Hyundai Motor Group told investors Tuesday that it plans to deploy more than 25,000 Atlas humanoid robots — developed by its US robotics subsidiary Boston Dynamics — across Hyundai and Kia manufacturing plants, absorbing 83 percent of the 30,000-unit annual production capacity the group is targeting by 2028. The announcement, made at a JPMorgan Chase investor session in Boston, is the first time Hyundai has disclosed a specific fleet figure for internal deployment — and it arrives as the Korean Metal Workers' Union has blocked Atlas from entering any Hyundai factory floor without a formal labor-management agreement.
25,000 Robots: From Production Target to Deployment Commitment
The JPMorgan session, held in Boston on Monday, brought together executives from six Hyundai affiliates — including Hyundai Motor, Kia, Hyundai Mobis, and Boston Dynamics, represented by interim CEO Amanda McMaster. Kia CEO Song Ho-sung confirmed at a separate overseas roadshow the same day that Atlas will first be deployed at Hyundai Motor Group Metaplant America in Savannah, Georgia, in 2028, with Kia's Georgia plant following in 2029.
The 25,000-unit figure represents a binding internal commitment: Hyundai will absorb the overwhelming majority of its own production output before any Atlas robots reach external customers. Additional customers are expected from 2027, when Boston Dynamics plans to open Atlas orders beyond its initial Hyundai and Google DeepMind deployments.
Hyundai is also building a US production facility for actuators through Hyundai Mobis — the automotive parts affiliate that already supplies Atlas's joints. Industry estimates place actuators at roughly 60 percent of each robot's material cost; the facility will have annual capacity of 350,000 units starting in 2028, enough to supply components for the full 30,000-robot production run per year.
What Atlas Actually Does — and How It Learned to Do It
Boston Dynamics published a technical blog on May 18, written by Alberto Rodriguez, Director of Robot Behavior for Atlas, alongside research engineers Shane Rozen-Levy and Vinay Kamidi, explaining for the first time the training methodology behind Atlas's ability to lift and carry loads exceeding 110 pounds.
The key advance is not raw strength — it is generalization. Earlier industrial robots handle fixed loads in controlled configurations. Atlas, by contrast, learned to pick up and carry an unfamiliar object without prior knowledge of its mass or center of gravity, using what Boston Dynamics calls proprioception: an internal body-awareness system that reads weight distribution, grip resistance, and balance across all joints simultaneously. As the engineers described in the blog, "the hardest part is not seeing the fridge or knowing how to lift it, but learning to adapt to whatever version of the fridge that Atlas will encounter in the real world."
The training pipeline works through GPU-parallel reinforcement learning simulation. Atlas practiced the lift with varied weights, floor friction values, grip conditions, and object positions across millions of simulated hours. The published demo uses a 50-pound mini-fridge; internal testing pushed the load past 110 pounds — above the training range — and Atlas adapted without additional training. That generalization is what roboticists call zero-shot sim-to-real transfer: a behavior developed entirely in simulation that deploys directly to physical hardware. The entire behavior was developed within weeks of Atlas's January 2026 debut.
The production robot that will perform this work has 56 degrees of freedom, fully rotational joints at the hips, waist, and neck, a 2.3-meter reach, and a certified maximum lift capacity of 50 kg (110 lbs). Its joints can rotate a full 360 degrees, allowing it to reposition loads in tight factory spaces without turning its entire body. Hyundai Mobis manufactures the actuators at automotive scale — the same supply chain that builds car components — which Boston Dynamics says enables the reliability and unit economics required for industrial deployment. The robot can autonomously swap its own batteries in under three minutes, keeping downtime negligible during continuous factory shifts.
Hyundai's Vertical Loop: From Parts to Production Line
Every piece of the Atlas deployment strategy runs through Hyundai's own infrastructure. Boston Dynamics develops the platform. Hyundai Mobis manufactures the actuators. Hyundai Motor Group operates the factories. The Robotics Metaplant Application Center (RMAC) — described by Boston Dynamics CEO Robert Playter as a "data factory" — serves as the training ground where Atlas units learn manufacturing tasks before being deployed at scale at the Savannah metaplant.
