In a surprising twist, Elon Musk has officially shut down Tesla’s ambitious DOJO supercomputer project—a move that has left many industry insiders and Tesla enthusiasts puzzled.

The decision comes just four years after DOJO was unveiled during Tesla AI Day in 2021, where it was hailed as a revolutionary supercomputer designed specifically for AI training.

Musk’s announcement has sparked widespread speculation about the reasons behind the abrupt cancellation and what it means for Tesla’s future in artificial intelligence.

The Birth of DOJO: A Bold Vision for AI Training

DOJO was introduced as a groundbreaking computing system that promised to redefine AI training.

Unlike conventional supercomputers, DOJO was designed to be superefficient, capable of performing massive amounts of AI training with minimal energy consumption.

Tesla envisioned DOJO as the backbone of its ambitious plans for self-driving cars and humanoid robots, both of which were announced alongside the supercomputer at Tesla AI Day.

The architecture of DOJO was unlike anything seen before. At its core was the D1 chip, Tesla’s first custom-designed AI chip, which consolidated the CPU, GPU, and RAM into a single piece of silicon—a design known as “system-on-a-chip” (SOC).

Each DOJO tile, about the size of a dinner plate, contained 25 D1 chips working in unison.

Multiple tiles could be stacked together to form a cabinet, creating the DOJO supercomputer cluster.

Tesla’s engineers touted DOJO’s unique design as a game-changer for AI training.

By reducing network latency and optimizing power efficiency, DOJO was expected to outperform traditional AI training systems powered by Nvidia GPUs.

However, despite the initial hype, DOJO never lived up to its lofty promises.

The Fall of DOJO: A Project Plagued by Setbacks

Fast forward to 2025, and the DOJO project is officially dead.

The downfall began with the departure of key personnel, including Ganesha, the architect behind the D1 chip and DOJO’s overall design.

Ganesha left Tesla in 2023 to start his own AI company, Destiny AI, taking with him several top engineers from the DOJO team.

His departure marked the beginning of a talent drain that Tesla struggled to recover from.

Further complicating matters, the DOJO project faced technical challenges that hindered its progress.

Despite Tesla’s claims of superior efficiency, DOJO failed to demonstrate significant advantages over Nvidia-powered systems.

Tesla’s existing AI training cluster, powered by Nvidia GPUs, continued to deliver reliable performance, raising questions about the necessity of DOJO.

In a leaked internal memo, Musk reportedly expressed frustration with the project’s lack of progress.

“DOJO was an experiment, and experiments don’t always work out,” Musk wrote.

“We need to focus our resources on proven technologies that drive real-world results.”

Elon Musk’s Bold Decision: A Strategic Pivot to AI6

Musk’s announcement to shut down DOJO was accompanied by a cryptic post on social media, where he hinted at Tesla’s next steps in AI development.

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

“DOJO 2 was now an evolutionary dead end. DOJO 3 arguably lives on in the form of a large number of AI6 SOCs on a single board.”

So, what is AI6? Tesla’s AI6 chip represents the next generation of in-car inference computing.

Unlike DOJO, which was designed for AI training, AI6 is optimized for real-time inference—allowing Tesla’s vehicles to process complex AI models locally without relying on external data centers.

This capability is critical for Tesla’s full self-driving (FSD) technology, which requires instantaneous decision-making to navigate roads safely.

Musk’s decision to scrap DOJO reflects a strategic shift in Tesla’s priorities.

By focusing on AI6 and other inference computing technologies, Tesla aims to enhance the performance of its vehicles while reducing costs associated with developing and maintaining custom AI training systems.

The Nvidia Factor: Why DOJO Was Doomed from the Start

One of the most compelling reasons behind DOJO’s cancellation is Tesla’s reliance on Nvidia GPUs for AI training.

Nvidia has emerged as the dominant player in the AI hardware market, supplying chips to nearly every major AI company, including Tesla.

Musk himself has become Nvidia’s largest customer, reportedly purchasing over a million GPUs for Tesla and his other ventures, such as XAI.

While DOJO was designed to compete with Nvidia, it never demonstrated a clear advantage.

Tesla’s existing AI training clusters, powered by Nvidia GPUs, continued to deliver robust performance, making DOJO redundant.

Moreover, Nvidia’s rapid advancements in GPU technology have made it increasingly difficult for competitors to catch up.

Musk’s decision to double down on Nvidia reflects a pragmatic approach to AI development.

“It doesn’t make sense for Tesla to divide its resources and scale two quite different AI chip designs,” Musk wrote in his post.

“The Tesla AI5, AI6, and subsequent chips will be excellent for inference and at least pretty good for training. All effort is focused on that.”

The Legacy of DOJO: Lessons Learned and Future Opportunities

While DOJO may be dead, its legacy lives on in Tesla’s AI strategy.

The supercomputer’s innovative architecture has provided valuable insights into chip design and system integration, which Tesla can apply to future projects.

Musk has hinted at the possibility of building a new supercomputer cluster based on DOJO’s design but using AI6 chips instead of D1 chips.

Meanwhile, Ganesha and his company, Destiny AI, are working on full-stack AI solutions that bear a striking resemblance to Tesla’s original vision for DOJO.

By developing chips, computers, data centers, and software for high-performance AI applications, Destiny AI is essentially rebuilding Tesla’s AI division outside the company.

This development raises intriguing questions about the competitive dynamics between Tesla and its former employees.

Tesla’s Roadmap: What’s Next for AI Development?

With DOJO out of the picture, Tesla is focusing its resources on inference computing and autonomous vehicle technology.

The upcoming AI5 chip, set to debut in 2026, promises significant performance improvements over the current AI4 chip.

Musk has described AI5 as a “game-changer” for FSD, paving the way for Tesla’s long-awaited robo-taxi fleet.

Beyond AI5, Tesla is already planning for AI6, which is expected to launch by 2028.

Musk has hinted that AI6 will deliver unprecedented capabilities, potentially enabling Tesla to achieve full Level 5 autonomy—a milestone that has eluded the company for years.

In addition to in-car computing, Tesla is expanding its AI training infrastructure with the Cortex supercomputer at Giga Texas.

Powered by Nvidia GPUs, Cortex is driving progress in FSD development and other AI applications.

Tesla is also planning a second-generation Cortex supercomputer, further solidifying its reliance on Nvidia.

Conclusion: A Strategic Gamble That Could Pay Off

Elon Musk’s decision to shut down DOJO marks a pivotal moment in Tesla’s AI journey.

While the supercomputer failed to deliver on its promises, its demise reflects Musk’s willingness to pivot when faced with challenges.

By focusing on inference computing and leveraging Nvidia’s proven technology, Tesla is positioning itself for long-term success in autonomous driving and AI development.

As Tesla continues to innovate, the lessons learned from DOJO will undoubtedly shape the company’s future projects.

Whether through AI6, Cortex, or other initiatives, Tesla remains committed to pushing the boundaries of what’s possible in artificial intelligence.

And while DOJO may be gone, its spirit lives on in the bold vision that defines Tesla’s approach to technology.