Shanghai (Gasgoo)- On July 21, Chinese autonomous driving technology company WeRide announced the official launch of its HPC 3.0 high-performance computing platform, developed in collaboration with Lenovo Vehicle Computing.
The platform, first deployed in WeRide's new-generation Robotaxi GXR, is powered by NVIDIA's latest DRIVE Thor X chip and aims to offer a cost-effective, scalable solution for commercializing autonomous driving technology.
Built on Lenovo Vehicle Computing's AD1 domain controller for Level 4 driving automation, the HPC 3.0 platform integrates dual NVIDIA DRIVE AGX Thor chips, delivering up to 2,000 TOPS of AI computing power. It supports demanding workloads such as generative AI and vision-language models, and runs on the certified DriveOS system. The training process leverages the NVIDIA DGX accelerated computing platform to ensure optimal system performance.
Designed with high integration in mind, the HPC 3.0 incorporates key functions such as Ethernet and CAN gateways, inertial navigation, and collision detection, significantly reducing the need for additional hardware. This streamlined architecture lowers manufacturing costs and simplifies after-sales maintenance. According to WeRide, the platform cuts autonomous driving suite costs by 50%, reduces total cost of ownership (TCO) by 84%, and slashes mass production pricing to one-quarter of the previous generation.
As a vehicle-grade solution, the HPC 3.0 meets certifications including AEC-Q100, ISO 26262, and IATF 16949. It features a multi-redundant design to satisfy ASIL-D safety requirements, with a failure rate of less than 50 FIT and a mean time between failures (MTBF) of 120,000 to 180,000 hours—covering a design lifespan of 10 years or 300,000 kilometers.
The platform also demonstrates strong environmental resilience, operating in temperatures from -40°C to 85°C and passing comprehensive testing for heat aging, mechanical shock, salt spray corrosion, and VOCs (Volatile Organic Compounds) compliance, making it well-suited for global deployment in regions with harsh climates, such as the Middle East and Southeast Asia.
WeRide founder and CEO Dr. Tony Xu Han noted that the HPC 3.0 significantly enhances the reliability and responsiveness of the Robotaxi GXR while lowering the barriers to commercial rollout. The GXR has already operated safely on public roads for over 2,000 days, and the new computing platform is expected to further drive its scalability.
Mr. Tang Xinyue, head of Lenovo Vehicle Computing, emphasized that the deployment of the AD1 domain controller marks a step toward standardization and mass adoption of Level 4 autonomous driving. The two companies plan to expand their partnership to include autonomous minibuses, sanitation vehicles, and other Level 4 platforms, targeting full coverage across urban mobility, logistics, and municipal services.
Ali Kani, the VP and general manager of the automotive team at NVIDIA, highlighted that the Level 4 systems built on NVIDIA's accelerated computing and DriveOS infrastructure are laying the foundation for safe and scalable autonomous mobility. He described the WeRide–Lenovo collaboration as a strong example of advanced technology transitioning into real-world applications.
The launch of the HPC 3.0 underscores the ongoing industry pursuit of balancing computational performance with cost efficiency. By optimizing chip selection, system integration, and production processes, WeRide said it aims to strike that balance. However, whether these cost savings can be sustained at scale—and how well the system adapts to global markets—will ultimately depend on real-world operational outcomes.
