Nvidia has introduced the Alpamayo artificial intelligence (AI) platform for autonomous vehicles, detailed its next generation of chips, and announced expanded partnerships for its DRIVE Hyperion platform. These developments aim to advance self-driving capabilities by integrating reasoning functions, enhancing safety, and providing open-source tools for developers. The initial Alpamayo announcement occurred at the Consumer Electronics Show (CES) in Las Vegas, with further details and partnership expansions later presented at Nvidia's GTC conference.
Alpamayo AI Platform for Autonomous Vehicles
Nvidia unveiled 'Alpamayo,' an AI platform designed to enable autonomous vehicles to perform reasoning functions, operate in complex environments, and provide explanations for their driving actions. The company states that Alpamayo introduces 'chain-of-thought reasoning' to autonomous vehicles, integrating visual input with language-like reasoning processes to assist vehicles in handling unexpected situations such as sudden roadworks or unusual driver behavior.
Jensen Huang, Nvidia CEO, characterized this development as a "ChatGPT moment for physical AI," highlighting a progression in machines' ability to comprehend, reason, and act in real-world settings.
The platform aims to enhance safety and scalability for applications like robotaxis by allowing vehicles to reason through rare scenarios and explain driving decisions.
Open-Source Access
Alpamayo is an open-source AI model, with its underlying code, known as Alpamayo 1 (a 10-billion-parameter VLA model), available on the machine learning platform Hugging Face. This allows autonomous vehicle researchers free access to retrain the model and use it as a foundation for development tools.
Alpamayo 1.5
A subsequent upgrade, Alpamayo 1.5, was introduced as an interactive, steerable reasoning model. It processes driving video, ego-motion history, navigation guidance, and natural language prompts to generate driving trajectories alongside reasoning traces. This version enables developers to modify vehicle behavior and specify constraints through prompts and navigation settings, and supports multi-camera configurations.
Vehicle Integration
Nvidia has partnered with Mercedes-Benz to produce a driverless car, the Mercedes-Benz CLA, powered by Alpamayo technology. This vehicle is scheduled for release in the United States in the coming months, with subsequent rollouts planned for Europe and Asia. A demonstration video showed an AI-powered Mercedes-Benz operating in San Francisco without manual driver intervention.
DRIVE Hyperion Platform and Expanded Partnerships
Nvidia announced increased adoption of its DRIVE Hyperion platform for Level 4 autonomous vehicle development and robotaxi services. The platform integrates chips, computers, sensors, and software necessary for advanced autonomous capabilities.
Automakers
BYD, Geely, Isuzu, and Nissan are developing Level 4 autonomous vehicles using the NVIDIA DRIVE Hyperion platform. BYD and Geely have joined Nvidia’s robotaxi program, with Geely incorporating Nvidia’s Thor chips into new Zeekr vehicles. Nissan is also developing robotaxi software with Wayve.
Mobility Providers
Uber and Nvidia announced an expanded partnership to launch a global robotaxi network across 28 markets and four continents by 2028, with initial rollouts planned for Los Angeles and San Francisco in the first half of 2027. Uber is collaborating with automakers like Lucid, Volkswagen, and Stellantis, which are building autonomous vehicles using Nvidia’s products. Lyft, Bolt, Grab, and TIER IV are also leveraging DRIVE Hyperion for global robotaxi development. Waymo additionally employs Nvidia’s products in its vehicles and cloud infrastructure.
Safety and Development Tools
Nvidia has introduced new initiatives to address safety and facilitate the development of autonomous vehicle systems.
Halos OS
The NVIDIA Halos OS has been introduced as a unified safety architecture for production-ready, scalable Level 4 autonomy on DRIVE Hyperion. Built on ASIL D-certified DriveOS foundations, it integrates safety middleware and deployable safety applications. The system is designed to intervene if AI models are poised to make unsafe decisions, guiding the vehicle to a safe state even if components fails. Companies including AEye, Flex, Gatik, Hesai, Lucid, MIRA, PlusAI, Qt Group, Saphira, and Valeo are joining the NVIDIA Halos AI Systems Inspection Lab.
AlpaSim
An open-source, end-to-end simulation framework available on GitHub, AlpaSim offers realistic sensor modeling, configurable traffic dynamics, and scalable closed-loop testing environments for AV development and policy refinement.
Physical AI Open Datasets
Nvidia provides a large-scale, open dataset comprising over 1,700 hours of driving data, collected across various geographies and conditions, including rare real-world edge cases. These datasets are available on Hugging Face.
Omniverse NuRec
For testing and validation of reasoning-based autonomous vehicles, NVIDIA Omniverse NuRec was announced. This set of 3D Gaussian Splatting technologies reconstructs and renders interactive simulations from real-world data. NuRec will be available on the NVIDIA NGC catalog and has been integrated into simulation solutions by providers such as 51WORLD, dSPACE, and Foretellix.
Next-Generation Chips
Nvidia also provided details on its next generation of chips. Jensen Huang announced that Nvidia's next generation of chips is in full production and is stated to deliver five times the computing power of previous Nvidia products for AI applications.
Vera Rubin Platform
Further details were disclosed regarding the new 'Vera Rubin platform,' expected to debut later this year. This platform consists of six separate Nvidia chips. The flagship server configuration will reportedly include 72 of the company’s graphics processing units (GPUs) and 36 of its new central processing units (CPUs). These can be configured into 'pods' containing over 1,000 Rubin chips, potentially improving the efficiency of generating AI 'tokens' by a factor of 10. The Rubin chips are reported to utilize a proprietary data type.
Industry Context and Reception
Analysts, including Paolo Pescatore of PP Foresight, commented that the announcements reinforce Nvidia's position in integrating AI hardware and software, marking a strategic shift towards becoming a platform provider for physical AI ecosystems.
Elon Musk, CEO of Tesla, responded to the Alpamayo announcement on social media, noting similarities with Tesla's Autopilot driver assistance software. Musk stated that achieving 99% functionality is straightforward, while resolving the remaining edge cases presents significant challenges.
Nvidia's automotive division generated $592 million in revenue out of a total of $51.2 billion in Q3 2025. The company faces competition in the market for delivering AI products to end-users from entities such as Advanced Micro Devices and Google. Expanded agreements with Chinese automakers BYD and Geely occur amidst ongoing trade discussions between the U.S. and China. While China holds a lead in electric vehicle production, the U.S. and China are more closely matched in the robotaxi sector, with Baidu operating in China and Waymo in the U.S.
Nvidia's stated long-term objective is for all cars and trucks to eventually become autonomous. The company also intends to launch its own robotaxi service by next year in partnership with an undisclosed entity. Public concerns about autonomous vehicle safety, stemming from incidents involving other manufacturers, are also a factor in the market.