Uber and NVIDIA Partner to Accelerate Autonomous Driving at CES 2025

Uber and Nvidia Partner to Accelerate Autonomous Vehicle Development

Uber is set to integrate Nvidia’s newly released generative world model simulation tool, Cosmos, alongside the cloud-based AI supercomputing platform, DGX Cloud. This collaboration aims to bolster the advancement of autonomous vehicle technology, as revealed at CES 2025.

Nvidia's Cosmos and DGX Cloud Capabilities

Cosmos is positioned as a valuable asset for companies operating in the robotics and autonomy sectors. It functions by generating physics-based videos from diverse inputs.

These inputs include a substantial dataset of 9,000 trillion tokens derived from 20 million hours of video content – potentially subject to copyright – which facilitates the simulation of realistic industrial and driving scenarios.

Nvidia DGX Cloud will provide Uber, and other organizations, with access to robust AI infrastructure. This infrastructure is designed for the training, refinement, and deployment of AI-powered self-driving models.

Uber's Strategy and Partnerships

Uber has disclosed limited specifics regarding its intended utilization of these Nvidia tools. Over the past year, the ride-hailing and delivery company has established 14 partnerships with autonomous vehicle companies spanning various areas.

• Robotaxis with Waymo
• Trucking with Aurora Innovation
• Sidewalk robotic delivery with Serve Robotics

It is plausible that Uber will leverage its extensive data concerning ride-hail and delivery usage patterns. This data could be used in conjunction with its partners to enhance their respective technologies.

The company currently favors an asset-light strategy, prioritizing partnerships over in-house AV development. This approach stems from a history of challenges with its own autonomous vehicle initiatives.

A Troubled Past with Autonomous Vehicle Development

Uber initiated its self-driving unit, Uber ATG, in 2015 through a collaboration with Carnegie Mellon University’s National Robotics Center.

In 2016, the company acquired Otto, a self-driving truck startup founded by Anthony Levandowski, a former Google engineer.

Subsequently, Uber discontinued its trucking technology division to concentrate on self-driving cars. However, the acquisition of Levandowski proved contentious.

Waymo accused Levandowski of misappropriating trade secrets, alleging their use by Uber. The legal dispute concluded with a settlement in 2018, and Levandowski received an 18-month prison sentence, later commuted by President Donald Trump.

Incidents and Subsequent Restructuring

In 2018, an Uber self-driving vehicle, operating in autonomous mode in Arizona, struck and fatally injured a pedestrian.

The following year, Uber ATG was spun out after securing $1 billion in funding from Toyota, only to be sold to autonomous vehicle startup Aurora Innovation in 2020.

Focus on Integration and Scalability

Currently, Uber appears dedicated to serving as the intermediary between riders and drivers, regardless of whether those drivers are human or robotic.

However, the company remains committed to rapid progress. During its third-quarter earnings report, Uber CEO Dara Khosrowshahi explained the company’s deliberate rollout of Waymo services in only two cities – Austin and Atlanta – this year.

“The goal is to achieve sufficient liquidity in a city upon launch,” Khosrowshahi stated. “This requires substantial investment in depots, infrastructure, mapping, and other essential elements to ensure a return on capital.”

Establishing the necessary infrastructure for AV deployment in a new city requires time. Nevertheless, Uber believes that Nvidia’s world models and cloud platform can expedite the process of achieving scalability.

“We are confident that our collaboration with Nvidia will accelerate the development of safe and scalable autonomous driving solutions for the industry,” Khosrowshahi affirmed.

Uber declined to provide further details to TechCrunch.