[SAN FRANCISCO] Tesla has disbanded the Dojo SuperComputer team, whose leaders are overturning the efforts of automakers to develop in-house chips for driverless technology, according to sources familiar with the issue.
Peter Bannon, who led Dojo, left, and CEO Elon Musk ordered efforts to close, according to sources who asked not to identify any discussion of internal issues. The team recently lost around 20 workers to its newly formed density, with the remaining dojo workers being reassigned to other data centers to calculate projects within Tesla, sources said.
Tesla plans to increase its reliance on external technology partners, including NVIDIA and advanced microdevice for computing, as well as Samsung Electronics for chip manufacturing, sources say.
Tesla, Musk and Bannon did not immediately respond to requests for comment.
The decision has shown significant changes in the creation of the program, with Dojo once positioned at the heart of Tesla's multi-billion dollar plan to acquire computing muscles in the Artificial Intelligence (AI) race.
The Dojo system is a Tesla-designed supercomputer used to train machine learning models behind electric vehicle (EV) manufacturers' autopilots and fully automated driving programs. The computer takes in data captured by the vehicle and processes it quickly to improve the company's algorithms. Analysts say Dojo could be a significant competitive advantage, with Morgan Stanley estimated that in 2023 it could add USD 500 billion to Tesla's value.
BT in your inbox
Deliver the latest news stories and analysis directly to your inbox, starting and ending daily.
Possibly poised to get out of stealth quickly, Dencessai is working on chips, hardware and software that powers AI data centers used in robotics, AI agents, and automotive applications. The company was founded by former Dojo head Ganesh Venkataramanan and former Tesla employees Bill Chang and Ben Floering.
Tesla's stock extended the postmarket decline after Bloomberg reported, falling less than 1% in New York at 5:29pm.
Talent drainage
Tesla has faced the departure of major talent this year as it has tackled increasing competition, lower sales and consumer backlash against mask political activity. Optimus's head of engineering Milan Kovac and software engineering vice president David Lau set out earlier this year, but Bloomberg reported in June that longtime Musk's best friend, Omead Afshar, had suddenly left.
The EV maker signed a US$16.5 billion deal with Samsung last month, protecting AI semiconductors until 2033. Future Texas factories will produce Tesla's next-generation AI6 chips, diversifying the procurement of Tesla from chipmaker Tesla, significantly exceeding Shipma Tewan semiconductor manufacturing.
Musk hints at a strategic pivot during Tesla's finest quarterly revenue call, suggesting that future iterations of the company's internal technology could converge with future iterations of partners. “When you think about Dojo 3 and AI6 inference chips, intuitively, they try to find convergence, which is basically the same chip,” Musk said in a call on July 23, 2023.
Tesla's CEO confirmed last year that the company has not pursued Dojo permanently and may instead be leaning more towards external partners.
“We are pursuing a dual path between Nvidia and Dojo,” Musk said in January 2024. “But I think Dojo is a long shot. Bloomberg
