Hyundai Motor Group is pushing back the launch of its in-house autonomous driving software by roughly two years, opting to lean on Nvidia technology in the meantime as it races to close the gap with Tesla and Chinese automakers.
The South Korean conglomerate now expects to roll out vehicles running its proprietary Atria AI platform in the second half of 2029, according to executives who briefed reporters at 42dot's headquarters in Seongnam on Sept. 11. The original target had been late 2027.
Before that happens, Hyundai will ship vehicles powered by Nvidia's Hyperion 10 platform, with Level 2+ driver-assistance capabilities arriving in the first half of 2028 and more advanced Level 2++ functions following in the second half of the year. Level 2+ generally covers hands-off driving on highways under driver supervision, while Level 2++ extends that capability to urban roads.
"Our partnership with Nvidia is not about leaving our destiny entirely in their hands," Park Min-woo, president of Hyundai Motor Group's Advanced Vehicle Platform division and CEO of 42dot, told reporters. He said Hyundai will co-design the technology with Nvidia and use data harvested from those systems to train and refine Atria.
Park, a former Nvidia executive who joined Hyundai in January, has driven the automaker's deepening relationship with the chipmaker. His approach marks a departure from predecessor Song Chang-hyeon, who had emphasized internal software development before leaving abruptly in December.
The strategic shift underscores how difficult it has become for traditional automakers to build competitive autonomous driving stacks from scratch. Hyundai and affiliate Kia Corp together rank as the world's third-largest automaker with annual sales exceeding 7 million vehicles, but that scale has not yet translated into a leadership position in vehicle autonomy.
Executives framed the delay as a deliberate choice rather than a setback. "We could release Atria AI sooner, but we strategically delayed the timeline," Park said. "Rather than rushing mass production to roll out a half-baked Level 2++, we decided to proceed with mass production using Nvidia's technology while focusing resources on the final goal."
At the core of Hyundai's strategy is what the company calls a "Data Flywheel" — a self-reinforcing cycle in which vehicles collect real-world driving data, that data trains AI models, improved models are deployed back to vehicles, and the cycle repeats. Hyundai believes its massive global fleet will let it accumulate data faster than rivals once autonomous systems reach mass production.
Kwon Jeong-hyeon, head of the group's Autonomous Driving Development Center, said Hyundai currently operates about 40 dedicated data-collection vehicles around the clock. But once production vehicles equipped with autonomous systems hit the road, the data volume should grow exponentially. The company expects to overtake competitors in accumulated driving data by 2033, roughly five years after mass production begins in 2028.
"The side that first completes the data loop from collection to learning, and from model training to model deployment, will have a competitive edge in automotive software and AI," Kwon said.
Hyundai is also unifying its data infrastructure across business units. The group's Data Union framework will standardize sensor specifications across Hyundai, Kia, 42dot and Motional — Hyundai's U.S. robotaxi joint venture — using Nvidia's Hyperion 10 platform. That alignment should allow the companies to share data more efficiently despite previously operating with different sensor stacks.
Several technical initiatives support the flywheel concept. Hard example mining automatically identifies driving scenarios the AI finds difficult. Continuous training feeds those examples back into the model to shorten development cycles. A special event recorder captures data when the system detects unusual situations, such as avoiding a vehicle stopped on a highway shoulder, or when a driver disengages autonomous mode.
Hyundai vehicles using Nvidia's Hyperion 10 platform will initially skip expensive lidar sensors in favor of cameras, ultrasonic sensors and a radar. However, the company is evaluating lidar for Level 3 systems, which would allow hands-off driving under specific conditions.
"To truly ensure safety, we believe sensor redundancy is critical: if one sensor fails, others must be able to maintain safe operation," Park said, without offering a timeline for Level 3 commercialization.
Beyond the Nvidia-based rollout, Hyundai is developing a Vision-Language-Action model that combines visual perception, language-based reasoning and action generation in a single system. The approach could improve the AI's ability to handle edge cases and explain its decisions, since language-based reasoning adds a layer of interpretability that purely visual end-to-end models lack.
Hyundai showcased the progress at the media event with video of an Ioniq 6 testbed vehicle equipped with Atria AI navigating dense urban traffic in Seoul's Gangnam district during rush hour and in Pangyo during rain. The vehicle, fitted with eight cameras and one radar, handled stopped vehicles on the shoulder, sudden cut-ins and oncoming traffic on narrow roads without driver intervention.
The company also plans to begin real-world testing in Gwangju by year-end, working with South Korea's Ministry of Land, Infrastructure and Transport to deploy autonomous test vehicles on public roads and gather large-scale validation data.
The Nvidia collaboration extends beyond driver assistance. Hyundai's relationship with the chipmaker now spans autonomous driving, AI data centers and humanoid robots developed by Boston Dynamics, the robotics company Hyundai acquired in 2021.
The competitive stakes are significant. Tesla's Full Self-Driving system has accumulated billions of miles of real-world data, while Chinese automakers including BYD and XPeng have moved aggressively to deploy advanced driver-assistance features across their lineups. Hyundai's dual-track approach — using Nvidia to enter the market quickly while building Atria for the long term — reflects a bet that its global manufacturing scale can eventually close the data gap.
Note: Level 2+ covers hands-off highway driving under driver supervision; Level 2++ extends to urban environments.
For Hyundai, the path forward hinges on execution. The company must deliver Nvidia-based systems on schedule in 2028 while simultaneously maturing Atria for its 2029 debut. The data flywheel strategy only works if millions of connected vehicles actually feed information back into Hyundai's AI pipeline — and if the engineering team can turn that data into demonstrably safer, more capable autonomous driving.
Park framed the challenge in terms of trust. "Autonomous driving competition is no longer about comparing specific features," he said. "Hyundai Motor Group will develop autonomous driving technology that customers can trust, based on a virtuous cycle of data, AI and validation."
Once added, BigGo Finance appears first in Google Search Top Stories, so you get the broadest, most up-to-the-minute, and most comprehensive global financial news first.