A Chronology of Tesla’s Dojo Development

Elon Musk aims to transform Tesla from simply a car manufacturer to a trailblazer in AI, mastering autonomous vehicles.

At the heart of this vision is Dojo, Tesla’s proprietary supercomputer, which spearheads the training of its Full Self-Driving (FSD) technology. Although FSD has not yet achieved complete autonomy and still necessitates an alert driver, Tesla believes that with an increase in data, computational power, and extensive training, full autonomy is within reach.

Dojo plays a pivotal role in this ambition.

Musk has gradually increased discussions about Dojo, highlighting its critical nature to Tesla, especially as the company seeks to reassure investors amidst declining EV sales by achieving autonomy. A chronology of Musk’s mentions and commitments concerning Dojo follows.

2019

Unveiling of Dojo

April 22 – During Tesla’s Autonomy Day, the focus was on Autopilot and FSD, powered by AI. Tesla took the opportunity to discuss its custom-designed chips for neural networks and autonomous driving.

Musk introduced Dojo as a powerhouse supercomputer dedicated to AI training, mentioning that all Tesla vehicles being produced were equipped with the necessary hardware for full self-driving and required only software enhancements.

2020

Dojo’s Promotional Campaign Begins

Feb 2 – Musk declares the imminent milestone of over a million Tesla vehicles worldwide, equipped for full self-driving, praising Dojo’s capabilities to process an immense volume of video training data and run sophisticated simulations efficiently.

“Anticipate more on Dojo’s prowess,” he stated.

August 14 – Reaffirming Tesla’s strategy, Musk describes Dojo as a “monster” system, aimed at deciphering enormous sets of video data. He forecasted that the system’s debut would be in about a year, around August 2021.

December 31 – Musk mentions Dojo’s non-essential but beneficial role in enhancing self-driving capabilities, stressing the goal for Autopilot to surpass human drivers significantly in safety.

2021

Formal Introduction of Dojo

August 19 – Tesla’s first AI Day spotlighted Dojo, aimed at attracting top engineering talent. The event also showcased Tesla’s D1 chip, integral to Dojo’s operation, set to complement Nvidia’s GPUs within Tesla’s AI cluster, which will consist of 3,000 D1 chips.

October 12 – The release of the Dojo Technology whitepaper provided insights into Tesla’s unique approach to computational algorithms for deep learning neural networks, capable of being implemented across both hardware and software platforms.

2022

Unveiling Dojo’s Advancements

August 12 – Musk suggests a gradual integration of Dojo, which would reduce Tesla’s dependence on additional GPUs in the following year.

September 30 – Details disclosed during Tesla’s second AI Day revealed the installation of the inaugural Dojo cabinet, undergoing significant testing phases. Tesla also disclosed its ambitious target to establish a fully operational Exapod cluster by Q1 2023, with plans to construct a total of seven Exapods in Palo Alto.

2023

A Venture Into the Unknown

April 19 – In a discussion with investors, Musk envisioned Dojo as a revolutionary advancement in reducing the costs of AI training, potentially evolving into a marketable service to rival offerings like Amazon Web Services.

Musk depicted Dojo as a daring venture but a gamble that Tesla was willing to take.

June 21 — The Tesla AI X account highlights the integration of Tesla’s neural networks in customer vehicles, demonstrating a projected increase in computational power, with Dojo’s operations flagged to commence in July 2023.

Future projections hinted at Tesla’s computational capabilities ranking among the world’s top five by February 2024 and achieving 100 exaflops by October 2024.

July 19 – Tesla announced the beginning of Dojo’s production phase, with Musk signaling Tesla’s investment exceeding $1 billion in Dojo through 2024.

September 6 – Musk communicated on X about the challenges posed by handling vast quantities of video data from Tesla vehicles, indicating the vital role of Nvidia and Dojo in overcoming these obstacles.

2024

Scaling Up Efforts

January 24 – Musk reaffirmed Dojo as a significant, albeit high-stakes, initiative during Tesla’s earnings call, highlighting the ongoing cooperation with Nvidia and the plans for future iterations of Dojo.

January 26 – Tesla’s commitment towards a $500 million investment in constructing a Dojo supercomputer in Buffalo was announced. Musk tempered expectations by comparing this investment to the costs associated with Nvidia hardware, emphasizing the considerable financial outlay required to remain competitive in AI technology.

April 30 – At TSMC’s symposium, the advancement to the next-generation training tile for Dojo, dubbed D2, was disclosed, signalling a significant leap in integrating the entire Dojo tile onto a singular silicon wafer.

May 20 – Musk revealed plans to include a supercomputer cluster in the Giga Texas factory extension, further underlining Tesla’s commitment to AI development.

As Tesla forges ahead in its AI endeavors with Dojo at the helm, the journey encapsulates the blend of ambitious technological advancement and the hurdles of scaling AI infrastructure—a testament to the company’s dedication to autonomy and innovation.