Tesla completed the deployment of its 50,000 Nvidia H100 GPU training cluster during Q4 2024.
First reported by TechCrunch, Tesla’s shareholder deck noted that the cluster, which the company has dubbed Cortex, came online at the company’s Gigafactory in Austin, Texas.
The Nvidia-powered Cortex cluster is different to Dojo, Tesla’s supercomputer dedicated to [full-scale driving] FSD and featuring custom D1 chips and custom systems layouts.
There was no mention of Dojo in the slide deck.
While the company did not specify when during the quarter the deployment went live, speaking on an earnings call in October 2024, CFO Vaibhav Taneja said the company was “on track to get 50k GPUs deployed in Texas by the end of this month.”
The project was reportedly delivered behind schedule, with Musk firing the construction manager in April.
“Cortex helped enable V13 of FSD (Supervised), which boasts major improvements in safety and comfort thanks to 4.2x increase in data, higher resolution video inputs, 2x reduction in photon-to-control latency, and redesigned controller, among other enhancements,” the slide deck read.
“Progress on Optimus hardware and software continued in Q4, including the latest generation hand, robust locomotion, and training on additional tasks, ahead of planned pilot production in 2025.”
Optimus is Tesla’s humanoid robot, which the company intends to market to businesses and consumers.
Speaking on the company’s Q4 earnings call in late January, Musk said: “We continue to invest in training infrastructure out of our Texas headquarters. The training needs for the Optimus humanoid robot are probably at least 10x of what is needed for the car to get to the full range of useful roles.”
Tesla spent more than $10 billion on capex in 2024, and expects a similar outlay in each of the next two years, though much of this relates to infrastructure for its electric vehicles. On the same Q4 earnings call, CFO Vaibhav Taneja said the company’s total AI-related capex outlay, including infrastructure, is now more than $5bn.
