168 AI Monetization…Data Collection + Supercomputing = Tesla AI Training Capacity +8.5x Tesla Dojo Custom Supercomputer – 6/21-9/24, per Tesla Note: Listing capacity in ‘H100-equivalent GPUs’ means Tesla converts the aggregate AI-training throughput of Dojo and its other accelerators into the number of NVIDIA Hopper H100 data-center GPUs that would deliver the same FP8/FP16 FLOPS, giving a single, industry-standard yard-stick for compute scale. Source: Tesla Q1:23 earnings call, Tesla Q3:24 investor presentation, Data Center Dynamics, Wikimedia Commons Tesla AI Training Capacity, H100 Equivalent GPUs We’re continuing to simultaneously make significant purchases of GPUs and also putting a lot of effort into Dojo [custom supercomputer], which we believe has the potential for an order of magnitude improvement in the cost of training… …Dojo also has the potential to become a sellable service that we would offer to other companies, in the same way that Amazon Web Services offers more web services, even though it started out as a bookstore. So, I really think that the Dojo potential is very significant. - Tesla Co-Founder & CEO Elon Musk, 4/23 AI Monetization = Compute Services Tesla AI Training Capacity (H100-Equivalent GPUs)