Elon Musk doesn’t want Tesla to be just an automaker. He envisions Tesla as an AI company focused on developing self-driving technology. Dojo, Tesla’s custom-built supercomputer, is central to this vision and is designed to train its full self-driving (FSD) neural networks.
The Role of Dojo in Tesla Vision
Despite its name, FSD is not fully self-driving and still requires human oversight. With more data, computing power, and training, Tesla believes FSD can achieve full autonomy. This is where Dojo plays a crucial role.
Early Mentions of Dojo (2019)
- April 22: At Tesla’s Autonomy Day, Musk first teases Dojo, describing it as a supercomputer for training AI.
Musk Begins the Dojo Roadshow (2020)
- February 2: Musk touts Dojo’s capabilities, emphasizing its ability to efficiently process vast amounts of video data.
- August 14: Musk calls Dojo a “beast” and mentions that the first version is about a year away.
- December 31: Musk notes that while Dojo isn’t necessary, it will enhance self-driving technology significantly.
Tesla Makes Dojo Official (2021)
- August 19: Tesla officially announced Dojo on its first AI Day, introducing the D1 chip, designed to power the Dojo supercomputer.
- October 12: Tesla releases a Dojo Technology whitepaper outlining technical standards for deep learning neural networks.
Tesla Reveals Dojo Progress (2022)
- August 12: Musk announces plans to phase in Dojo, reducing the need for incremental GPUs.
- September 30: At Tesla’s second AI Day, the company reveals the first Dojo cabinet and demonstrates its capabilities onstage, setting ambitious targets for 2023.
A ‘Long-Shot Bet’ on Dojo (2023)
- April 19: Musk describes Dojo as a “long-shot bet” but one worth making. It has the potential to significantly reduce the cost of training and become a sellable service.
- June 21: The Tesla AI X account posts about Dojo’s ongoing tasks and projects that Tesla’s compute will rank among the top five globally by early 2024.
- July 19: Tesla reports the start of Dojo production and plans to spend over $1 billion on the project through 2024.
- September 6: Musk highlights AI training compute limitations but expresses confidence that Nvidia and Dojo will overcome these challenges.
Plans to Scale Dojo (2024)
- January 24: Musk acknowledges Dojo as a high-risk, high-reward project, with plans to scale up and develop future versions.
- January 26: Tesla announces a $500 million investment to build a Dojo supercomputer in Buffalo, emphasizing its importance in its AI ambitions.
- April 30: At TSMC’s North American Technology Symposium. It’s revealed that Dojo’s next-generation training tile, the D2, is already in production.
- May 20: Musk confirms that part of the Giga Texas factory extension will house a dense, water-cooled supercomputer cluster.
- June 4: A CNBC report reveals Musk’s strategic allocation of Nvidia chips between Tesla and other ventures, emphasizing the importance of Dojo in Tesla’s AI training.
- July 1: Musk discusses potential hardware upgrades for Tesla vehicles to accommodate the next-gen AI model.
- July 23: During Tesla’s second-quarter earnings call, Musk acknowledged the challenge of securing Nvidia hardware, reaffirming the importance of Dojo in achieving Tesla’s AI goals.
- July 30: Musk mentions that AI5, Tesla’s next-generation AI model, is approximately 18 months away from high-volume production.
- August 3: Musk shares a walkthrough of Tesla’s supercomputer cluster at Giga Texas, highlighting its massive storage capacity for video training.
