LAS VEGAS, January 06 (AJP) -Nvidia unveiled a next-generation “reasoning AI” for autonomous driving at CES 2026 on Monday, signaling a deeper push into the self-driving vehicle market with real-world deployments planned as early as the first quarter of this year.
At a press conference at the Fontainebleau Live Theater in Las Vegas, Nvidia CEO Jensen Huang introduced the technology, dubbed Alpamayo, describing it as “the world’s first self-driving AI that thinks and reasons.”
“The ChatGPT era of physical AI has arrived,” Huang said onstage. “Machines have begun to understand the real world directly and to reason and act on their own.”
Huang said Alpamayo is designed to address autonomous driving’s so-called “long-tail” problem — rare, complex scenarios that are difficult to anticipate through conventional data collection alone.
Unlike traditional systems that react primarily to sensor inputs, Alpamayo reasons step by step, Huang said, and can explain why it chose a particular action.
“Alpamayo not only takes sensor input to control steering, braking and acceleration, it also reasons about what action to take and tells you why,” he said. “It can respond safely even in complex long-tail scenarios.”
The model is a vision-language-action system built on Nvidia’s physical AI platform, Cosmos, which the company describes as a world foundation model. Nvidia said Alpamayo was trained using human driving demonstrations and trillions of miles of AI-generated data.
During the presentation, Huang showcased a live demonstration of a vehicle equipped with Alpamayo navigating downtown San Francisco, verbally explaining its decision-making process in real time.
“The long tail of driving can’t be solved by collecting every scenario in the real world,” Huang said. “But it can be solved by breaking it down into smaller, normal situations and reasoning through them.”
A video shown onstage featured an AI-powered Mercedes-Benz vehicle driving through San Francisco traffic while a passenger sat behind the steering wheel with their hands resting in their lap.
“It drives so naturally because it learned directly from human demonstrators,” Huang said. “In every single scenario, it reasons about what it’s going to do and why.”
Nvidia said it plans to release Alpamayo as open source, with the underlying code made available on the machine-learning platform Hugging Face, allowing autonomous-vehicle researchers to access and retrain the model.
Several companies and research groups have already lined up to use Alpamayo for Level 4 autonomous driving development, including Lucid Motors, Jaguar Land Rover, Uber and Berkeley DeepDrive.
“Robotaxis will be the first to benefit,” Huang said, adding that Mercedes-Benz vehicles equipped with Alpamayo are expected to begin operating on U.S. roads in the first quarter.
Nvidia also said it is planning a robotaxi pilot service in 2027 with partners including Uber. The company declined to name additional partners or specify locations for the rollout.
“Our vision is that someday, every single car, every single truck, will be autonomous,” Huang said.
Copyright ⓒ Aju Press All rights reserved.
![[CES 2026] Nvidias Jensen Huang, AMDs Lisa Su to outline AI strategies](https://image.ajunews.com/content/image/2026/01/05/20260105074115590653_278_163.jpg)
