Tesla has announced a new $500 million project to build a giant Dojo supercomputer cluster at Gigafactory New York in Bufallo.
Dojo is a new supercomputer designed from the ground up by Tesla specifically to train AI with videos.
The project suffered significant delays, but it seemed to be getting some momentum last year as the first Dojo cluster came online in the summer.
However, we also learned that Tesla let go of some of the program’s top leadership last month.
Furthermore, CEO Elon Musk described the project during Tesla’s earnings call this week as a “long shot” with a “not a high probability” of success.
I would think of Dojo as a long shot. It’s a long shot worth taking because the payoff is potentially very high. But it’s not something that is a high probability.
At the same time, Musk confirmed that Tesla is heavily investing in NVIDIA AI processors to build traditional supercomputer clusters to power its AI efforts.
These new comments about Dojo sort of threw some cold water on the program for many Tesla shareholders this week.
That’s why it’s surprising to now hear Tesla investing $500 million in a new Dojo cluster just a few days later.
New York State Governor Kathy Hochul announced the new project in Bufallo this week, where Tesla operates Gigafactory New York.
Musk quickly confirmed the news, but he added that Tesla is investing even more in NVIDIA hardware:
The governor is correct that this is a Dojo Supercomputer, but $500M, while obviously a large sum of money, is only equivalent to a 10k H100 system from Nvidia. Tesla will spend more than that on Nvidia hardware this year. The table stakes for being competitive in AI are at least several billion dollars per year at this point.
The CEO also confirmed that Tesla is going to start buying hardware for AMD as well.
Tesla is investing in all the processing power it can acquire in order to support its AI programs, like Full Self-Driving and Optimus.