NVIDIA details its three-computer robotaxi stack as partners target driverless fleets at scale
NVIDIA says every major commercial robotaxi program runs on its modular stack — a three-computer platform for training, simulation and in-vehicle computing — in a market projected to reach $400 billion by 2035.
- $400 billion
- over 6 million
- 28 cities by 2028
- 14 HD cameras, 9 radars, 3 lidars, 12 ultrasonics
What Happened
The global robotaxi market is projected to reach $400 billion by 2035, with more than 6 million commercial vehicles in operation, and driverless fleets are already moving people through some of the world's busiest streets. Deploying a driverless vehicle is one challenge; scaling a fleet means delivering the same safe, reliable performance across thousands of vehicles. Meeting that demand requires enormous compute across the development lifecycle, from preparing and training AI models to simulating and validating driving behavior, plus real-time processing in the vehicle. NVIDIA provides an open platform for AI training, simulation and safety validation, with libraries, software development kits, workflows and models developers can use alongside their own technology stacks — and it says every major robotaxi program operating at commercial scale today runs on its modular stack.
$400 billion
Physical AI's first commercial breakthrough, with over 6 million commercial vehicles expected in operation.
- Training computer: driving models can be trained on NVIDIA DGX systems, with the NVIDIA Alpamayo portfolio of open reasoning vision language action (VLA) models, simulation frameworks and physical AI datasets, plus reinforcement learning blueprints and recipes for post-training and distillation.
- Simulation and validation computer: NVIDIA Omniverse NuRec reconstructs real-world driving scenarios from sensor data and NVIDIA Cosmos world foundation models generate variations, turning thousands of corner cases into millions of combinations of behavior, traffic, weather, lighting and sensor conditions on NVIDIA RTX PRO Servers; the AlpaSim framework supports closed-loop simulation of reasoning-based models.
- In-vehicle computer and sensor architecture: NVIDIA DRIVE Hyperion is the modular, level-4-ready reference architecture, and NVIDIA Halos provides a production-ready safety foundation through Halos OS plus a validation and certification framework from cloud to car.
- Dual NVIDIA DRIVE AGX Thor systems-on-a-chip built on the NVIDIA Blackwell platform
- 14 high-definition cameras
- Nine radars, three lidars and 12 ultrasonics for 360-degree sensor fusion
- Uber is scaling its fleet of NVIDIA DRIVE Hyperion, with plans to reach 28 cities by 2028, and is building a robotaxi AI data factory on NVIDIA Cosmos to curate fleet driving data for rare scenarios; it is collaborating with Autobrains, Avride, Lucid, May Mobility, Mercedes-Benz, Momenta, Nissan, Nuro, Pony.ai, Stellantis, Waabi, Wayve, WeRide and Zoox to bring NVIDIA-powered robotaxi services to its platform.
- May Mobility plans to operate autonomous ride-hailing services through Uber's network while developing its software stack on the NVIDIA DRIVE platform.
- Bolt uses NVIDIA technologies to develop and scale AVs across Europe.
- Lyft plans to use NVIDIA DRIVE Hyperion as a reference architecture for future autonomous fleets; May Mobility vehicles are currently operating on Lyft's network in Atlanta powered by NVIDIA DRIVE.
- Through its partnership with Grab, WeRide plans to bring its DRIVE Hyperion- and DRIVE AGX Thor-based GXR to key markets across Southeast Asia.
- Wayve, Nissan and Uber are developing a global robotaxi program using a prototype vehicle that combines Nissan's vehicle engineering, Wayve's embodied AI and the NVIDIA DRIVE Hyperion platform.
- Autobrains is developing robotaxi programs with Uber in Munich and VinFast in Southeast Asia, built on NVIDIA DRIVE Hyperion and enabled by its Agentic AI technology.
- Zoox uses NVIDIA DRIVE for in-vehicle computing and cloud-based training and simulation.
- Momenta is developing its software stack based on NVIDIA DRIVE AGX running on DriveOS, and Pony.ai developed its new-generation autonomous-driving domain controller with DRIVE Hyperion and DRIVE AGX Thor.
- Tensor is developing its level 4 Robocar with eight NVIDIA DRIVE AGX Thor systems-on-a-chip in its in-vehicle supercomputer.
Other developers include Waabi, which is expanding into the robotaxi market through a deployment collaboration with Uber and builds its Waabi Driver platform on NVIDIA DRIVE AGX Thor, and TIER IV and Isuzu, which are deploying level 4 autonomous buses on DRIVE Hyperion and DRIVE AGX Thor. Lenovo is supplying its NVIDIA DRIVE AGX Thor-based AD1 level 4 domain controller for a next-generation robotaxi program with SWM, while DeepRoute.ai is developing a new generation of robotaxis on the DRIVE Hyperion platform with DRIVE AGX Thor. Among automakers, Tesla trains its autonomous-driving neural networks on NVIDIA supercomputers, and Mercedes-Benz and NVIDIA are collaborating with Uber on a robotaxi ecosystem based on the new S-Class, built on DRIVE Hyperion, full-stack NVIDIA DRIVE AV L4 software and Alpamayo open AI models. Stellantis, Wayve and Uber are collaborating on level 4 driverless mobility services using DRIVE Hyperion, and Lucid, Nuro and Uber are developing a global robotaxi service on DRIVE AGX Thor. Hyundai Motor and Kia are expanding their NVIDIA collaboration on data-driven autonomous-driving systems built on DRIVE Hyperion, with NVIDIA also exploring expanded collaboration with Hyundai Motor Group's joint venture Motional for level 4 robotaxis; Geely plans to develop and commercialize robotaxis using DRIVE Hyperion, and its brand Zeekr has adopted DRIVE AGX Thor for a centralized domain controller.
Previously from NVIDIA Corporation
In August 2026, NVIDIA released Alpamayo 2 Super, described as the top-ranked open reasoning model for autonomous driving, made commercially available under a permissive license. Earlier, in June 2026, the company introduced Halos OS, a certified safety foundation for robotaxis that addresses regulatory requirements for reliable autonomous vehicle operation. Both layers reappear in today's story: the Alpamayo open models and datasets sit in the training and simulation stack, and Halos provides the safety foundation for the in-vehicle platform.
- NVIDIA Alpamayo 2 Super Open Model for Robotaxis and AVs Now Commercially Available
- NVIDIA Halos OS Provides Safety Foundation for Robotaxi Deployments
Background drawn from MotorClaw's earlier coverage of NVIDIA Corporation's official releases.
Why this matters
Robotaxi developers, mobility platforms and automakers are leaning on shared computing platforms to train models, simulate rare scenarios and run in-vehicle AI instead of building every layer themselves. The partners named span North America, Europe, Asia and the Middle East, so the architecture chosen now could shape which driverless services reach riders first.
Terms in This Story
- robotaxi
- A driverless vehicle that carries paying passengers without a human driver.
- VLA (vision language action)
- A type of AI model that combines visual perception, language-style reasoning and the selection of driving actions.
- level 4
- A level of driving automation at which a vehicle can perform all driving tasks within a defined operating area without human intervention.
- long-tail
- Rare or unusual driving situations that appear infrequently in real-world data but matter greatly for safety.
Summarised from the linked release; details can be imperfect — always verify against the original source.