During his keynote at GTC this week, Nvidia CEO Jensen Huang declared that "the Chat-GPT moment for self-driving cars has arrived."
Following up on the open autonomous platform the company teased back in January, Nvidia has officially unveiled a massive expansion of its self-driving ambitions. This includes a fleet of new automotive partners, a massive update to its underlying AI architecture, and a direct shot at Tesla's Cybercab: a global robotaxi network.
The Uber Robotaxi Network
The biggest news from the presentation was the announcement that Nvidia is partnering with ride-hailing giant Uber to launch a global network of Level 4 autonomous robotaxis.
The service is slated to begin rolling out in Los Angeles and San Francisco in early 2027. But Nvidia and Uber aren't planning to stop on the West Coast.
The aggressive roadmap outlines a rapid expansion to 28 cities by 2028. This move changes the landscape of the robotaxi race, transforming Nvidia from a hardware supplier into a direct competitor against Tesla's upcoming autonomous fleet.
A Massive Automotive Alliance
To put these robotaxis on the road, Nvidia is continuing to aggressively license its technology to legacy automakers and EV startups alike. At GTC, Huang announced that the company is officially expanding its self-driving partnership to include BYD, Nissan, Hyundai, and Geely.
This adds to an already staggering roster of automotive partners utilizing Nvidia's platform, which currently includes GM, Toyota, Mercedes-Benz, Jaguar Land Rover, Volvo, Rivian, XPENG, Polestar, NIO, Lucid, Li Auto, Isuzu, Zeekr, Xiaomi, and Stellantis.
While Tesla executives like Elon Musk and Ashok Elluswamy have previously weighed in on the extreme difficulties of creating a generalized self-driving system that works across dozens of different vehicle architectures, Nvidia is betting heavily that its platform can become the universal brain for the rest of the automotive industry.
Alpamayo 1.5: 10-Billion Parameters
Underpinning all of these partnerships is a massive update to Nvidia's autonomy software stack. Dr. Marco Pavone, Nvidia's Director of Autonomous Vehicle Research, detailed the release of Alpamayo 1.5 to the developer community.
Building on the foundation introduced at CES earlier this year, Alpamayo 1.5 is an open, 10-billion-parameter chain-of-thought reasoning model. Built on the Cosmos-Reason2 backbone and post-trained with reinforcement learning, this new model adds crucial capabilities for automakers.
Because Nvidia has to account for so many different vehicle designs from its partners, Alpamayo 1.5 now natively supports highly flexible, configurable multi-camera setups. Pavone described the result as an "interactive, steerable reasoning engine."
Nvidia is also expanding its open data platform by releasing highly requested reasoning labels for its PhysicalAI Autonomous Vehicles dataset, giving developers the tools they need to train and adapt the model for their specific fleets.
With a 10-billion-parameter reasoning engine, an army of legacy manufacturing partners, and the logistics network of Uber at its disposal, Nvidia has made it clear that the robotaxi race is officially well underway.
NVIDIA’s Keynote
You can watch Nvidia’s keynote below. The autonomous vehicle section begins at the 2-hour, 6-minute mark.
