Hyundai Motor Group puts its AI Data Flywheel into full operation, targeting Level 2++ production by 2029

Track One: NVIDIA collaboration

Integrates NVIDIA's validated vehicle AI computing platform and autonomous driving software into the Group's SDV architecture; Level 2+ production targeted for the first half of 2028 and Level 2++ for the second half of 2028; sensor systems standardized across Hyundai Motor, Kia, 42dot and Motional around NVIDIA DRIVE Hyperion 10 for more consistent data collection and utilization.

Track Two: Atria AI

Proprietary end-to-end autonomous driving system jointly developed by the AVP Division and 42dot under an integrated development framework; production of Atria AI-powered Level 2++ vehicles targeted for the second half of 2029, with capability progressing based on real-world driving data collected from production vehicles.

More than 7 millionvehicles

Sold by Hyundai Motor and Kia across approximately 190 countries and regions, the starting advantage the Group cites for autonomous driving data collection.

• Hard Example Mining: automatically identifies challenging driving situations, or edge cases, that AI models find difficult to recognize or interpret, and prioritizes them for training.
• Continuous Training Pipeline: continuously incorporates newly acquired real-world driving and validation data into model training, and feeds vehicle evaluation findings back into data collection and model development to shorten development cycles.
• Virtual Validation: reconstructs real-world driving data into three-dimensional environments using technologies such as 3D Gaussian Splatting to recreate scenarios that are difficult or unsafe to reproduce in real testing, and to check that newly trained models do not degrade existing performance.
• Follow-the-Sun Development: connects development centers in South Korea and the U.S. so teams sequentially carry out data collection, issue analysis and model improvement for continuous 24-hour development.
• Special Event Recorder (SER) Integration: automatically records and stores significant events and related data during autonomous driving for use in AI model training and performance improvement, with the Group exploring ways to enhance it.
• Data Union: a framework based on standardized sensor architectures and data structures that lets data from multiple vehicles and organizations accumulate under common standards for AI training, initially focused on Hyundai Motor, Kia, 42dot and Motional.

• Executive Ride-Along: Minwoo Park and Seonggyun Jeong traveled through central Seoul as the vehicle navigated expressways, major thoroughfares, bridges and urban streets, discussing Atria AI's development process, technical capabilities and decision-making mechanisms.
• One-Take Urban Driving: three unedited sequences of roughly two to four minutes each, filmed without edits other than playback speed adjustments, covering congested Gangnam morning rush-hour traffic, high-density Jamsil traffic with frequent interactions with large vehicles such as buses, and rainy urban driving in Pangyo.
• Edge Case Handling: ten representative scenarios, including avoiding vehicles parked along the roadside, responding to sudden vehicle cut-ins, navigating unprotected left turns, detecting pedestrians in congested areas and identifying oncoming vehicles on narrow neighborhood roads.