Tesla's $500M Dojo Supercomputer, NVIDIA AI Chip Deal

On January 29th, Tesla announced the successful launch of its inaugural Dojo supercomputer last year. Presently, the company is poised to invest an additional $500 million to construct a cutting-edge Dojo supercomputer, driven by NVIDIA AI GPUs, with the primary objective of expediting advancements in artificial intelligence (AI) development.

Tesla's CEO, Elon Musk, shared on the X platform that the company plans to allocate substantial financial resources to the Dojo supercomputer initiative, initiating a $500 million project at the Gigafactory located in New York.

Musk underscored that the projected budget of $500 million would only cover the purchase of approximately 10,000 NVIDIA H100 units. However, he emphasized that Tesla's overall expenditure is expected to surpass this estimate, potentially reaching "several billion dollars" at a certain juncture. Musk affirmed, "Tesla's investment in NVIDIA hardware this year will exceed the aforementioned figure, as the company strives to sustain competitiveness in the AI sector, necessitating an annual outlay of at least several billion dollars."

In addition, Musk disclosed Tesla's intention to incorporate AMD chips, though specific details about the chip model were not provided. Speculation suggests it might be from the Instinct MI300 series, potentially the MI300X. This strategic move indicates Tesla's commitment to diversifying its hardware sources for AI applications.