Hyundai Motor confident it will surpass rivals in autonomous driving data by 2033

Hyundai Motor Group, the world's third-largest automaker by sales, is moving aggressively to lead the autonomous driving AI race by leveraging driving data generated by the roughly 7 million vehicles it sells each year on roads around the world.

The strategy centers on unifying data that Hyundai Motor, Kia, 42dot and Motional have each accumulated separately, creating what the company calls a "data flywheel" — a virtuous cycle in which selling more vehicles generates more data, and more data makes the AI smarter.

Hyundai Motor unveiled the medium- and long-term strategy for software-defined vehicles and autonomous driving at its 2026 CEO Investor Day, held Wednesday at the Conrad Seoul hotel in Yeouido.

Park Min-woo, president and head of Hyundai Motor Group's Advanced Vehicle Platform division and CEO of 42dot, said the group's greatest advantage in scaling the data flywheel lies in its sales volume and business network. "We sell more than 7 million vehicles a year and have a network spanning more than 190 countries," he said.

He added that as SDV architecture and standardized sensor systems are applied to more production vehicles, real-world experience and data accumulated across diverse countries and road environments will grow rapidly. "Through this, we expect to surpass competitors in cumulative data volume by around 2033," he said.

Park said the key question is how quickly the company can increase the share of vehicles collecting and using data through a unified structure.

To that end, Hyundai Motor is building a closed-loop system in which AI learns from data generated by vehicles, and improved functions are then pushed back to those vehicles to collect even more data — a cycle the company has named the "data flywheel." The process repeats continuously: data collection, analysis, AI and service improvement, and over-the-air software updates, steadily raising AI performance with each iteration.

Park used ChatGPT and YouTube as examples to explain the concept. "What they have in common is that they get better the more you use them," he said. "Who can accumulate more real-world experience, identify the meaningful data within it, connect that data to AI training faster, and turn it into a better product — that will be the new standard of competition."

He said the automotive industry must undergo the same transformation. "Cars also need to evolve from products whose features are fixed at the time of delivery to products that learn and improve with use," he said. "Where vehicle performance, production scale and quality were once the pillars of competitiveness, what matters going forward is how quickly data can be accumulated and connected to AI training and service improvement."

In autonomous driving, Hyundai Motor plans to integrate data gathered separately by Hyundai Motor, Kia, 42dot and Motional. To do so, the group will standardize sensor systems across its subsidiaries based on the Nvidia ecosystem and build a "data union" that allows data from different vehicles and business units to be used under a consistent framework.

The AI training will draw on internally collected test-drive data, real-world driving data from customer vehicles, and data secured by partner companies. Issues and unexpected situations encountered during driving will be analyzed to identify root causes, and the findings will be fed back into the group's proprietary autonomous driving AI, Artria AI.

Artria AI is an end-to-end autonomous driving AI that integrates perception, decision-making and control into a single system. Hyundai Motor has set a goal of developing in-house the full autonomous driving lineup from Level 2+ through Level 4.

"The performance of autonomous driving AI depends on how thoroughly it can acquire and learn from rare but difficult-to-handle unexpected situations," Park said. "Hyundai Motor Group is focused on securing exactly that kind of edge-case data."

Starting later this year, the company will also pursue "rare data" collection on public roads. Artria AI will be deployed in a Gwangju autonomous driving pilot project to directly gather edge-case data that is difficult to encounter in ordinary driving — the kind of scenarios that determine the real-world capability of autonomous driving AI.

Full-scale data collection from production vehicles is set to begin in 2028. Through a strategic collaboration with Nvidia, Hyundai Motor plans to apply Level 2+ autonomous driving technology to its first SDV production vehicle, due for release that year. Standardized sensor systems and an AI computing platform will also be rolled out sequentially across production vehicles at the same time.

"Hyundai Motor Group's annual sales of more than 7 million vehicles and its global network will serve as the foundation for scaling the data flywheel," Park said. "Beyond sheer scale, we believe we can also secure differentiated competitiveness in terms of data quality and diversity."

Data collected will feed directly into the performance improvement cycle. Hyundai Motor will use a dashboard to analyze issues arising during driving, automated parking performance and the stability of key driving scenarios. When anomalies are identified, the relevant data will be used to retrain Artria AI. The improved AI is then quantitatively validated before being redeployed to production vehicles.

"When a key issue or anomaly is detected, we use the driving data we have secured to refine Artria AI and then verify how much the quantitative indicators have improved," Park said. "The improved Artria AI is then distributed to customers, allowing more of them to enjoy a safer driving experience."

Hyundai Motor aims to roll out Artria AI progressively across its production vehicle lineup, ultimately building a full autonomous driving range from Level 2+ through Level 4.

Hyundai Motor Group is also building large-scale computing infrastructure to handle the surge in data. The group plans to bring its Saemangeum AI data center online in 2029, when autonomous driving data is expected to begin growing in earnest. The facility will be built at a capacity of 100 MW and will be large enough to accommodate more than 50,000 GPUs.

Park described the Saemangeum AI data center as a means of processing and training on the vast data collected worldwide without relying on external cloud services. "It means internalizing the entire pipeline — from data collection and verification to deployment," he said.

The same strategy applies to generative AI operating inside vehicles. Hyundai Motor is rolling out its next-generation infotainment system, Pleos Connect, and generative AI agent Gleoo AI, starting with the facelifted Grandeur and other models.

The company analyzes which features customers use, how frequently, where they encounter friction, and in what contexts they activate specific functions — all to improve the AI and its services. Analysis of real customer data has shown that Gleoo AI's task completion rate and voice recognition success rate have both risen while response times have shortened. Improved features are then pushed back to vehicles via over-the-air updates.

"The competitiveness of the future mobility industry will be determined not by any single technology, but by the capability to acquire data, learn from it and connect it to the customer experience," Park said. "The capabilities we build up in this way will not be confined to vehicles — they will serve as a foundation that can extend over the long term into manufacturing, robotics and beyond."