Elon Musk has officially christened Tesla’s new supercomputer cluster at Giga Texas as ‘Cortex.’ This behemoth of a system will be fueled by a staggering 100,000 NVIDIA H100/H200 AI GPUs, paired with ‘massive storage’ for video training its FSD (Full Self-Driving) and Optimus robotics projects.
The announcement of the supercomputer’s new name came from Musk himself in a post on X, a move that isn’t surprising considering his leadership roles at SpaceX, Tesla, and X. He divulged that Cortex will boast around 100,000 NVIDIA H100 and H200 AI GPUs, which will be dedicated to training the neural networks powering FSD and Optimus robotics.
Musk has emphasized the sheer scale of the cooling system required for this supercomputer cluster, stating that it needs around 130MW of power for cooling this year alone. This requirement is set to skyrocket to over 500MW in the next 18 months. This week, the cooling system’s fans were installed, captured in drone shots by Giga Texas (now Cortex) observer Joe Tegtmeyer.
The new Cortex supercomputing cluster also utilizes four gigantic water tanks, providing crucial assistance in cooling the system. To accommodate this massive cooling infrastructure, Tesla had to construct a dedicated structure on the second floor of the building.
