Elon Musk Unveils Dojo Supercomputer to Rival Nvidia
Tesla's Dojo aims to mitigate GPU shortages and enhance AI training capabilities, targeting performance equivalent to 8,000 Nvidia H100 GPUs.
Overview
- Musk announced the Dojo supercomputer during Tesla's Q2 earnings call, emphasizing its necessity due to Nvidia GPU shortages.
- Dojo will use Tesla's in-house D1 chips, with plans to upgrade to D2 chips in the future.
- By the end of 2024, Tesla aims to have AI training capacity equivalent to 90,000 Nvidia H100 GPUs.
- The Dojo supercomputer will support Tesla's Full Self-Driving technology and other AI initiatives.
- Musk shared images of the Dojo system, highlighting its significant processing power and efficiency.