Tesla steps up a gear, plans Dojo supercomputer to train driverless cars


Elon Musk said that Tesla wants to invest over $1 billion on its so-called Project Dojo by the end of 2024. The project plans to create the supercomputer needed to run smarter software for self-driving cars.

According to Musk, who was speaking during a conference call with investors on Wednesday and is quoted by Bloomberg, the Dojo supercomputer is being designed to handle huge amounts of data, including video from Tesla cars needed to create autonomous-driving software.

Tesla’s owner added that the firm has a “staggering amount” of video available. This is because customers have been eagerly using Autopilot, a camera-based driver-assistance system, and a related feature marketed as Full Self-Driving Beta that has collected over 300 million miles of data.

ADVERTISEMENT

In its latest earnings release, the company said that it had already begun production of its Dojo training computer.

“Four main technology pillars are needed to solve vehicle autonomy at scale: extremely large real-world dataset, neural net training, vehicle hardware, and vehicle software. We are developing each of these pillars in-house,” said Tesla.

“This month, we are taking a step towards faster and cheaper neural net training with the start of production of our Dojo training computer.”

The automaker already has a large NVIDIA GPU-based supercomputer, which is one of the most powerful in the world, but the new Dojo custom-built computer is using chips designed by Tesla.

According to Musk, that’s because NVIDIA – which recently crossed the $1 trillion valuation threshold – cannot supply more of its advanced A100 tensor core GPU clusters.

“Frankly if they could deliver us enough GPUs, we might not need Dojo,” Musk told analysts on Wednesday during Tesla’s Q2 earnings call. “But they can’t, because they have got so many customers.”

Musk has previously claimed that Dojo will be capable of an exaflop, or 1 quintillion (​​1018) floating-point operations per second. That is a huge amount of power.

“To match what a one exaFLOP computer system can do in just one second, you’d have to perform one calculation every second for 31,688,765,000 years,” Network World wrote back in 2020.

ADVERTISEMENT