NVIDIA INTRODUCES ALPAMAYO AS PART OF ITS PHYSICAL AI STRATEGY

Nvidia used CES 2026 to outline a new phase of its autonomous driving efforts, unveiling Alpamayo, a family of AI models designed to support decision-making in self-driving vehicles.

- 

The system is positioned as part of Nvidia's broader “Physical AI” strategy, which focuses on applying large-scale artificial intelligence models to machines operating in real-world environments, where uncertainty, safety, and edge cases remain persistent challenges.

Alpamayo is built around a 10-billion-parameter architecture and is intended to improve how autonomous vehicles interpret complex situations on the road. Nvidia said the model is trained on extensive real driving data and is designed to make its decision processes more transparent, an area that has drawn growing attention from regulators, automakers, and safety researchers.

The announcement reflects Nvidia's continued effort to move beyond supplying hardware alone, offering software platforms and data resources that automakers and developers can adapt for their own autonomous systems.

MODEL DESIGN, DATA, AND OPEN ACCESS

Nvidia said Alpamayo relies on a large corpus of real-world driving data, with more than 1,700 hours of recorded driving used to train and validate the system. That dataset is being released as part of Nvidia's Physical Al Open Datasets, alongside simulation tools intended to help developers test and refine autonomous behaviors before deploying them in vehicles.