According to the report, Tesla is splitting up the team behind the Dojo supercomputer, putting a stop to the automaker’s efforts to create in-house processors for autonomous technology.
Dojo’s leader, Peter Bannon, is departing the firm, and the remaining team members will be relocated to other Tesla data center and computing initiatives. The termination of Tesla’s Dojo initiatives comes after about 20 employees departed the manufacturer to launch DensityAI, an AI startup. According to reports, the new business will soon emerge from stealth and is developing chips, hardware, and software that will power AI data centers used in robotics, by AI agents, and in automotive applications. Former Dojo chief Ganesh Venkataramanan and former Tesla workers Bill Chang and Ben Floering formed DensityAI.
Additionally, it occurs at a crucial moment for Tesla.
Even after a limited robotaxi launch in Austin this past June that included Model Y vehicles with a human in the front passenger seat and led to several documented instances of the vehicles displaying problematic driving behavior, CEO Elon Musk has pushed for shareholders to see Tesla as an AI and robotics company.
The closure of Dojo, which Musk has been discussing since 2019, represents a significant change in Tesla’s approach. Due to its capacity to “process truly vast amounts of video data,” Dojo would be the cornerstone of Tesla’s AI goals and its aim of becoming fully self-driving, according to CEO Elon Musk. He only mentioned Dojo in passing during the company’s second-quarter earnings call.
Morgan Stanley estimated in 2023 that Dojo could increase the company’s market value by $500 billion by opening up new income streams including software services and robotaxis. Musk stated just a year ago that in the run-up to the October robotaxi unveiling, Tesla’s AI team will “double down” on Dojo.
But in August 2024, Musk started promoting Cortex, Tesla’s “giant new AI training supercluster being built at Tesla HQ in Austin to solve real-world AI,” and the discussion about Dojo stopped.
The Dojo project combined in-house chip manufacturing with a supercomputer. At its inaugural AI Day in 2021, Tesla publicly launched Dojo and showcased its D1 chip. Venkataramanan demonstrated the device, which Tesla said will power the Dojo supercomputer in conjunction with Nvidia’s GPU. Furthermore, the automaker said that it was developing a next-generation D2 microprocessor that would address any information flow issues with its predecessor.
According to sources, Tesla now intends to rely more on Nvidia and other outside tech partners, including as AMD for computing and Samsung for chip production. In order to produce its AI6 inference chips, which are expected to expand from powering FSD and Tesla’s Optimus humanoid robots to high-performance AI training in data centers, the company last month inked a $16.5 billion contract with Samsung.
Musk alluded to possible layoffs during Tesla’s second-quarter earnings conference.
“Inferentially, we want to look for convergence between Dojo 3 and the AI6 inference processor, which is essentially the same chip, Musk stated.
The announcement comes as Tesla’s board offers Musk a $29 billion compensation package to keep him at Tesla and assist push the company’s AI initiatives ahead, rather than becoming diverted by his other firms, notably the more pure-play AI startup xAI.
