Nvidia is partnering with Uber to support up to 100,000 robotaxis starting in 2027 with its Nvidia Drive AV software platform and Drive AGX Hyperion 10 autonomous vehicle development platform, according to one in a series of announcements directly affecting the sensors community made Tuesday by CEO Jensen Huang at GTC Washington DC. Stellantis, Lucid and Mercedes-Benz are also collaborating on Hyperion 10 for passenger mobility.
Hyperion 10 is both a reference production computer and sensor set architecture to make any vehicle level 4 autonomous driving ready, Nvidia said in a pre-brief with reporters. It will work with cars, trucks and vans equipped with validated hardware and sensors that host compatible autonomous-driving software. Uber’s plan is to provide a ride-hailing service for both human and robot drivers powered by Hyperion 10 with Uber supplying the handheld ride-hailing software and network.
Sensors partners working with the Nvidia automotive ecosystem include 13 companies so far: Aeva, Arbe, Aumovio, Bosch, Cepton, Forvia, Hesai, Innoviz, Luminar, onsemi, Robosense, Sony, and Valeo. Those sensors companies are listed alongside 26 OEMs and Neighborhood Electric Vehicle (NEVs) makers that include smaller and slower-moving vehicles. The biggest names in the vehicle industry are included, such as Ford and GM, but also Hyundai, Toyota, Volvo, BMW and Mercedes Benz. Aurora and Iveco are in a group of 10 truck makers while Zoox and Uber and Nuro are listed in a group of 10 robotaxi makers.
Nvidia also lists other companies offering essential software and hardware. They include 11 AV software makers including Nuro and Wayve, and 11 tier 1 suppliers like Bosch and Foxconn and Lenovo, also 10 separate simulation software makers as well as six base software makers like Android and QNX and Red Hat. Smart factory partners include BMW, GM, Hyundai JLR, Lucid, Mercedes Benz, Schaeffler, Toyota and Volvo. In-cabin software makers in the Nvidia ecosystem include Cerence among four others. Cloud compute is from among eight providers including Alibaba Cloud, AWS, ByteDance, Baidu AI Cloud, Google Cloud, Micrsoft Azure, Oracle Cloud and Tencent. Siemens and Autodesk are part of a group of six engineering and design suppliers to Nvidia.
Sensors for robotaxis and trucks can include optical, radar and infrared sensors, depending on the OEM involved. Vehicles also have traditionally included accelerometers, location and vibration electronics. Nvidia said it has engineered Hyperion 10 to make any vehicle level 4 AV ready, which includes redundancy. In other words, if any sensor fails in guiding a robotaxi, a redundant system will step in so that any moving vehicle can get to a safe stop, said Ali Kani, vice president and general manager of the Nvidia automotive platform.
Hyperion 10 features the Drive AGX Thor System-on-Chip and the Drive OS operating system as well as a multimodal sensor suite with 14 high-def cameras, nine radars, one lidar and 12 ultrasonics, as well as a qualified board design. The prequalified sensor suite architecture helps accelerate development, lower costs and gives automakers a “running start” with access to Nvidia’s development expertise, Nvidia said in a press release. There are actually two Drive AGX Thor in-vehicle platforms based on Nvidia Blackwell, each GPU delivering more than 2000 FP 4 teraflops of real-time compute . Drive AGX Thor fuses diverse 360-degree sensor inputs and is optimized for transformer, vision language action models and generative AI workloads.
Also, an Nvidia Halos system delivers safety guardrails from cloud to car. Bosch, Nuro, Wayve and Aumovio are inaugural members of the Nvidia Halos AI System inspection lab which aims to accelerate level 4 automated driving. L4 is often called “high automation” to allow a vehicle to perform all driving tasks and also monitor the driving environment without human intervention in some limited conditions. If the L4 system encounters a situation it cannot handle or the driver does not respond to a warning, it will safely pull over.