Nvidia is building an automotive technology business. A self-driving test car is pictured at the company’s automotive garage on June 5, 2023 in Santa Clara, California.
Bloomberg | Bloomberg | Getty Images
Nvidia The company on Monday announced plans to work with partners to test a robotaxi service as early as 2027, underscoring the chipmaker’s ambitions to become a major player in the world of self-driving cars.
Nvidia officials said at a self-driving demonstration in San Francisco last month that the service will be offered in collaboration with partners and will employ vehicles with “Level 4” driving. This means it can be driven in predefined areas without human intervention. The company did not say where it would operate or who its partners would be.
“We’ll probably start with limited supply, but we’ll work with our partners to gain a foothold,” Xinzhou Wu, Nvidia’s vice president of automotive, said at the event.
Nvidia has been offering automotive chips and other technology under the Drive brand since 2015, but that remains a small part of the company’s business. Automotive and robotic chip sales for the quarter ended October were just $592 million, about 1% of Nvidia’s total sales. Nvidia announced a robotaxi partnership with Uber in October.
The chipmaker announced in December that it had developed software that could power self-driving cars, and said Mercedes-Benz models launching in late 2026 would be able to use Nvidia’s technology to navigate cities like San Francisco.
Self-driving cars remain one of the key areas where Nvidia can show growth outside of AI infrastructure. CEO Jensen Huang said robotics, including self-driving cars, is the company’s second most important growth area after artificial intelligence.
“I imagine that someday all of the 1 billion cars on the road will be self-driving,” Huang said Monday at a launch event at the CES conference in Las Vegas. “You can tune it up and rent it from someone else, or you can own it yourself.”
In addition to the chips that go into self-driving cars, Nvidia sells access to its famous AI chips and simulation software to car companies to help them train and develop their self-driving models.
Nvidia says the Drive AGX Thor in-vehicle computer, which costs about $3,500 per chip, will help automakers save on research and development costs and bring self-driving features to market faster. Nvidia said it is working with automakers to fine-tune its technology, including determining how fast to accelerate for specific vehicles.
“Some people say, ‘I need your help training and optimizing the software on the chip, but I’ll do the simulation myself,'” said Ali Kani, Nvidia’s general manager of automotive platforms.
Car companies such as Mercedes-Benz want to adapt Nvidia’s technology and market it as part of the in-car experience, selling it as part of or with new cars.
Robotaxis have gained popularity over the past year. of the alphabet Waymo operates driverless commercial taxi services in five U.S. markets, including San Francisco.
Nvidia’s robotaxi announcement suggests that it is targeting self-driving cars in addition to private cars that consumers can buy.
Ride an Nvidia robotaxi
In December, NVIDIA offered reporters and analysts an hour-long drive around San Francisco in a 2026 Mercedes-Benz CLA sedan.
My car was driven by a safety driver hired by Mercedes-Benz, who said the car was driving itself for 90% of the trip.
The drive was uneventful. San Francisco is a difficult city to drive in, with large hills, frequent red lights, and trucks unloading in the middle of the road, but I didn’t feel stressed and was able to concentrate on the conversation.
However, there was one big problem. The driver took the wheel in an awkward situation where two buses and a self-driving Waymo were trying to pass on a four-lane road with on-street parking and unloading trucks on both sides. Drivers had to back up their cars and wait until the traffic cleared.
Nvidia said my drive was “Level 2 Plus Plus.” This means that NVIDIA technology can: Tesla’s Fully automatic driving mode. While Nvidia-powered cars like Mercedes-Benz have enhanced self-driving capabilities, the responsibility for keeping everyone safe still lies with the driver, who must remain alert at all times.
The chipmaker said the system would eventually be able to drive “park-to-park,” meaning driving from one parking space to another, but Mercedes-Benz CLA cars will not have that feature.
“If you’re scared of a parking situation, the car will solve it for you,” Mercedes-Benz Group CEO Ola Källenius said Monday at an NVIDIA event.
The Mercedes-Benz model that Nvidia demoed was launched in Europe last year and will be available in the U.S. this year, Kani said.
Kani said Mercedes-Benz vehicles are equipped with lane-keeping and driver-assistance features that help drivers stay in their lanes. Lane-switching functionality will be added to these vehicles through software updates, and this year will include hands-free highway driving, city driving, and parking lot-to-park functionality.
Nvidia said it uses two AI systems in Drive-equipped vehicles to ensure safety. Cars primarily run on an “end-to-end” system called a vision language model that uses AI to decipher visual sensors and plot a route.
The company said it has also built a second, safety-oriented “stack” that uses strict rules, such as stopping at stop signs, to take over when the AI doesn’t know what to do.
Still, Nvidia hopes recent advances in generative AI, powered by the company’s graphics processing units, will further improve the capabilities of its self-driving algorithms. Nvidia is targeting 2028 for point-to-point self-driving capabilities in consumer vehicles. Ultimately, NVIDIA said it wants the car itself to feel like a real driver, one that users can simply talk to.
“With transformers and generative AI, we can do even more,” Wu said.
