Tesla CEO Elon Musk said that the automotive company will "double down" on its Dojo custom AI chip supercomputer, at the same time as installing more Nvidia GPUs.
Shares in Tesla fell more than eight percent in premarket trading after the company reported second-quarter earnings that missed expectations, declining revenue, a fall in profits of 45 percent, and a delay in its 'robotaxi' launch event.
Tesla CFO Vaibhav Taneja said that despite a dip in Q2 capex, the company still expected capital expenditures to pass $10 billion. "As we increase our spend to bring a 50,000 GPU cluster online, this new cluster will immensely increase our capabilities to scale [full-scale driving] FSD and other AI initiatives."
Musk said that the company had "been working 24/7 to complete the South extension on the Tesla Giga factory in Texas. That South extension is what will house 50,000 H100s and we're beginning to move the H100 server racks into place there."
His comment came in response to a question about internal Nvidia emails that revealed plans to ship thousands of GPUs reserved for Tesla to his private ventures X and xAI. "The Tesla data centers were full," he said.
"That was in Tesla's interest, not contrary to Tesla's interest. [It] does Tesla no good to have GPUs that it can't turn on. That South extension is able to take GPUs, which is really just this week. We are moving the GPUs in there and we'll bring them online."
Musk did not comment on Nvidia emails that noted the $10bn capex figure “conflicts with bookings and FY 2025 forecasts."
In the earnings call, Musk also said that the company was "going to double down on Dojo, and we do see a path to being competitive with Nvidia with Dojo." Dojo is Tesla’s supercomputer dedicated to FSD, with custom D1 chips and custom systems layouts.
"What we are seeing is that the demand for Nvidia hardware is so high that it's often difficult to get the GPUs," Musk said. "I'm quite concerned about actually being able to get state-of-the-art Nvidia GPUs when we want them.
"And I think this therefore requires that we put a lot more effort into Dojo in order to ensure that we've got the training capability that we need. We are going to double down on Dojo, and we do see a path to being competitive with Nvidia with Dojo. We kind of have no choice because the demand for Nvidia is so high and - it's obviously their obligation essentially to raise the price of GPUs to whatever the market will bear, which is very high. I think we've really got to make Dojo work and we will."
On X, Musk said that "we also use the Tesla HW4 AI computer in the training loop with Nvidia GPUs, currently at roughly a 1:2 ratio. Also, we’re changing the name from Hardware 4 (HW4) to Artificial Intelligence 4 (AI4)."
He added: "Dojo 1 will have roughly 8,000 H100-equivalent of training online by end-of-year. Not massive, but not trivial either."
Also on X, Musk launched a poll asking users whether Tesla should invest $5 billion in his AI startup xAI. "Board approval & shareholder votes are needed, so this is just to test the waters," Musk said.
