Hyundai Delays Its Driving Software to 2029, Uses Nvidia for 2028

Hyundai Motor Group faces a two-year gap between its original software timetable and the revised arrival of its proprietary system. Vehicles using Hyundai’s Atria driver-assistance software are now planned for late 2029 rather than late 2027. To get supervised driving technology into production sooner, the group intends to introduce Nvidia-based systems during 2028.

The change creates a deliberate tension between speed and control. Hyundai gains earlier access to a production-ready computing and sensor platform, but it will depend on an outside technology supplier for the first rollout. Its stated strategy is not to abandon Atria. Hyundai plans to co-design the interim system with Nvidia, learn from mass production and feed the resulting vehicle data back into its own development program.

Hyundai detailed that broader approach in its September 13, 2026 announcement on an AI-powered data collection and development cycle. The company calls this process a Data Flywheel: vehicles collect information, developers use it for training and validation, and updated software returns to vehicles for further deployment and data gathering.

“Our partnership with Nvidia is not about leaving our destiny entirely in their hands,” Hyundai Motor Group president Park Min-woo said. Park, who joined Hyundai in January after working at Nvidia, described the chipmaker as an interim development partner while Hyundai continues building Atria.

The 2028 systems will still need an attentive driver

Hyundai plans to deploy Nvidia-based Level 2+ vehicles in the first half of 2028, followed by Level 2++ vehicles in the second half. These labels describe more capable forms of driver assistance, not vehicles that can operate without human oversight.

Level 2+ is intended to support more extensive highway assistance, while Level 2++ is aimed at more complex urban driving. Depending on the system and operating conditions, those capabilities may allow periods without the driver’s hands continuously on the wheel. They do not remove the driver’s responsibility to supervise the vehicle and be prepared to take control.

That boundary matters because Level 3 is a different operating proposition. A Level 3 system can perform the driving task under specified conditions, but Hyundai has not provided a commercialization timetable for that capability. The company is considering additional sensing for future Level 3 vehicles, yet that consideration should not be read as a confirmed production decision.

Cameras and radar come first, not lidar

The initial Hyundai vehicles built around Nvidia’s Hyperion 10 architecture are expected to use 10 cameras, one forward-facing radar and 12 ultrasonic sensors. The package will not initially include lidar, a more expensive sensing technology that measures distance using laser light.

This combination gives the system several types of environmental input. Cameras provide visual information, radar measures the position and motion of objects in difficult visibility conditions, and ultrasonic sensors cover short-range detection around the vehicle. The standardized hardware also has to support software updates over a vehicle life expected to span seven or eight years.

Hyundai is evaluating lidar for higher automation levels because redundant sensing can provide different ways to observe the same driving environment. “To truly ensure safety, we believe sensor redundancy is critical: if one sensor fails, others must be able to maintain safe operation,” Park said. That statement explains the design objective, but it does not establish that lidar has been selected for a production Hyundai or Kia model.

Fleet data is the bridge back to Atria

Hyundai and Kia sell more than 7 million vehicles annually, giving the group a potentially large deployment base. Data from Nvidia-based vehicles could help engineers identify difficult situations, validate updates and refine Atria before its planned launch in the second half of 2029. The earlier rollout should also give Hyundai experience with integrating sensors, computing hardware, software updates and factory processes at production scale.

Hyundai expects its fleet to help it surpass competitors in accumulated driving data by 2033. That remains a company forecast, not an achieved advantage. Vehicle sales alone do not automatically produce useful training information; the data must be collected consistently, filtered, labeled or otherwise processed, validated and converted into dependable software improvements.

The 2028 Nvidia deployment is therefore more than a placeholder, but it is not a replacement for Hyundai’s in-house program. It is a bridge intended to buy production experience and fleet data while Atria continues through development. For drivers, the immediate limit is simpler: both the Nvidia-based systems and the first Atria release remain supervised assistance, with the human still responsible behind the wheel.

By Jonathan Barrett — Editor for AMI’s future mobility and autonomous systems section, with two decades covering robotics, e-mobility, drone-vehicle convergence, and transport mechanical systems.