Uber will use Cosmos, Nvidia's new generative world model simulation tool, and DGX Cloud, a cloud-based AI supercomputing platform, to support the development of self-driving car technology, the companies announced at CES 2025.
Cosmos is sold to robotics and self-driving companies as a tool to generate physically-based videos from a variety of inputs. It contains 9,000 trillion tokens generated from 20 million hours of (presumably copyrighted) video and can be used for realistic industrial and driving simulations. environment. Nvidia DGX Cloud provides Uber and other companies with access to high-performance AI infrastructure to train, fine-tune, and deploy AI self-driving models.
Uber provided few details about how it plans to use these Nvidia tools. Last year, the ride-hailing and delivery giant forged 14 partnerships with AV companies in a variety of sectors, from robotaxis company Waymo to trucking company Aurora Innovation to sidewalk robot delivery company Serve Robotics.
Uber will likely work with these partners, handing over vast amounts of data about where and how people use rides and deliveries. The company does not develop its own AV technology, preferring instead an asset-light approach of partnering with other companies. It's not for lack of effort. Uber's history of AV development is problematic.
Uber launched its self-driving division, Uber ATG, in 2015 through a partnership with Carnegie Mellon University's National Robotics Center. A year later, the company acquired Otto, a self-driving truck startup founded by Anthony Levandowski, one of Google's star engineers. Uber has since shut down its trucking technology division to focus on self-driving cars. However, introducing Lewandowski did some damage. Waymo accused him of stealing trade secrets, which Uber then used. The case was settled in 2018, and Levandowski was later sentenced to 18 months in prison before being pardoned by then-President Donald Trump.
In 2018, an Uber self-driving car testing self-driving mode struck a pedestrian in Arizona and died from injuries. The following year, Uber spun out Uber ATG after completing a $1 billion financing from Toyota, but sold it to self-driving car startup Aurora Innovation in 2020.
Now, Uber seems committed to building a bridge between passengers and drivers, both human and robot drivers. But that doesn't mean Uber doesn't mind moving slowly.
During the company's third-quarter earnings call, Uber CEO Dara Khosrowshahi said that due to startup investments, the company will only partner with Waymo in two cities this year: Austin and Atlanta. He explained that he would start.
“What we always want to have is to provide adequate liquidity to cities,” Khosrowshahi said in response to an analyst's question. “We want to go to cities with the right investments in things like warehousing, infrastructure and mapping to get a return on capital.”
Laying the groundwork for AV deployment in new cities will take time, but Uber seems to think it can scale more quickly using Nvidia's world model and cloud platform.
“We believe that by working with Nvidia, we can significantly accelerate the industry's timeline for safe, scalable self-driving solutions,” Khosrowshahi said in a statement.
Uber does not share further details with TechCrunch.