Tesla begins production of Dojo supercomputer to train driverless cars

Tesla says it has begun production of its Dojo supercomputer to train its fleet of autonomous vehicles.

In its second-quarter earnings report for 2023, the company outlined “four core technology pillars needed to solve vehicle autonomy at scale”: extremely large real-world datasets, neural net training, vehicle hardware and vehicle software.

“We are developing each of these pillars in-house,” the company said in its report. “This month, we are taking a step toward faster and cheaper neural net training with the production start of our Dojo training computer.”

The automaker already has a large Nvidia GPU-based supercomputer that is one of the most powerful in the world, but the new Dojo custom-built computer is using chips designed by Tesla. In 2019, Tesla CEO Elon Musk gave this “super powerful training computer” a name: Dojo.

Previously, Musk claimed that Dojo would be capable of performing one exaflop, or 1 quintillion (1018) floating-point operations per second. This is an incredible amount of power. “To compare what an exaflop computer system can do in just one second, you would have to perform one calculation every second for 31,688,765,000 years,” network world wrote,

On Tesla’s AI Day in 2021, Dojo was still a work in progress. Officials revealed its first chip and training tiles, which will eventually evolve into a full dojo cluster or “Xapod”. Tesla said it would combine 2 x 3 tiles in one tray and two trays in a computer cabinet, for more than 100 petaflops per cabinet. In a 10-cabinet system, Tesla’s Dojo will break the exaflop barrier of exapod compute.

A year later, at AI Day 2022, Tesla revealed some progress on the Dojo, including a full system tray. At the time, the automaker talked about having the full cluster by early 2023 – although it now looks like early 2024 is more likely.