to right.

The timeline extends over several years. Seoul Economic Daily reported on September 22 that Atlas is set to join HMGMA's manufacturing processes starting in 2028, beginning with parts sorting and expanding to parts assembly by 2030. The robot will also handle high-intensity tasks such as repetitive motions and lifting heavy objects. The AI Insider reported on September 22 that Boston Dynamics will begin exploring Atlas uses outside automotive next year and is in discussions with existing Spot and Stretch customers in manufacturing, aerospace, semiconductors, logistics, food and beverage, and life sciences to expand data collection and other tasks supporting Atlas training in 2027.

Analysis

What this really means is that Boston Dynamics is using Hyundai's own factory as a proving ground for physical AI at enterprise scale. The staged approach is deliberate. Atlas begins with parts logistics and sequencing, which are lower-risk roles compared with tasks closer to the vehicle build. Automotive World reported on September 22 that starting with logistics tasks suggests a carefully planned deployment path, proving Atlas in lower-risk roles before trusting it with tasks closer to the vehicle build. That sequencing reduces the chance that early failures damage confidence or production schedules.

The bigger picture here is that the 30,000-unit US factory points beyond Hyundai's internal needs. Automotive World's analysis notes that a domestic factory at that scale signals manufacturing-scale ambition beyond internal use, suggesting Boston Dynamics is preparing to supply humanoid robots to customers outside the Group. Hyundai's own plan calls for 25,000 Atlas units across Hyundai Motor and Kia plants, so a 30,000-robot annual capacity implies additional demand. Early cross-industry discussions with Spot and Stretch customers hint that automotive is the proving ground for Atlas, not its exclusive long-term market.

The competitive context is intensifying. Seoul Economic Daily reported that Tesla is training its Optimus humanoid at its Fremont, California plant, and Japan's Toyota has announced plans to deploy 400,000 units of its in-house humanoid Ellie at production sites starting in 2028. Those figures show that major automakers see humanoid robots as a strategic technology. Boston Dynamics is differentiating through deep integration with an existing manufacturing operation and through a training facility that generates real production data. The Robot Report's description of RMAC as a Robot Behavior Factory captures that advantage: real parts, real fixtures and real workcells feed the learning process.

Challenges remain. High-dexterity tasks such as handling cosmetic or fragile parts, bimanual manipulation and insertion into high-density cubbies are difficult. Training methods must generalize from teleoperation and simulation to the unpredictable conditions of a working plant. Jackowski's comment that production design assumptions have been close to right is encouraging, but it also implies that the next phase will test those assumptions under sustained production pressure. The move to a building ten times larger in 2027 will provide more space and more data, but it will also raise expectations.

Why It Matters

For Hyundai Motor Group, the RMAC and the planned deployments could reshape how its factories operate. A fleet of 25,000 Atlas units across Hyundai and Kia plants would be one of the largest humanoid robot deployments ever attempted in manufacturing. The robots are intended to take on repetitive motions and strenuous work such as lifting heavy loads, which could improve safety and productivity. Jackowski said RMAC reinforces a shared ambition to create a future where robots amplify human potential by making work safer and more productive.

For the United States, the planned production facility capable of building 30,000 robots annually represents a significant manufacturing investment. Georgia is already home to HMGMA, and the expansion of RMAC there, including the move to a building about ten times its current size in 2027, deepens the region's role in robotics. The center also supports hiring, as Boston Dynamics has moved staff from Waltham, Massachusetts, to Savannah, Georgia, and is hiring. That flow of talent and investment links traditional automotive manufacturing with advanced robotics.

For the wider robotics industry, the project is a test of whether humanoid robots can move from demonstrations to dependable industrial work. The combination of real factory data, staged deployment and cross-industry interest from aerospace, semiconductors, logistics, food and beverage, and life sciences suggests that lessons learned at HMGMA could travel well beyond cars. If Atlas succeeds in parts logistics and then assembly, it will provide a template for other companies considering humanoid labor.

Next Up

In 2027, Boston Dynamics plans to move RMAC into a new building at HMGMA, expanding it to roughly ten times its current size. The same year, the company expects to begin exploring Atlas use cases outside automotive and to expand data collection with existing Spot and Stretch customers in manufacturing, aerospace, semiconductors, logistics, food and beverage, and life sciences. These discussions are aimed at supporting Atlas training and other tasks.

In 2028, Atlas is set to join HMGMA's manufacturing processes, starting with parts sorting. By 2030, the applications will extend to component assembly. Over time, the robots will also take on repetitive motions and heavy lifting. Hyundai Motor Group's broader plan includes 25,000 Atlas units across Hyundai Motor and Kia plants globally over the next few years and a US factory capable of producing 30,000 robots annually. The first phase at RMAC is already operational, making the next few years a critical proving period for Atlas and for physical AI in manufacturing.