Tesla Shifts Gears: Dojo Team Disbanded as Musk Eyes New AI Partnerships

Tesla has decided to dismantle its Dojo supercomputer team, signaling a significant shift in its strategy for developing autonomous driving technologies. The Dojo project, which was initially seen as a pivotal component in Tesla's quest for AI supremacy, will be terminated as CEO Elon Musk has ordered its closure.

Peter Bannon, who led the Dojo team, is set to leave the company, further underscoring this strategic turnaround. Around 20 team members have already transitioned to a newly formed entity, DensityAI, while the remaining staff will be reassigned to Tesla's other data center and computing projects.

This pivot marks a departure from Tesla's previous path as it plans to increase its reliance on external technology partners. Collaborations with Nvidia and AMD will provide enhanced computing solutions, while chip manufacturing responsibilities will shift to Samsung Electronics.

Dojo had been a cornerstone of Tesla’s ambition to boost computational power in the competitive AI field. Analysts had long championed its potential, with some estimating it might contribute up to $500 billion to Tesla’s market value. As a bespoke supercomputing system, Dojo was key to training models for Tesla's Autopilot, Full Self-Driving (FSD) capabilities, and the Optimus humanoid robot.

The transfer of Dojo’s talent pool to DensityAI coincides with its emergence onto the scene. The startup, led by former Tesla engineers including Ganesh Venkataramanan and Bill Chang, is focusing on developing chips and software geared towards AI data centers, targeting applications in robotics and automotive sectors.

This shifting strategy stems from a year brimming with challenges for Tesla, including intensifying competition, declining sales, and criticisms regarding Musk's political activities. The company has seen a significant exodus of key personnel, with high-profile departures such as Milan Kovac, head of the Optimus project, and David Lau, VP of Software Engineering.

Last month, Tesla inked a $165 billion deal with Samsung to ensure a steady supply of AI semiconductors through 2033. Samsung's forthcoming Texas factory is set to produce Tesla’s next-gen AI6 chips, creating a diversified supply chain alongside TSMC.

Musk hinted at strategic adjustments during the latest earnings call, suggesting future iterations of Tesla’s proprietary technologies might integrate with partner solutions. "With Dojo 3 and AI6 inference chips, pragmatically we should seek convergence—ideally they merge into one chipset," he shared.

Despite earlier ambitions, Musk conceded in January that the Dojo initiative’s continuity was uncertain. He remarked, “While pursuing parallel paths with Nvidia and Dojo, Dojo represents a high-risk, long-term bet with substantial potential upside.”