Tesla Abandons Dojo to Bet Big on AI6’s Unified Power

Tesla CEO Elon Musk has outlined a new direction for the automaker’s AI hardware ambitions, emphasizing the development of AI5 and AI6 chips as central to its future. The AI6 chip, in particular, is described as the “best AI chip to date,” while AI5 is positioned as an “epic-level chip” designed to enhance Tesla’s Full Self-Driving (FSD) and robotics capabilities. These announcements come amid the recent discontinuation of the Dojo AI supercomputer project, which Musk described as an “evolutionary dead end” following six years of development [1].

The decision to pivot away from the custom-built Dojo supercomputer reflects a strategic reevaluation of Tesla’s AI infrastructure. Dojo was initially intended to serve as a proprietary AI training system for FSD, designed to process vast quantities of video data from Tesla’s global vehicle fleet. However, the project faced technical, financial, and staffing challenges, including key departures such as Peter Bannon, the project’s lead, and 20 team members who went on to form a new AI startup, DensityAI [1].

Musk cited the convergence of development paths toward AI6 as a key factor in the shutdown of Dojo. “Once it became clear that all paths converged to AI6, I had to shut down Dojo and make some tough personnel choices, as Dojo 2 was now an evolutionary dead end,” he posted on X [1]. The AI6 chip is being developed in collaboration with Samsung and is intended to serve dual roles in both training and inference, potentially streamlining Tesla’s AI operations across autonomous vehicles, robotics, and data center applications. The AI5 chip, meanwhile, is focused on inference tasks, allowing for more efficient decision-making in real-world environments [1].

Tesla’s shift to AI6 aligns with broader industry trends where companies are increasingly favoring scalable, off-the-shelf solutions combined with custom silicon. The company has also expanded its use of Nvidia’s H100 and H200 GPU clusters in its Cortex supercomputer, which has been deployed at Gigafactory Texas. This move enhances Tesla’s flexibility and capacity for AI training and model development, particularly for FSD and Optimus, its humanoid robot [1].

The cancellation of Dojo has not dampened Tesla’s long-term AI ambitions. In fact, the company is investing $500 million into a new supercomputer in Buffalo, New York, though it will not be Dojo. Instead, the funds will support a system leveraging AI6 and GPU-based compute power. Analysts note that the pivot reflects a pragmatic approach to balancing innovation with practicality, especially given the complexities of developing and maintaining a custom supercomputer [1].

Public and investor reactions to the Dojo shutdown have been mixed. Critics argue that it represents another missed opportunity for Musk, particularly as Tesla grapples with challenges in its electric vehicle (EV) sales and the rollout of its robotaxi service. Others view the decision as a necessary pivot from a high-risk, self-reliant hardware strategy to a more flexible model that integrates partnerships with leading chip manufacturers like Nvidia, AMD, and Samsung [1].

Moving forward, Tesla’s focus on AI5 and AI6 underscores its commitment to advancing AI capabilities in autonomous driving and robotics. Musk has previously stated that Optimus robots will account for 80% of Tesla’s future valuation, highlighting the strategic importance of these technologies. The integration of AI6 into Tesla’s infrastructure could potentially enhance the performance and scalability of its AI models, enabling faster iterations and broader deployment of autonomous features [3].

As Tesla continues to refine its AI roadmap, the lessons from the Dojo project remain integral. The project’s legacy lies in the insights it provided into the challenges of proprietary AI hardware and the benefits of hybrid approaches combining custom and commercial solutions. With AI5 and AI6 now at the forefront, Tesla aims to maintain its leadership in AI-driven innovation while addressing the logistical and economic realities of modern semiconductor development [1].

Source:

[1] Tesla Dojo: The rise and fall of Elon Musk's AI supercomputer (https://techcrunch.com/2025/09/02/tesla-dojo-the-rise-and-fall-of-elon-musks-ai-supercomputer/)

[2] Tesla Powers Down Dojo: The End of Elon Musk's Ambitious AI Supercomputer (https://opentools.ai/news/tesla-powers-down-dojo-the-end-of-elon-musks-ambitious-ai-supercomputer)

[3] Elon Musk's Unfulfilled Promises: From Robots to Supercomputers (https://forklog.com/en/elon-musks-unfulfilled-promises-from-robots-to-supercomputers/)