In the rapidly advancing world of technology, one name stands out for its ambition to redefine the automotive industry: Tesla. Elon Musk envisions a future where Tesla is not merely a car manufacturer but a pioneering force in artificial intelligence, particularly in the realm of autonomous driving. This vision hinges on the development of advanced AI systems capable of enabling vehicles to navigate without human intervention.
Central to this ambitious goal is a groundbreaking supercomputer known as Dojo, specifically engineered to enhance the training of Tesla’s Full Self-Driving (FSD) neural networks. While FSD currently offers a range of automated driving features, it still necessitates human oversight. Tesla believes that with increased data, computational power, and rigorous training, the leap from semi-autonomous to fully autonomous driving is within reach.
As the narrative of Dojo unfolds, it has been a topic of intrigue and speculation, especially as discussions about its capabilities intensified throughout 2024. However, as the project progressed, a new supercomputer named Cortex emerged, shifting the focus away from Dojo.
Here’s a detailed timeline chronicling the key moments and announcements surrounding Dojo, illustrating its journey and the evolving landscape of Tesla’s AI ambitions.
April 22 – During a significant event focused on autonomy, Tesla showcased its AI team, discussing the intricacies of Autopilot and FSD. Musk introduced the concept of Dojo, describing it as a supercomputer dedicated to AI training. He emphasized that all Tesla vehicles produced at that time were equipped with the necessary hardware for full self-driving, pending only a software update.
February 2 – Musk announced that Tesla would soon surpass one million connected vehicles globally, equipped with the sensors and computational capabilities essential for full self-driving. He highlighted Dojo’s potential, stating it would process extensive video training data and operate hyperspace arrays with remarkable efficiency.
August 14 – Musk reiterated Tesla’s commitment to developing Dojo, describing it as a powerful tool for processing vast amounts of video data. He projected that the first iteration of Dojo would be ready within a year, aiming for an August 2021 launch.
December 31 – Musk acknowledged that while Dojo was not strictly necessary, it would significantly enhance self-driving capabilities, emphasizing the need for Autopilot to be substantially safer than human drivers.
August 19 – At Tesla’s inaugural AI Day, the company officially unveiled Dojo, aiming to attract top engineering talent. Alongside this announcement, Tesla introduced its D1 chip, which would power the Dojo supercomputer, with plans to integrate 3,000 D1 chips into its AI cluster.
October 12 – Tesla released a whitepaper detailing Dojo’s technology, outlining a new standard for binary floating-point arithmetic used in deep learning neural networks, which could be implemented in various configurations.
September 30 – At the second AI Day, Tesla revealed the installation of its first Dojo cabinet, showcasing its capabilities through a demonstration of an AI-generated image. The company set an ambitious target for completing a full Exapod cluster by the first quarter of 2023.
April 19 – During a quarterly earnings call, Musk highlighted Dojo’s potential to drastically reduce training costs and hinted at its future as a sellable service, akin to cloud computing offerings.
June 21 – Tesla’s AI division announced that its neural networks were already operational in customer vehicles, with projections indicating that Dojo production would commence in July 2023.
July 19 – In its second-quarter earnings report, Tesla confirmed the initiation of Dojo production, with plans to invest over $1 billion in the project through 2024.
January 24 – Musk reiterated the high-risk, high-reward nature of Dojo, stating that it was actively performing training tasks and that plans for future iterations were already in motion.
August 3 – Musk provided an update on the Tesla supercompute cluster, now referred to as Cortex, which would consist of a vast number of Nvidia GPUs, aimed at enhancing FSD and other AI applications.
August 26 – A video showcasing Cortex was shared, highlighting its role as a new AI training supercluster designed to tackle real-world AI challenges.
January 29 – In a quarterly earnings call, Tesla announced the completion of Cortex, which utilized a significant number of Nvidia GPUs to support the latest version of FSD.
July 23 – Musk indicated that Dojo 2 was expected to operate at scale by 2026, hinting at potential overlaps with future AI chip designs.
August 7 – Reports surfaced that Tesla had disbanded the Dojo team, signaling a shift in focus towards integrating AI capabilities into a unified chip design.
August 10 – Musk confirmed the decision to shut down Dojo, stating that the convergence of efforts towards a new chip design was necessary for the company’s future.
September 1 – Tesla shared its Master Plan Part IV, emphasizing the importance of AI in its future endeavors, although Dojo and Cortex were notably absent from the discussion.
This narrative was initially published on August 10, 2024, with the final update on the timeline occurring on September 2, 2025.