Hyundai Physical AI: Atria AI, CODA, Gleo AI and Atlas
Hyundai Motor Group presented Atria AI, CODA, Gleo AI, Boston Dynamics robots and smart-factory tools as parts of a single Physical AI strategy.
Hyundai Motor Group has brought its Atria AI autonomous-driving system, new CODA electronic architecture, in-car LLM Gleo AI, Boston Dynamics robotics and tools for software-defined factories into one public technology framework. At the HMG Tech Talent Forum, the group said it intends to evolve from an automaker into a technology company working with Physical AI — artificial intelligence that acts not only on a screen, but also in vehicles, robots and manufacturing.
One of the most striking figures in the presentation was 7–8 million vehicles, which the group produces each year across different countries, climates and markets. Advanced Vehicle Platform head Minwoo Park describes that scale as one of Hyundai’s advantages in developing Physical AI. There is an important caveat, however: this does not mean that all of those vehicles are already sending road data to a neural network.
In the Data Flywheel now in operation, around 40 dedicated vehicles collect road information around the clock. Hyundai and Kia do sell more than 7 million vehicles a year in roughly 190 countries and regions, but the huge production fleet currently represents potential for future scaling rather than an active data source made up of millions of cars. A recent review of Data Flywheel and Atria AI explains this mechanism in more detail.
Atria AI itself is an end-to-end autonomous-driving system developed by Hyundai Motor Group and 42dot. Production vehicles with Atria AI Level 2++ are planned for the second half of 2029. An earlier track based on NVIDIA solutions is targeted for 2028: Level 2+ in the first half and Level 2++ in the second. The labels Level 2+ and Level 2++ are not separate levels in SAE J3016, which defines Levels 0 through 5; with Level 2 systems, the driver must continuously supervise the driving task.
At the forum, Hyundai showed that autonomous driving is only one layer of a much larger structure. CODA is the next E/E architecture for software-defined vehicles, Gleo AI handles conversational interaction in the cabin, while Spot, MobED and Atlas bring related perception and decision-making technologies into the physical world. In parallel, the company is developing agentic AI and digital tools for factories. The approach continues the auto industry’s broader shift toward software-defined vehicles and manufacturing.
The practical meaning of Physical AI is therefore broader than a conventional autopilot. Hyundai is trying to build a common loop in which real vehicles and robots provide experience, models are trained and validated virtually, and new algorithms return to cars and manufacturing. But the forum does not mean a new customer feature is launching now: the company did not announce when CODA or Gleo AI will reach specific production models. Official Hyundai Motor Group material