Tesla's (TSLA 1.50%) quarterly earnings calls are always a highlight. CEO Elon Musk often says some wild things, but praise for competitors isn't often among them.
That's why it was noteworthy that Musk mentioned none other than AI chip leader Nvidia (NVDA -1.99%) on the call a couple of times. The implications for Nvidia shareholders are significant.
Tesla's Dojo will use Nvidia AI chips after all
A couple of years ago, Tesla officially announced its newest supercomputer, Dojo -- a massive data center used to train its full self-driving (FSD) software for its cars. Along with the announcement, it was revealed the electric vehicle leader had also developed its own chip, called the D1, to train the AI algorithms needed for FSD. This was a big deal, because up to this point, Tesla had been using Nvidia GPUs to power this AI training.
This is in keeping with Tesla's practice of designing its own hardware and software in-house when it deems it necessary, either to cut costs and dependence on suppliers, or when something it needs isn't available on the market. Back in 2018, Tesla announced it had designed its own chips for the on-board computers in its cars, ditching the systems Nvidia had designed for the EV maker (a lineup of GPUs it had dubbed "Tesla").
As far as I could tell, the Dojo D1 custom chip announcement was once again a defeat for Nvidia, which has tons of competitors trying to duplicate what it has accomplished in AI chip design. Few, if any, have been successful like Tesla has been in these efforts.
But then again, there's more to this story.
Musk doles out praise for Nvidia
Turns out that this new Dojo data center will be using plenty of Nvidia GPUs after all. Tesla needs to process an enormous amount of real-world footage recorded by its cars on the road (used to train the FSD algorithm, which relies entirely on cameras, versus a blend of cameras and other sensors adopted by other automotive companies). In commenting on this, Musk clarified that Dojo will use a blend of Nvidia and custom-designed D1 chips.
Later in the call, Musk expanded on those comments, doling out praise for Nvidia and its co-founder and CEO Jensen Huang:
We're using a lot of Nvidia hardware. We'll continue to use -- we'll actually take Nvidia hardware as fast as Nvidia will deliver it to us. Tremendous respect for Jensen and Nvidia. They've done an incredible job. And frankly, I don't know, if they could deliver us enough GPUs, we might not need Dojo. But they can't. They've got so many customers. They've been kind enough to, nonetheless, prioritize some of our GPU orders.
What Tesla will spend on Dojo in the next year
Musk also said that Tesla will spend "well over $1 billion on Dojo" through the end of 2024 (not just with Nvidia, but no doubt with other suppliers too). That's also significant, given that Dojo is already up and running now. This speaks to the value of the newest Nvidia AI systems' ability to quickly shred its way through massive amounts of video data. Dojo is in operation, but the latest and greatest hardware (as well as some Tesla in-house-designed software) can apparently offer a big upgrade.
As to Nvidia not being able to get enough GPUs to Tesla, those "many other customers" Musk refers to include all the public cloud computing infrastructure giants (each of which operate many data centers around the world): Amazon AWS, Microsoft Azure, Alphabet Google Cloud, and Oracle Cloud, to name just the biggest publicly traded four. All of these companies are seemingly in a race to add Nvidia hardware to their data centers so as not to be left behind in a new AI era.
The implied reality here is that Tesla's own D1 chip isn't as good as Nvidia's, but Tesla will have to make do with what's available.
Such is the position of strength Nvidia finds itself in right now. With generative AI (including self-driving car AI software) all the rage right now, the GPU designer is the premium option, and it can't keep up with demand. The question now is how long the line out the door will last for what is now the world's most valuable semiconductor company.
