Tesla Pulls the Plug on Dojo: A Strategic Shift in AI and Chip Development

Tesla has quietly dismantled its ambitious Dojo supercomputer project, marking a dramatic pivot in the company’s AI strategy and sparking new opportunities for a startup led by former Tesla engineers.

For years, Dojo was seen as Tesla’s crown jewel in the race toward full self-driving (FSD) technology — a powerful in-house supercomputer built to process the mountains of video data generated by Tesla’s global fleet. But according to Bloomberg, the automaker has now dissolved the entire Dojo team, with its leader set to leave the company. This move not only signals a change in technical direction but also underscores a growing reliance on external partners like Nvidia, AMD, and Samsung for AI and computing solutions.

the Situation

Tesla has officially disbanded its Dojo supercomputer division, a project launched in 2019 with the goal of creating a proprietary AI training platform. This team, led by Peter Bannon, is now being reassigned to other data center and compute initiatives within the company. Bannon’s departure follows the earlier exit of around 20 Dojo members who left to form DensityAI — a stealth-mode AI hardware startup founded by Ganesh Venkataramanan, Dojo’s former head, alongside ex-Tesla engineers Bill Chang and Ben Floering.

DensityAI is gearing up to unveil its plans for developing chips, hardware, and software optimized for AI-driven data centers, robotics, AI agents, and automotive applications. The company aims to deliver scalable, energy-efficient computing solutions that could compete directly with industry heavyweights.

The decision to shutter Dojo comes just as Tesla’s board approved a staggering \$29 billion compensation package for Elon Musk, intended to ensure his focus remains on Tesla’s AI and robotics ambitions over other ventures like his AI startup, xAI.

Tesla’s shift coincides with challenges in its self-driving program. Its limited robotaxi pilot in Austin, Texas — using Model Y vehicles with human safety drivers — faced public criticism after reports of erratic driving behavior.

Originally unveiled during Tesla’s AI Day in 2021, Dojo was built around Tesla’s custom D1 chip, designed to work in tandem with Nvidia GPUs, with plans for a next-gen D2 chip to address data bottlenecks. In 2023, Morgan Stanley analysts predicted Dojo could add \$500 billion to Tesla’s market value by enabling high-margin robotaxi and software services. But by August 2024, Musk began pivoting toward “Cortex,” a new AI training supercluster based at Tesla’s Austin headquarters, signaling the beginning of Dojo’s decline.

Tesla has now deepened its relationship with outside suppliers, most notably through a \$16.5 billion partnership with Samsung to produce AI6 inference chips — versatile processors designed for everything from autonomous driving to humanoid robots. The company is also expanding collaborations with Nvidia and AMD to scale its AI infrastructure, with Musk hinting at merging aspects of Dojo 3 into the AI6 chip to streamline development.

What Undercode Say:

The shuttering of Dojo is more than just a corporate reorganization — it’s a symbolic moment in the ongoing evolution of Tesla’s identity. On one hand, it represents a pragmatic decision to double down on proven partnerships with tech giants like Nvidia and Samsung rather than fighting an uphill battle in the fiercely competitive custom chip market. On the other, it signals the loss of an audacious in-house vision that, if successful, could have given Tesla a unique competitive edge.

Dojo’s failure to reach commercial maturity despite years of investment highlights the extreme difficulty of building cutting-edge AI infrastructure from scratch. While Tesla is a master at hardware-software integration in vehicles, competing with specialized chipmakers — each backed by billions in R\&D and decades of engineering — is an entirely different arena.

From a business strategy perspective, this pivot could be seen as a resource optimization move. Instead of pouring capital into internal chip development, Tesla can now leverage best-in-class technology from partners while focusing on what it does best: integrating those systems into vehicles, robotics, and services. This decision might also accelerate timelines for key projects like the Optimus humanoid robot or its long-promised fully autonomous fleet.

DensityAI’s emergence adds an intriguing subplot. The ex-Dojo engineers are not retreating from the AI hardware race — they’re simply re-entering it on their own terms, free from Tesla’s corporate priorities. If DensityAI succeeds, it could compete with Nvidia in niche AI hardware markets, or even supply Tesla in the future. Ironically, the company that killed Dojo might one day buy chips from the very people who built it.

Financially, the \$29 billion compensation package for Musk is a bold move by the board to keep his attention locked on Tesla. But with Musk’s attention divided between SpaceX, Neuralink, The Boring Company, and xAI, it’s fair to question whether this package will translate into sustained focus.

The move toward Cortex and AI6 suggests Tesla is not abandoning AI — far from it. The company is betting that external partnerships plus internal system integration will get them to autonomy faster than going solo with a massive chip R\&D project. However, this also means Tesla will be more dependent on supplier roadmaps and supply chain risks, a notable shift from Musk’s usual preference for vertical integration.

In the bigger picture, the Dojo shutdown is a reminder that visionary projects must eventually meet the hard realities of cost, competition, and execution. For investors, it’s a signal that Tesla is entering a new phase — less about moonshot hardware bets, more about execution and scaling with partners.

🔍 Fact Checker Results

✅ Bloomberg and multiple industry insiders confirm Dojo’s shutdown and leadership departure.
✅ DensityAI is indeed founded by former Dojo engineers and preparing to exit stealth mode.
❌ No evidence yet that Dojo’s closure was due to a technical failure — strategy appears to be the main driver.

📊 Prediction

Tesla will increasingly market Cortex as its flagship AI infrastructure, positioning it as more efficient and scalable than Dojo. Within the next 2–3 years, AI6 chips from Samsung will likely become the backbone of Tesla’s FSD and robotics programs, reducing dependence on Nvidia over time. DensityAI may secure early partnerships with smaller robotics firms before targeting large-scale automotive contracts — and could eventually return to Tesla’s supply chain in a surprising twist.

If you want, I can also add a section on how this affects Tesla’s stock price trajectory and investor sentiment. That would make the analysis even sharper for a news blog.