Tesla shuts down Dojo, the AI training supercomputer that Musk said would be key to full self-driving

Tesla is dismantling the team behind its Dojo supercomputer, signaling a significant change in its approach to developing in-house chips for autonomous driving technology. According to reports from Bloomberg, the project's lead, Peter Bannon, is departing the company, and the remaining members of the team will be reassigned to other computing and data center initiatives within Tesla. This shift follows the exit of approximately 20 employees who have ventured out to establish their own AI startup, DensityAI. This new company, founded by former Dojo head Ganesh Venkataramanan along with ex-Tesla staff Bill Chang and Ben Floering, is set to unveil its operations soon. DensityAI aims to develop chips, hardware, and software that will enhance AI capabilities across various sectors, including robotics and automotive applications. The decision to close down the Dojo project comes at a critical juncture for Tesla. CEO Elon Musk has been advocating for shareholders to view Tesla as a leader in AI and robotics, despite the challenges faced during the limited rollout of its robotaxi service in Austin earlier this year, which included problematic driving behaviors in the vehicles. Musk had previously emphasized the importance of Dojo as a foundational element of Tesla's AI strategy, claiming its ability to process extensive video data was vital for achieving full self-driving capabilities. In light of this, analysts from Morgan Stanley had projected that Dojo could potentially enhance Tesla's market value by $500 billion by opening up new revenue avenues. However, discussions surrounding Dojo have waned since Musk began promoting Cortex, a new AI training supercluster being developed at Tesla's Austin headquarters. Initially announced at AI Day 2021, Dojo was designed as both a supercomputer and an in-house chip development project. Tesla's D1 chip was introduced during this event, intended to work alongside Nvidia's GPU for the Dojo setup. With the discontinuation of the Dojo initiative, Tesla is expected to increase its collaboration with external tech partners, such as Nvidia and AMD, to fulfill its computing needs. Recently, Tesla secured a $16.5 billion agreement with Samsung to produce AI6 inference chips, which are designed to support a wide range of applications from full self-driving technology to advanced AI training environments. During a recent earnings call, Musk alluded to possible redundancies within the company, indicating a strategic reevaluation of the resources allocated to AI initiatives. As Tesla navigates these changes, the board has proposed a lucrative $29 billion compensation package for Musk, aiming to keep him focused on Tesla’s ambitions in AI while managing his other ventures, including the AI-centric startup xAI.