The RMAC's purpose is strategic: every robot trained there contributes to a proprietary dataset of humanoid manufacturing skills. Hyundai's deployment plan starts conservatively, with parts sequencing at the Savannah plant — which builds the Ioniq 5 and Ioniq 9 electric vehicles — and expands to component assembly and heavier repetitive work by 2030. Hyundai describes this approach as human-centered automation, with robots handling hazardous and repetitive tasks while human workers train, supervise, and maintain robotic systems.
Google DeepMind: Teaching Atlas What to Do Next
The second launch partner addresses Atlas's cognitive capability rather than its physical throughput. Google DeepMind is receiving a fleet of Atlas robots to integrate its Gemini Robotics AI foundation models — including Gemini Robotics 1.5, a vision-language-action model for direct robot control, and Gemini Robotics-ER 1.5, which handles higher-level planning and can access digital tools in natural language. "We developed our Gemini Robotics models to bring AI into the physical world," said Carolina Parada, Senior Director of Robotics at Google DeepMind. "We are excited to begin working with the Boston Dynamics team to explore what's possible with their new Atlas robot as we develop new models to expand the impact of robotics, and to scale robots safely and efficiently."
The partnership is not exclusive. Google has separately tested Apptronik's Apollo humanoids, and Boston Dynamics has prior AI research ties with Toyota Research Institute. But pairing Atlas's hardware with Gemini Robotics on a dedicated Atlas fleet is the most direct path available to expanding what the robot can learn and how quickly it can be reassigned to a new task.
Korean Union Draws Line Before First Robot Arrives
The labor conflict is specific and documented. The Hyundai Motor branch of the Korean Metal Workers' Union issued a public statement on January 22, 2026, declaring that Atlas will not be permitted to enter Hyundai factories without a labor-management agreement. The union cited the robot's anticipated unit cost — estimated at roughly 200 million Korean won (approximately $145,000) based on internal union calculations — as evidence that management views Atlas as a labor-cost reduction tool. "Keep in mind that not a single robot can enter the workplace without labor-management agreement," the union stated. At that price, the union calculated, a robot operating round-the-clock could replace the work of three human workers while costing less over two years than the wages of one.
Boston Dynamics CEO Robert Playter has publicly acknowledged that robots will take over some categories of work. "The really repetitive, really back-breaking labor is going to end up being done by robots," he told 60 Minutes. Hyundai Vice Chair Jaehoon Chang has maintained that human workers will shift to higher-value roles — training, supervising, and maintaining robotic systems — and that Atlas responds to chronic labor shortages in manufacturing rather than displacing existing positions. Both claims may prove simultaneously accurate: chronic shortages can be filled by robots while existing workers are still displaced when roles are eliminated rather than retrained into new ones. The union has identified the 2026 summer contract negotiations as the expected escalation point.
Competitive Landscape: First to Scale, Not Alone
Atlas arrives in commercial production alongside a field that has moved faster than most forecasters expected two years ago. Figure AI ran its Figure 02 robots during an 11-month pilot at BMW's Spartanburg, South Carolina plant, during which the robots helped produce more than 30,000 BMW X3 vehicles before the generation was retired in November 2025. Agility Robotics' Digit has moved over 100,000 totes at GXO Logistics warehouse facilities. Tesla's Optimus is in internal factory testing. Apptronik's Apollo is in pilot use at Mercedes-Benz facilities in Berlin and Hungary.
What distinguishes Atlas in the current field is payload capacity and joint range: 50 kg maximum and 56 degrees of freedom exceed published competitor specifications, and Atlas is the first of these platforms to have its full-body reinforcement learning methodology explained in public technical detail by the named engineers who built it — a disclosure that gives the broader research community a concrete benchmark.
Safety standards have not kept pace. The US Occupational Safety and Health Administration has no regulations specific to humanoid robots operating alongside human workers. ISO 25785-1, a new safety standard specifically for dynamically stable walking robots, is under development; the US delegation is led by Federico Vicentini of Boston Dynamics, Kevin Reese of Agility Robotics, and Carole Franklin of the Association for Advancing Automation, with publication expected in 2026 or 2027 at the earliest.
Hyundai's JPMorgan commitment makes the deployment timeline concrete for the first time. Whether 25,000 robots begin arriving in Hyundai and Kia factories on schedule — and whether Hyundai reaches a labor-management agreement before the 2028 deployment window opens — will determine whether this week's investor presentation marks the start of a new chapter in industrial automation, or another ambitious target that encounters the resistance every new category of factory tool eventually faces.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
