Tesla Halts Dojo AI Supercomputer, Key to Musk’s Self-Driving Vision
▼ Summary
– Tesla is disbanding its Dojo supercomputer team, ending in-house chip development for driverless tech, with lead Peter Bannon leaving and remaining members reassigned.
– Around 20 former Dojo employees left Tesla to start DensityAI, a new AI company focused on chips, hardware, and software for AI data centers in robotics and automotive applications.
– Elon Musk had previously positioned Dojo as central to Tesla’s AI and full self-driving ambitions, but the project is now being replaced by reliance on Nvidia, AMD, and Samsung for chips and compute.
– Tesla recently signed a $16.5 billion deal with Samsung to manufacture its AI6 inference chips, shifting away from in-house chip development like the D1 and planned D2 chips.
– The move coincides with Tesla’s board offering Musk a $29 billion pay package to keep him focused on Tesla’s AI efforts amid his involvement in other ventures like xAI.
Tesla has reportedly disbanded its Dojo supercomputer team, marking a significant pivot away from its in-house AI chip development for self-driving technology. The project, once touted by Elon Musk as central to the company’s autonomous driving ambitions, is being shelved as key personnel depart and resources shift toward external partnerships.
According to sources, Peter Bannon, Dojo’s lead, is exiting Tesla, with remaining team members reassigned to other data center initiatives. The move follows the departure of roughly 20 employees who left to launch DensityAI, a startup focused on AI hardware and software for robotics and automotive applications. The exodus includes former Dojo head Ganesh Venkataramanan, signaling a broader talent drain from Tesla’s AI division.
This strategic shift arrives at a critical juncture for Tesla. Musk has aggressively positioned the company as an AI and robotics leader, despite setbacks in its full self-driving (FSD) efforts. Earlier this year, Tesla’s limited robotaxi rollout in Austin faced scrutiny after reports of erratic driving behavior, highlighting the challenges of achieving true autonomy.
Dojo, first announced in 2021, was designed to process massive volumes of video data to accelerate FSD development. Musk once described it as a game-changer, with analysts predicting it could unlock $500 billion in market value through robotaxis and AI services. Yet, by mid-2024, attention shifted to Cortex, Tesla’s new AI training supercluster in Austin, suggesting internal priorities had evolved.
Originally, Dojo combined supercomputing with custom chip development. Tesla unveiled its D1 chip at AI Day 2021, promising a next-gen D2 chip to address performance bottlenecks. Now, the company appears to be leaning harder on Nvidia GPUs and partnerships with AMD and Samsung for its AI infrastructure. A recent $16.5 billion deal with Samsung to produce Tesla’s AI6 inference chips underscores this external reliance.
During Tesla’s Q2 earnings call, Musk hinted at consolidation, suggesting Dojo and AI6 chip development might merge. The timing coincides with Tesla’s board approving a $29 billion pay package to keep Musk focused on the company’s AI goals amid his sprawling ventures, including xAI.
While Tesla hasn’t officially commented, the dissolution of Dojo raises questions about the automaker’s ability to deliver on its autonomous driving promises, and whether external solutions will fill the void left by its abandoned in-house ambitions.
(Source: TechCrunch)
