FILE PHOTO: A view shows the General Motors plant, which exports vehicles to Canada and the U.S., in Ramos Arizpe, Mexico January 29, 2025. REUTERS/Daniel Becerril/File Photo
(Reuters) - General Motors will use artificial intelligence chips and software from Nvidia to develop autonomous vehicle technology for its vehicles and improve workflow at its factories, the companies said on Tuesday.
Traditional automakers have struggled to commercialize autonomous technology that has been more challenging and expensive than expected but has emerged as a way to boost sales and rake in subscription revenue from motorists.
The companies plan to work together to build AI systems using Nvidia's platforms to train AI manufacturing models for factory planning. GM also plans to use Nvidia's autonomous tech for future advanced driver-assistance systems.
Asked about financial terms, a GM spokesperson said the companies have a strategic collaboration on using AI in manufacturing and GM will buy chips from Nvidia for driver-assistance technology.
A slew of automakers and suppliers, including Toyota and Hyundai, have partnered this year with Nvidia to develop their autonomous driving capabilities in the face of competition from Tesla, which uses proprietary technology to run its Full Self-Driving system.
GM used Nvidia's chips to power its self-driving Cruise robotaxis. But it started developing custom chips in-house to reduce cost and dependency before it shut down the robotaxi business last year to focus instead on AV technology for personal vehicles.
GM has forecast that its Super Cruise driver-assistance technology would earn about $2 billion in total annual revenue within five years. Super Cruise is free for three years, after which customers are offered subscriptions for $25 a month or $250 a year.
Other companies that have partnered with Nvidia for their advanced driver-assistance systems include Chinese EV maker BYD, German carmaker Mercedes-Benz, and U.S. EV startup Rivian.
GM had previously used Nvidia's platforms for training AI models, including for simulation and validation.
(Kalea Hall reporting in Detroit and Abhirup Roy reporting in San Francisco; Editing by Rod Nickel)
