Elon Musk Ends Tesla’s Dojo Supercomputer Project, Declares AI Shift Toward AI6 Chip Evolution

In a major shift signaling the end of one of Tesla’s most ambitious AI endeavors, CEO Elon Musk announced over the weekend that the company has officially shut down its Dojo supercomputer project. The announcement, shared via a series of posts on X (formerly Twitter), marks a strategic redirection of Tesla’s AI resources and priorities.

The decision concludes a four-year effort to develop an in-house supercomputing system designed to support Tesla’s Full Self-Driving (FSD) technology. Musk confirmed the disbanding of the Dojo team, comprised of around 20 engineers, stating that Dojo 2 had become “an evolutionary dead end.”

Why Did Elon Musk Shut Down Tesla’s Dojo Supercomputer?

Tesla’s Dojo project, first revealed during the company’s AI Day in 2021, was hailed as a transformative step in the race to build proprietary AI infrastructure. Powered by the custom-designed D1 chip—manufactured by TSMC—Dojo was engineered to train Tesla’s autonomous driving models using vast volumes of real-world driving data.

However, internal challenges and technological roadblocks prompted a critical reassessment. Musk explained that with the development of more advanced AI chips like the AI5 and AI6, it no longer made sense to continue parallel investment in Dojo. The AI6 chip, co-developed with Samsung, is now positioned as Tesla’s unified hardware solution capable of handling both inference (real-time AI decisions) and training (deep data processing), making the Dojo architecture obsolete.

“Once it became clear that all paths converged to AI6, I had to shut down Dojo and make some tough personnel choices,” Musk wrote.

This move also reflects Tesla’s effort to consolidate resources, reduce complexity, and accelerate its AI roadmap. Musk emphasized that stacking AI5 and AI6 chips in supercomputer clusters would simplify design and reduce network cabling costs “by a few orders of magnitude.”

What Was Tesla’s Dojo and Why Did It Matter?

Dojo was envisioned as a Tesla-built alternative to NVIDIA’s AI infrastructure, promising greater bandwidth, lower latency, and reduced costs. Its core mission was to process terabytes of video and sensor data collected from Tesla vehicles, effectively enabling the self-driving AI to learn at unprecedented speed.

The project was also strategic: a successful in-house AI supercomputer would lessen Tesla’s reliance on external suppliers. But that vision hit multiple roadblocks.

In the past year, Dojo experienced internal turbulence, including the exit of key engineers. While Tesla had made progress with its D1 chip, the fast pace of AI chip innovation rendered the architecture less competitive. The company ultimately concluded that Dojo could no longer deliver the necessary performance gains to justify its operational and development costs.

How Will Tesla’s New AI Chips Shape the Future?

With the AI5 and AI6 chips now taking center stage, Tesla aims to create a scalable, unified AI hardware platform. The AI5 chip, built in partnership with TSMC, is optimized for deployment in Tesla vehicles and robots. Meanwhile, the AI6 chip—produced by Samsung—is engineered for large-scale AI training, potentially serving across Tesla’s supercomputing infrastructure.

By pivoting to these next-generation chips, Tesla is betting on efficiency, scalability, and speed. Unlike Dojo, these chips can be used across multiple platforms, reducing the need for distinct architectures and enabling Tesla to deploy AI advancements more rapidly.

This pivot is also timely, given Tesla’s broader business context. The company’s electric vehicle revenue fell by 16% in the last quarter, with its U.S. market share dropping below 50%, down from 75% in 2022. As Tesla repositions itself, streamlining its AI strategy becomes essential.

Expert Analysis: A Strategic AI Realignment

Analysts suggest this strategic realignment could sharpen Tesla’s focus on real-world AI deployment. “Dojo was always an ambitious project, but it’s not uncommon in tech to pivot once better technology emerges,” said a senior chip engineer familiar with the company’s operations.

The shift may also strengthen Tesla’s position in the broader AI market, especially as competition intensifies. With powerful chips like AI6 at its disposal, Tesla can now prioritize integration and application over infrastructure development—an approach more in line with the rapid iteration cycles of AI innovation.

What’s Next for Tesla’s AI Development?

While the Dojo project may be over, Tesla’s AI ambitions are far from diminished. The new chip lineup suggests a leaner, faster AI infrastructure that could support not only FSD but also Tesla’s other ventures, including Optimus (the humanoid robot) and broader robotics initiatives.

Musk’s announcement hints at a future where Tesla’s AI systems are more integrated, modular, and adaptable. The move away from a standalone supercomputer to a chip-centric model signals the company’s intent to stay ahead in both the automotive and AI arms race.

Elon Musk may have shut down Dojo, but the vision of an AI-driven Tesla future is still very much alive—just running on a different chip.

For your information:

What was Tesla’s Dojo supercomputer used for?
Tesla’s Dojo was designed to train its Full Self-Driving (FSD) AI using massive amounts of real-world driving data, enabling more accurate autonomous navigation capabilities.

Why did Elon Musk call Dojo 2 an ‘evolutionary dead end’?
Musk concluded that advancements in Tesla’s new AI6 chip made the Dojo 2 architecture obsolete, leading to the project’s termination to streamline development efforts.

What is the Tesla AI6 chip?
AI6 is Tesla’s next-generation AI chip developed with Samsung, designed for both training and inference tasks, making it more versatile than previous chip models like D1.

How does this shift affect Tesla’s AI strategy?
By consolidating around AI5 and AI6, Tesla simplifies its AI development, reduces costs, and accelerates the deployment of self-driving and robotics capabilities.

What impact does this have on Tesla’s EV business?
The shift allows Tesla to redirect engineering talent and resources toward solving current challenges in its core electric vehicle segment, which has seen revenue decline.

Is Tesla still pursuing supercomputing for AI?
Yes, but instead of a standalone Dojo system, Tesla will now use clustered AI5/AI6 chips to build more cost-effective and scalable AI infrastructure.