Nvidia expands Drive Hyperion platform adoption with BYD, Geely, Isuzu, and Nissan

Nvidia says several global automakers are expanding their development of Level 4 autonomous vehicles using the company’s Drive Hyperion platform, a hardware-and-software reference architecture designed to support autonomous-driving systems. Automakers, including BYD, Geely, Isuzu, and Nissan, are building next-generation vehicles based on the platform, while mobility providers such as Uber, Bolt, Grab, and Lyft are using the technology to support robotaxi programs.

The announcement was made during Nvidia’s GTC developer conference and highlights the company’s continued push into the autonomous vehicle technology sector. Nvidia’s Drive platform combines computing hardware, sensor integration, networking, and software components designed to support automated driving systems. The company positions the platform as a standardized architecture intended to simplify development and deployment across different vehicle programs.

According to Nvidia, standardizing on the Drive Hyperion platform enables manufacturers to reduce development complexity by leveraging a common reference design that integrates the key components for automated driving. The company says the approach can help partners shorten validation cycles and support fleet-scale learning by collecting and analyzing data across multiple vehicles operating with similar hardware and software configurations.

Level 4 autonomous vehicles are designed to operate without human intervention within specific conditions or environments. These systems can perform all driving tasks within a defined operational design domain, such as certain geographic areas or weather conditions. The technology is often associated with robotaxi fleets and automated transportation services that operate within mapped service zones.

Several automakers developing vehicles on the platform are working toward these types of applications. Nvidia says BYD, Geely, and Nissan are building Level 4 vehicle programs based on the Drive Hyperion architecture. Nissan’s effort is expected to incorporate autonomous-driving software from the British startup Wayve, which focuses on machine-learning–based driving systems that learn directly from data rather than relying solely on preprogrammed rules.

Commercial vehicle manufacturer Isuzu is also working with autonomous driving software company TIER IV to develop an autonomous bus platform. Their project uses Nvidia’s Drive AGX Thor system-on-a-chip, a high-performance computing processor designed to handle the large volumes of sensor data required for autonomous driving systems.

Beyond passenger vehicles, Nvidia’s platform is also being integrated into broader transportation services, including ride-hailing networks. The company says Uber is working with Nvidia to build an autonomous ride-hailing fleet powered by the full-stack Nvidia Drive autonomous vehicle software system.

Under the partnership, Nvidia and Uber plan to deploy autonomous vehicles in 28 cities across four continents by 2028. The rollout is expected to begin with Los Angeles and the San Francisco Bay Area during the first half of 2027. These vehicles will rely on Nvidia’s computing platform, AI models, and safety architecture to support automated driving capabilities.

Other mobility providers are also participating in the broader development ecosystem. Companies including Bolt, Grab, and Lyft are using the Drive Hyperion platform to support their own autonomous mobility initiatives. Nvidia says the participation of multiple ride-hailing and transportation companies reflects a broader shift toward software-defined robotaxi fleets built around standardized computing platforms.

As part of the announcement, Nvidia also introduced a new safety architecture called Halos OS. The operating system is designed to provide a unified safety framework for autonomous vehicles built on the DriveHyperion platform. Nvidia says the system is built on its ASIL D-certified DriveOS foundation and includes a three-layer safety architecture that integrates middleware and deployable safety applications.

The system also includes an active safety stack intended to meet five-star New Car Assessment Program (NCAP) requirements. Nvidia says the goal of the architecture is to ensure that autonomous-driving software running on the platform can be verified and validated to meet automotive safety standards.

To support safety validation and system development, several companies have joined the Nvidia Halos AI Systems Inspection Lab. Participants include sensor manufacturers, automotive technology companies, and simulation providers such as AEye, Hesai, Valeo, Flex, Gatik, PlusAI, and Qt Group. These companies will contribute to testing and validating AI-driven vehicle systems using the Halos safety architecture.

Nvidia also introduced an updated version of its autonomous driving AI model, Alpamayo 1.5. The model is part of a larger collection of AI models, datasets, and development tools intended to help developers build reasoning-based autonomous driving systems.

Alpamayo 1.5 is designed to process driving video, vehicle motion history, navigation instructions, and natural-language prompts. Based on these inputs, the system generates driving trajectories along with reasoning traces that describe how the AI model reached its decisions. Nvidia says the feature allows developers to analyze and refine how autonomous systems interpret road situations and respond to them.

The updated model also supports multiple camera configurations and adjustable sensor parameters. Nvidia says this capability allows the same AI driving stack to be adapted across different vehicle platforms without requiring significant redesigns of the underlying software architecture.

Another area Nvidia addressed is simulation and validation. The company introduced a technology called Omniverse NuRec, which is designed to create high-fidelity simulation environments for autonomous driving development. The system uses a technique known as 3D Gaussian Splatting to reconstruct digital environments from real-world data collected by vehicles and sensors.

Simulation environments are a key part of autonomous vehicle development because they allow engineers to test driving systems against rare or potentially dangerous scenarios that may be difficult to reproduce on public roads. Nvidia says NuRec allows developers to reconstruct real-world driving environments and replay them in interactive simulation, allowing engineers to analyze system behavior and train AI models under controlled conditions.

Several companies involved in autonomous vehicle development have already integrated NuRec into their toolchains. Simulation and testing providers, including 51WORLD, dSPACE, and Foretellix, have added support for the technology in their platforms. Other companies, such as Parallel Domain and Voxel51, are using the technology to improve simulation pipelines and AI training workflows.

Research institutions are also participating. Mcity, an autonomous vehicle research center operated by the University of Michigan, is using the technology to build a digital twin of its test track. The virtual environment enables researchers and industry partners to simulate autonomous-driving scenarios based on real-world infrastructure and traffic patterns.

Nvidia says the broader goal of these technologies is to provide a full-stack development ecosystem for autonomous driving, spanning hardware, AI models, safety frameworks, and simulation tools. By offering a standardized architecture for these systems, the company aims to simplify development for automakers and mobility providers building autonomous vehicle fleets.

While the company describes the technology as a foundation for scalable Level 4 autonomy, many of the systems and features announced remain in development. Nvidia notes that some technologies will be made available only when they reach production readiness, and that deployment timelines may change.

Autonomous vehicle technology has been under development for more than a decade, with automakers, technology companies, and startups investing heavily in software, sensors, and computing platforms designed to automate driving tasks. Nvidia’s Drive platform represents one of several computing ecosystems competing to provide the underlying hardware and software infrastructure for future autonomous vehicles.

With partnerships now spanning global automakers, ride-hailing companies, and software developers, Nvidia is positioning the Drive Hyperion platform as a shared foundation for those efforts. Whether the technology ultimately reaches large-scale commercial deployment will depend on continued advances in safety validation, regulatory approval, and real-world operational performance.