Hyundai partners with Nvidia on a unified AI platform for autonomous cars and robots

Hyundai is taking the most ambitious technological step in its history by striking a strategic partnership with Nvidia. The Korean group is purchasing 50,000 Blackwell GPUs and investing €2.78 billion to build its own AI data center. This will let the company assemble, for the first time, a full-fledged platform for autonomous driving and robotics without leaning on outside suppliers. The decision clearly points to a desire to own the critical computing core rather than patching together third-party parts.

The new graphics processors can execute nearly two quintillion operations per second, creating a compute backbone on par with what Tesla uses. Unlike its rival, Hyundai plans to make up for limited real-world road data with high-precision virtual modeling via Omniverse and Cosmos, training its algorithms on millions of simulated scenarios. In practice, a simulation-first approach can speed up iteration and expand edge-case coverage—if those synthetic miles translate reliably to the street.

At the same time, the company is widening its robotics ambitions. The Isaac Sim virtual environment will be used to train future humanoid robots, while the Drive AGX Thor platform will serve as a single computing nucleus for both cars and mechanical assistants. Hyundai has already approved plans for a U.S. factory dedicated to robotics with an annual capacity of 30,000 units. A shared “brain” across products simplifies software upkeep and hints at a tighter, more unified roadmap.

The infrastructure now taking shape will knit vehicles and robots into one network, with data flowing continuously back to the AI center to refine models. Hyundai expects that by 2028 its cars will carry high-performance computers co-developed with Nvidia—hardware designed to underpin its driver-assistance today and fully autonomous systems tomorrow. If the company hits that target, it shifts from fast follower to a true systems player.