Tesla's first supercomputer to enter production in July and cut reliance on Nvidia

Amanda Liang; Judy Lin, DIGITIMES Asia 0

Credit: Tesla

Despite Tesla CEO Elon Musk's strong criticism of the downside of artificial intelligence (AI), Tesla's AI team recently announced on Twitter the progress of Tesla's custom supercomputer platform Dojo, indicating that the computer will go into production in July 2023 and that Dojo is expected to be one of the world's five most advanced supercomputers by early 2024.

According to reports from Electrek and Teslarati, this is seen as another major move by Tesla in the AI field. Although Nvidia A100 and H100 GPU chips are the dominant chips in the AI field at this stage, Tesla's self-developed AI training and inference chips may reduce its dependence on traditional chip makers.

Dojo, a supercomputer dedicated to AI machine learning that Tesla launched at AI Day 2021, uses Tesla-designed chips and the entire infrastructure, as well as video data from the Tesla fleet, to train the neural network that is critical to supporting Tesla's machine vision technology for autonomous driving.

Although the Dojo system may not take final shape until 2024, Tesla CEO Elon Musk is pleased with the performance of his AI team, saying that Tesla's AI advancements, both in software and hardware, are far beyond what some experts have even realized.

Software is the key to autonomous driving, and Tesla is already using a large supercomputer powered by Nvidia GPUs to process FSD autonomous driving data, one of the world's most powerful supercomputing clusters.

Tesla announced its new supercomputer Dojo project as early as 2021, indicating a powerful computing center capable of handling a large number of AI tasks, designed to process large amounts of video data, accelerate the iterative computing of Tesla's Autopilot and FSD systems, and provide computing power for Tesla's humanoid robot Optimus.

Tesla's chief engineer Tim Zaman told the public that their computing cluster has only 0.3% idle time, 84% of which is spent on high-priority tasks. Therefore, they are in dire need of more computing resources, and Dojo supercomputers will play an important role in this regard.