Tesla Disbands Dojo Supercomputer Team in Major AI Strategy Shift  

Friday, Aug 8, 2025 9:22 am ET1min read
AI Podcast:Your News, Now Playing
Aime RobotAime Summary

- Tesla disbands its Dojo supercomputer team, ending a multiyear in-house AI project led by Peter Bannon, who is leaving the company.

- Around 20 former Dojo employees joined startup DensityAI, while remaining staff will shift to other Tesla computing projects.

- The strategic pivot prioritizes external partners like Nvidia, AMD, and Samsung for AI chips, signaling reduced reliance on proprietary hardware.

- Tesla shares fell 0.7% premarket, extending 2024 losses, as the move follows high-profile executive exits and intensifying industry competition.

- Elon Musk emphasized focusing on next-gen AI chips for vehicles, calling Dojo a "long shot" with uncertain but potentially high rewards.

According to people familiar with the matter,

is disbanding its Dojo team, with the project’s leader set to depart—a move that derails the automaker’s years-long effort to build an in-house supercomputer for self-driving development.  

Peter Bannon, who led the Dojo initiative, is leaving the company, and CEO Elon Musk has ordered the project shut down, according to insiders who requested anonymity to discuss internal matters. Roughly 20 employees from the team have recently joined the newly formed AI startup DensityAI, while the remaining staff will be reassigned to other Tesla data centers and computing projects.  

The decision marks a dramatic pivot for a project once considered central to Tesla’s multibillion-dollar bid to lead the AI race. The company now plans to rely more heavily on external partners, including

(180.77, +1.35, +0.75%) and for computing power, and Samsung Electronics (005930.KS) for chip manufacturing. Neither Tesla nor Bannon responded to requests for comment.  

After Bloomberg News first reported the development, Musk confirmed the strategy change on X, writing that Tesla’s next-gen AI chips—destined for its vehicles—“will be excellent for inference and at least pretty good for training. All effort is focused on that.”

Tesla shares fell as much as 0.7% in premarket trading on Friday in New York, extending this year’s decline to 20%.  

Designed to train machine-learning models for Tesla’s Autopilot, Full Self-Driving (FSD) system, and Optimus humanoid robot, Dojo was built around custom D1 chips tailored for AI training. The supercomputer processed vast amounts of vehicle-captured video data to refine Tesla’s algorithms—a capability analysts once predicted could become a $500 billion market value booster, per

.  

The Dojo team’s dissolution follows a wave of high-profile departures amid slumping sales, intensifying competition, and backlash over Musk’s political activities. Optimus engineering head Milan Kovac and software VP David Lau left earlier this year, while longtime Musk lieutenant Omead Afshar exited in June, Bloomberg reported.  

Tesla recently inked a $16.5 billion deal with Samsung to secure AI semiconductors through 2033, with a new Texas factory slated to produce its next-gen AI6 chips—diversifying supply beyond

.  

Musk hinted at the strategic shift during Tesla’s Q2 earnings call, suggesting a convergence between internal and partner tech: “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.”

As early as January 2024, Musk acknowledged Dojo was a “long shot,” telling investors: “It’s a long shot worth taking because the payoff is potentially very high. But it’s not something that is highly probable. It’s not like a sure thing at all.”

Comments



Add a public comment...
No comments

No comments yet