Following the reported end of its Dojo supercomputer project, Tesla is pivoting its AI strategy and will now partner with companies like Nvidia and Samsung for chips used in its autonomous driving tech.
Elon Musk has seemingly taken a U-turn in his artificial intelligence (AI) strategy for Tesla. The company has reportedly shut down its ambitious Dojo supercomputer project, with key talent departing the company for a new startup. The decision marks a major move away from CEO Elon Musk's long-held vision of developing in-house chips for autonomous driving technology.
Dojo Project Disbanded Amid Talent Exodus
The Dojo team has been officially disbanded, according to sources. The project's leader, Peter Bannon, is leaving the company, and the remaining employees are being reassigned to other data centers and compute projects within the company.
This strategic change comes on the heels of a mass exodus of key personnel. Approximately 20 workers from the Dojo team have departed to found a new AI startup named DensityAI. This new company, led by former Dojo head Ganesh Venkataramanan, is reportedly focused on building chips, hardware, and software for AI applications in data centers and robotics, a mission that directly mirrors the original goals of the Dojo project.
Shifting from In-House to External Collaboration
For years, Musk promoted Dojo as the cornerstone of Tesla's AI ambitions, a custom supercomputer designed to process truly vast amounts of video data needed to train its Full Self-Driving (FSD) software. The project's aim was to reduce Tesla's dependence on external suppliers like Nvidia and AMD by creating its own dedicated AI accelerator chips.
However, the reported shutdown indicates that Tesla is now moving toward a more collaborative approach. The company plans to significantly increase its reliance on external technology partners, including:
- Nvidia and AMD for its high-performance computing needs.
- Samsung Electronics for chip manufacturing. Tesla recently signed a substantial $16.5 billion deal with Samsung to produce its next-generation AI6 inference chips at a plant in Texas. These chips are intended to power FSD and the Optimus humanoid robots.
This shift suggests that despite Musk’s recent claims of progress on Dojo, the company has found it more efficient to leverage the expertise and scale of established semiconductor giants.
The Rise of Cortex and Strategic Priorities
The news follows a period where Musk’s public comments on Dojo began to wane, with the CEO instead starting to champion a different project: Cortex. Described as a giant new AI training supercluster being built at Tesla HQ in Austin, Cortex appears to be the new focal point for Tesla's AI development, suggesting a move away from the specific architecture of Dojo.
During a recent earnings call, Musk hinted at this strategic convergence, stating that Tesla would seek to align future iterations of its technology, such as the AI6 inference chip, with partner technologies to create a unified system. "Thinking about Dojo 3 and the AI6 inference chip, it seems like intuitively, we want to try to find convergence there, where it’s the same chip," Musk said.
The news of Dojo’s end comes as Tesla's board has offered Musk a new $29 billion pay package. This offer is reportedly designed to keep him at Tesla and push forward the company's AI efforts, rather than allowing him to be sidetracked by his other companies, such as his pure-play AI startup, xAI. The board's move underscores its belief that Musk's leadership is critical to Tesla's future as an AI and robotics company, despite the closure of a key project he once championed.
Impact and Outlook
While the closure of Dojo represents a notable setback for Tesla's in-house engineering efforts, it also highlights the company's willingness to adapt and innovate in the face of challenges. By outsourcing key hardware components, Tesla can potentially reduce costs, accelerate development timelines, and mitigate the risks associated with building cutting-edge chips from scratch.
However, this move also cedes some control and potential long-term differentiation to its suppliers and new competitors like DensityAI, which is staffed by former Tesla talent with intimate knowledge of the company’s original plans.
