Tesla Builds Largest AI Supercomputer. Why Is It Important?

In July 2023, Tesla announced that the long-awaited launch of its Dojo supercomputer has finally gone into production. Dojo is a custom-built platform, specifically designed for AI machine learning and video training from its vehicle fleet.

Reading Time: 2 minutes

tesla dojo

Illustration: Lenka T.

What will Tesla’s Dojo be used for?   

Unlike Tesla’s existing NVIDIA GPU-based supercomputer, Dojo utilizes chips and an infrastructure created entirely in-house. This new supercomputer is set to significantly enhance Tesla’s capacity to train neural nets using video data, revolutionizing the computer vision technology that drives their self-driving efforts. 

The initial unveiling of Dojo took place during Tesla’s AI Day in 2021, but the company was still in the process of ramping up its efforts at the time. Throughout 2022, Tesla made considerable progress on Dojo, culminating in a full system tray showcased at AI Day 2022. 

However, the original timeline for its operation by Q1 2023 was delayed. 

Now, Tesla has confirmed the production commencement of Dojo through its Tesla AI Twitter (or X) account. The plan is to continuously add trays and cabinets to the system, ultimately transforming it into one of the most powerful supercomputers worldwide by early 2024. 

Moreover, Tesla has ambitious intentions to surpass this achievement significantly by the end of the following year. 

Namely, with Dojo’s capabilities, Tesla aims to leverage its vast real-world driving scenario database, amassed over the years through millions of vehicles. 

This supercomputer’s deployment will enable Tesla to harness the full potential of its accumulated data, propelling their self-driving technology to new heights. 

Why is this important?   

Now, let’s put things into perspective. 

Let’s start from the burning question: 


Is Musk using chips designed by Tesla because they’re better than the ones NVIDIA provides, or because he can’t access them anymore?

According to Fortune, the latter is correct; NVIDIA isn’t valued at more than trillion dollars only because of the AI hype but rather because their chips are indeed super powerful and the best on the market, without any close competition. 

That’s why, the demand is huge from companies worldwide. In other words, the supply is limited, and Musk isn’t happy with the situation. His solution was to produce chips himself. That’s not to say Tesla’s product isn’t good, it’s just that we’re yet to see how everything will play out. 

From what we know so far, Musk was never modest when it comes to announcements or patient enough to think things through – he prefers to do them overnight, as we’ve seen with Twitter becoming X, and many other drastic changes. 

Either way, Tesla is not the only company that was dissatisfied with running out of AI chips. In fact, many are falling down and slowing their AI innovation. The fact that Tesla managed to produce their own chips puts them into an incomparable advantage. 

But that’s not all. 

Another key information lies in the fact of how Dojo works. 


Part of Dojo’s job is to help train the car’s ‘neural nets’. It does this by watching hours of Tesla camera videos and explaining what’s happening in them. These videos will also be used to train Musk’s Open AI competitor ‘x.AI’. This means x.AI has access to a new and unique type of data set that no other A.I model will have access to.

Source: Synthetic Mind 

In other words, thanks to Dojo, Musk has just entered the top 5% of AI leaders club. 

Needless to remind you, he has this big dream of a taxi service completely run by self-driving cars, called “Robotaxi”. The only thing stopping them until now is that Tesla needed more power than was possible for a computer to generate. 

But, now that they’ve made this huge progress with Dojo, the company has come close to making a huge leap in the autonomous cars industry and has become the ultimate leader in the race. 

A journalist by day and a podcaster by night. She's not writing to impress but to be understood.