Ilon Musk has published a 20-second video showing rows of servers with Nvidia H100 GPUs. They are designed for Tesla's new Cortex AI supercomputer. A total of 350,000 GPUs will be used to build the supercomputer. It is being built at Tesla's headquarters in Texas. Musk described the project on social network X as “a giant new supercluster for AI training and creating true artificial intelligence.” In the first phase, which is nearing completion, the system will include 50,000 Nvidia GPUs and specialized AI hardware. This is expected to provide fully autonomous driving functionality for Tesla vehicles. Musk estimates that the giant data center will consume 130 MW of energy, with power consumption rising to 500 MW as it develops. This is not Musk's only supercomputer project: another cluster is being built in parallel for startup xAI, which will utilize 100,000 Nvidia H100 GPUs and will be the basis for the next version of GrokAI, a chatbot for X's premium subscribers.