SHANGHAI/SINGAPORE (Reuters) – Tesla plans to use data from China that can be processed domestically to drive global development of self-driving systems as part of Elon Musk’s strategic shift, people familiar with the matter said. It is said that progress is underway. of the work.
As part of that effort, Tesla plans to build a data center in China to train algorithms needed for more fully self-driving cars, said two people, who requested anonymity because the work is private. It is said that they are formulating the
Until recently, Tesla had not received approval from Chinese regulators to transfer data generated by EVs in China overseas for its “Full Self-Driving” (FSD) system, the two people and additional people said. The company has been focusing on efforts to obtain the following. of the problem.
It’s unclear whether Tesla will pursue both data transfer and local data center options for processing self-driving data from China, or whether it’s developing parallel plans as a hedge.
Tesla’s efforts highlight the speed with which electric car makers are betting on advances in AI as demand for EVs slows and competition intensifies.
The company’s efforts to make the most of Chinese vehicle data to develop driving artificial intelligence for automakers comes as the U.S. government attempts to crack down on AI technology transfers from U.S. companies to China. .
Tesla has not been able to offer the full version of the FSD, which costs about $9,000 worth, in China.
A growing market for FSDs in China would boost the company’s revenue and profits at a time when pressure from Chinese rivals such as BYD has squeezed automakers’ revenues and margins.
Tesla did not respond to a request for comment.
Tesla will need to work with a Chinese partner to set up a data center in China for FSD development, two of the people said. Hardware procurement also presents potential challenges.
The company is in talks with Nvidia about acquiring graphics processing equipment for data centers in China, according to one of the people briefed on the talks. U.S. sanctions prevent NVIDIA and its partners from selling cutting-edge chips in China.
Nvidia declined to comment on whether it had held discussions with Tesla.
Tesla’s campaign to leverage more data from China began in earnest when Mr. Musk made a whirlwind visit to Beijing last month, meeting with officials including Premier Li Qiang.
According to two sources familiar with the matter, Mr. Musk tried to smoothly approve Tesla’s data transfer in his talks with Mr. Lee. The possibility of Tesla investing in data centers in China has also been raised.
Musk also discussed the possibility of Tesla licensing its FSD system to Chinese EV makers, one of the people said. Musk said in April that Tesla was in talks with another “major” automaker to license FSD, without naming the company.
China’s State Council Newspaper Office did not respond to a request for comment.
china effect
China, the world’s largest car market, has the largest fleet of cars equipped with sensors that can collect data from crowded cities with complex traffic patterns, and the data it generates is valuable to automakers and AI providers. It has become valuable.
Mr. Musk has previously expressed opposition to China-based data centers, arguing that transferring data to the United States is the most efficient option, two people familiar with the matter said.
Tesla has been storing data collected by its Chinese EVs in Shanghai since 2021. Tesla’s China team has meanwhile been seeking approval from Chinese regulators to transfer data out of the country, two people familiar with the matter said.
Under a one-year pilot project, companies in Shanghai’s Lingang district, where Tesla’s factory is located, will be allowed to transfer certain data without requiring further security assessments, Reuters reported on Friday. That’s what it means.
Some analysts believe Mr. Musk is trying to make China a launching pad for self-driving, similar to how Tesla bet on the Shanghai Gigafactory in 2019 to enable its rise as a mass-market electric car maker. I suspect that this is the case.
“If Tesla can deploy FSD in China and use Chinese data for algorithm training, it will definitely be a milestone for Tesla,” said Yale, managing director of Automotive Foresight, a Shanghai-based consultancy. Mr. Chan said.
“China has played a key role in expanding Tesla’s EV production at its Shanghai factory, and will again play a key role in expanding the mass adoption of self-driving technology,” he said.
Many industry experts predict that it will be years before fully self-driving cars become widespread, but predictions vary widely.
The driver assistance features currently available in China are “Level 2” systems, meaning they require a driver who is ready to take over the wheel. Tesla’s less advanced options of his FSD and Autopilot are also Level 2 systems that require an attentive driver.
A fleet of more fully automated vehicles operated by Baidu, China’s largest search engine operator, and self-driving startup Pony.ai is driving in a limited test zone. .
But Chinese EV makers, including BYD, have made autonomous driving and advanced driver assistance systems a priority. Mercedes and BMW have received a license to test a Level 3 system that allows drivers to take their hands off the wheel and look away on a wide range of roads in China.
At least five companies – Hyundai, Mazda, Toyota, Volkswagen and Nissan – have received approval to transfer some of their data outside of China, but none of those approvals involve data used to train AI systems. has not been approved, the lawyer said in a state media report. And the analysts.
An organization representing foreign companies in China says penalties for violating China’s data privacy law, which took effect in 2021, are a major risk factor for data operations.
Asked about competition from Chinese EV makers during a conference call with investors after Tesla’s quarterly results last month, Musk said Tesla should be seen more as an AI company.
He said he believes Tesla’s FSD system “will work fairly well in almost all markets without modification.” Musk said “country-specific” training would be more effective.
(Reporting by Zhang Yan and Fanny Potkin; Editing by Sonali Paul)
