A strategic collaboration among VinFast, Autobrains, and NVIDIA is forging a next-generation Level 4 autonomous driving program for Southeast Asia, built on NVIDIA DRIVE Hyperion to scale robotaxi operations.

The alliance emphasizes an ecosystem approach to move from pilots to everyday, scalable robotaxi transportation while maintaining momentum for NVIDIA’s AI platform beyond data centers.

Executives stress accessibility of advanced mobility, practical on-road implementation in dynamic traffic, and real-time, on-vehicle reasoning as core goals.

Autobrains brings Agentic AI technology and a robust patent portfolio to enable autonomous driving on standard sensors and automotive-grade compute.

The press release provides background on robotaxis and related technologies, including FAQs detailing partner roles and the tech components involved.

The program tackles barriers in autonomous driving—cost, complexity, and real-world performance—through a modular, agent-based architecture that integrates vehicle systems with high-performance compute.

Aimed at Southeast Asia’s dense, varied traffic, the initiative seeks to make advanced autonomous driving more affordable and practical in challenging environments.

The modular approach combines vehicle integration, high-end compute, and Autobrains’ Agentic AI to reduce costs and improve real-world reasoning.

Autobrains CEO Igal Raichelgauz describes the project as the first Agentic Level 4 designed to handle real-world complexity at scale on production vehicles, enabling real-time on-vehicle reasoning.

The effort builds on prior collaboration, transitioning from Level 2++ capabilities to defined-condition full autonomy, following joint work since January and culminating in a June announcement.

Southeast Asia serves as a demanding validation environment due to dense traffic, diverse road behavior, and dynamic urban settings.

NVIDIA notes that software-defined vehicles can operate more safely and reliably in complex real-world conditions with iterative improvements over time.