Image: Nvidia
Nvidia is making a major push into autonomous driving, unveiling what CEO Jensen Huang calls the world’s first “thinking, reasoning” autonomous vehicle AI — and it’s hitting U.S. roads later this year.
At CES 2026, Nvidia announced Alpamayo, a next-generation autonomous driving model that introduces Vision-Language-Action (VLA) AI to production vehicles, starting with the all-new Mercedes-Benz CLA. According to Nvidia, Alpamayo is trained end-to-end, taking raw camera input and generating driving actions while also exposing the reasoning behind each decision.
Alpamayo sits at the core of Nvidia’s newly launched DRIVE AV software stack, which is debuting in the Mercedes-Benz CLA in the U.S. before a broader rollout across the automotive industry. Nvidia says the system supports advanced Level 2 point-to-point driving, including complex urban navigation, proactive collision avoidance, and automated parking — all powered by large reasoning models and a dual-stack safety architecture.
“With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision,” Nvidia explained, noting that the model’s open weights and tools allow developers to adapt it for production vehicles and validation systems.
Ali Kani, Vice President of Automotive at Nvidia, framed the launch as a major milestone for AI-defined vehicles, saying: “As the automotive industry embraces physical AI, NVIDIA is the intelligence backbone that makes every vehicle programmable, updatable and perpetually improving through data and software. Starting with Mercedes-Benz and its incredible new CLA, we’re celebrating a stunning achievement in safety, design, engineering and AI-powered driving that will turn every car into a living, learning machine.”
The CLA is the first Mercedes vehicle built on the new MB.OS platform and recently earned a five-star Euro NCAP safety rating, with Nvidia-powered active safety features contributing to that result. Nvidia’s DRIVE AV uses an end-to-end AI driving stack paired with a parallel classical safety system built on its Halos safety framework, adding redundancy and guardrails for real-world deployment.
The announcement also comes as Nvidia revealed its next-generation Rubin AI chips, which Tesla CEO Elon Musk described as “a rocket engine for AI.” Tesla relies heavily on Nvidia GPUs to train both Full Self-Driving and Optimus, underscoring how deeply Nvidia remains embedded across the autonomy and robotics landscape — even as it now powers what could become a direct competitor to Tesla’s FSD.
With Alpamayo, Nvidia is positioning itself not just as a hardware supplier, but as a full-stack autonomy platform provider — one that could shape how future self-driving systems think, reason, and explain themselves on public roads.
