Tesla has confirmed a $500 million funding to assemble a Dojo supercomputer at its Gigafactory in Buffalo, New York.
The Dojo supercomputer, first unveiled at Tesla’s AI Day in 2021, is an bold challenge aimed toward processing intensive video information from Tesla’s fleet of automobiles. This information is essential for coaching the AI algorithms that energy Tesla’s Full Self-Driving (FSD) Beta software program, a key part within the firm’s quest for a completely autonomous driving answer.
Regardless of latest setbacks, together with the departure of key challenge leaders, the corporate has introduced it’ll spend $500 million to carry a Dojo supercomputer to Giga New York. The funding was first introduced by New York State Governor Kathy Hochul at a press convention on Friday, information which was later confirmed by Musk on X.
Musk additionally confirmed that along with the Dojo challenge, Tesla is considerably investing in NVIDIA AI processors, with plans to spend extra on NVIDIA {hardware} than the allotted $500 million for the Dojo supercomputer in 2024.
The governor is right that this can be a Dojo Supercomputer, however $500M, whereas clearly a big sum of cash, is barely equal to a 10k H100 system from Nvidia.
Tesla will spend greater than that on Nvidia {hardware} this yr. The desk stakes for being aggressive in AI are at…
— Elon Musk (@elonmusk) January 26, 2024
Tesla final yr introduced plans to spend “nicely over $1 billion on Dojo”, with the aim of it turning into one of many world’s prime 5 supercomputers by early 2024. The supercomputer is projected to achieve 100 exaflops by later this yr. To place it into perspective, an exaflop signifies a supercomputer’s potential to carry out a minimum of 10^18, or one quintillion, floating-point operations per second.
After the departure of challenge lead Ganesh Venkataramanan late final yr, former Apple govt and present director Peter Bannon has taken over the reigns.