Part 3/11:
Tesla's strategy involves assembling multiple "training tiles" into larger Exapods, which are then integrated into cabinets, forming the backbone of their supercomputing infrastructure. The plan originally forecasted the first Dojo systems to be operational by early 2023, with scale-up to around seven exapods at Tesla’s Palo Alto facility by the end of 2024. If Tesla achieves its objectives, they could deploy substantial Nvidia A100 GPU performance—potentially matching or exceeding Nvidia’s total shipments—marking a significant step towards faster autonomous driving training.