Tesla Abandons Ambitious Dojo AI Supercomputer Project, Reassigns Team
Tesla is reportedly dismantling its dedicated Dojo supercomputer team, marking a significant pivot away from its ambitious in-house chip development strategy for autonomous driving technology. According to Bloomberg, which cited anonymous sources, the automaker is disbanding the team responsible for the AI training supercomputer, with its lead, Peter Bannon, departing the company. Remaining team members are expected to be reassigned to other data center and compute initiatives within Tesla.
This development follows the departure of approximately 20 employees from the Dojo project who have since founded a new AI company named DensityAI. This startup is poised to launch soon, focusing on building chips, hardware, and software for data centers supporting AI applications in robotics, AI agents, and automotive sectors. DensityAI was established by former Dojo head Ganesh Venkataramanan, alongside ex-Tesla employees Bill Chang and Ben Floering.
The decision comes at a critical juncture for Tesla, as CEO Elon Musk actively promotes the company as a leader in AI and robotics. Despite recent robotaxi launches that have encountered reported driving anomalies, Musk has consistently positioned Tesla’s AI capabilities, powered by projects like Dojo, as central to its future. Introduced in 2019, Dojo was envisioned by Musk as the cornerstone of Tesla’s AI strategy, particularly for processing vast amounts of video data crucial for achieving full self-driving capabilities. He referenced Dojo as recently as the company’s second-quarter earnings call.
The potential impact of Dojo was considerable; Morgan Stanley predicted in 2023 that it could add $500 billion to Tesla’s market value by enabling new revenue streams such as robotaxis and advanced software services. Last year, Musk emphasized a significant investment in Dojo leading up to Tesla’s robotaxi reveal in October.
However, the focus on Dojo began to shift around August 2024, when Musk started highlighting “Cortex,” a new, expansive AI training supercluster being constructed at Tesla’s Austin headquarters. This strategic shift appears to involve increased reliance on external partners. Sources indicate Tesla plans to bolster its collaborations with Nvidia, AMD for compute solutions, and Samsung for chip manufacturing. Notably, Tesla recently secured a $16.5 billion deal with Samsung for the production of its AI6 inference chips, designed for scalability from powering Full Self-Driving and Optimus humanoid robots to high-performance AI training.
During the second-quarter earnings call, Musk hinted at a convergence of AI efforts, stating, “Thinking about Dojo 3 and the AI6 inference chip, it seems like intuitively, we want to try to find convergence there, where it’s basically the same chip.” This news also emerges as Tesla’s board proposes a substantial $29 billion pay package for Musk, aimed at retaining his focus on the company’s AI advancements amidst his involvement with other ventures like xAI.
TechCrunch has reached out to Tesla for further comment on these developments.
