Nvidia wants to power robotaxi fleets with chips, software by 2027
Nvidia is building an automotive technology business. Seen here are autonomous vehicle test cars at the company’s auto garage in Santa Clara, California, on June 5, 2023.
Bloomberg | Bloomberg | Getty Images
Nvidia It said Monday it is working with robotaxi operators, hoping to use the company’s AI chips and Drive AV software stack to power fleets of autonomous vehicles as soon as 2027.
Xinzhou Wu, Nvidia’s vice president of automotive, said at an autonomous driving demonstration in San Francisco last month that the chipmaker hopes to support co-deployment of “Level 4” robotaxis during this period, meaning vehicles that can drive without human intervention in predefined zones.
“We will probably start with a limited availability but will work with our partner to get our place,” Wu said.
Nvidia has been offering chips and other technologies for cars under the Drive brand since 2015, but that remains a small part of the company’s business. Automotive and robotics chips generated just $592 million in sales in the quarter ending in October; This accounted for approximately 1% of Nvidia’s total revenue. Nvidia announces its partnership with robotaxi Uber in October.
The chipmaker said in December that it was developing software that could power a driverless car, and that Mercedes-Benz models due in late 2026 could use Nvidia’s technology to navigate cities like San Francisco.
Self-driving cars remain one of the main areas where Nvidia can grow outside of its AI infrastructure. CEO Jensen Huang said robotics, including driverless cars, is the company’s second most important growth category after artificial intelligence.
“We dream that one day all one billion cars on the road will be autonomous,” Huang said at a launch event at the CES conference in Las Vegas on Monday. “It could be a robotaxi that you orchestrate and rent from someone, or you could own it.”
In addition to the chips built into self-driving cars, Nvidia also sells simulation software to automotive companies along with its famous AI chips so they can train self-driving models and develop technology.
Nvidia says automakers can use its Drive AGX Thor automotive computer, which costs about $3,500 per chip, to save on research and development costs and bring self-driving features to market faster. Nvidia said it works with automakers to tune its technologies for certain vehicles, such as determining how much the vehicle should accelerate.
“Some people say, ‘Hey, I need your help training and optimizing my software on your chip, but I’ll handle the simulation myself,'” said Ali Kani, general manager of Nvidia’s automotive platform.
Auto companies like Mercedes-Benz want to tweak Nvidia’s technology, market it as part of the in-car experience, and sell it as part of or alongside a new car.
Robotaxies gained popularity last year, led by alphabet Waymo operates a driverless commercial taxi service in five U.S. markets, including San Francisco.
Nvidia’s robotaxi announcement signals that it is targeting driverless fleets as well as personal cars that consumers can buy.
Riding in an Nvidia driverless car
In December, Nvidia offered reporters and analysts a one-hour ride around San Francisco in its 2026 Mercedes-Benz CLA sedan.
While my car had a Mercedes-Benz-appointed safety driver behind the wheel, the driver said the car drove itself for 90% of the drive.
The ride was smooth. San Francisco is a tough city to drive in, with big hills, frequent red lights, and trucks unloading in the middle of the street, but I never felt stressed and was able to focus on conversation.
But there was one big snag: The driver took control in a tricky situation where two buses and the driverless Waymo were trying to cross a four-lane road with street parking on either side of trucks unloading goods. The driver had to back up the car and wait for the congestion to clear.
Nvidia says my driver is “Level 2 Plus Plus”, which means its technology has similar features Tesla’s Full Self-Driving mode. Nvidia-powered cars like Mercedes-Benz will increasingly have autonomous driving capabilities, but the responsibility to keep everyone safe still falls on the driver, who must always be careful.
The chipmaker said the system could eventually handle “park-to-park” operation, meaning going from one initial parking space to another, but Mercedes-Benz CLA cars won’t initially come with this feature.
“Any parking situation that you think is scary, this car will solve it for you,” Mercedes-Benz Group CEO Ola Källenius said at the Nvidia event on Monday.
Kani said that the Mercedes-Benz model shown by Nvidia was launched in Europe last year, but will also be launched in the United States this year.
Kani said Mercedes-Benz cars come with lane keeping and driver assist features to help drivers stay in their lane. Vehicles that gained the lane changing feature with a software update will also have hands-free highway driving, city driving and park-to-park features this year.
Nvidia said it uses two artificial intelligence systems in Drive-enabled vehicles to ensure security. The car mostly drives with an “end-to-end” system called a vision language model, which uses artificial intelligence to decode visual sensors and plot a path.
The chipmaker said it’s also building a second, security-focused “stack” that uses strict rules (like stopping at a stop sign) to take over when the AI isn’t sure what to do.
Still, Nvidia hopes that recent advances in generative AI driven by the company’s graphics processing units will allow self-driving algorithms to become more capable. Nvidia is targeting 2028 for point-to-point autonomous driving features in consumer vehicles. Ultimately, Nvidia said it wants to make the tool feel like a real driver that users can easily talk to.
“We can do so much more with transformers and generative AI,” Wu said.
Correction: An earlier version of this story incorrectly described Nvidia’s plans to work with robotaxi partners.
