According to Sawyer Merritt on X, Tesla stated that "Cortex 2 is now online and has started running training workloads," underscoring an accelerated ramp of onsite training infrastructure to secure compute for AI products and services, and continued investment in custom silicon development (source: Sawyer Merritt). According to Tesla’s statement shared by Merritt, the move signals deeper vertical integration across model training and inference, enabling lower latency, cost control, and faster iteration cycles for autonomy and robotics use cases (source: Sawyer Merritt). As reported by the same post, expanding in‑house training clusters and custom chips positions Tesla to reduce dependence on external cloud GPUs and improve training throughput for FSD and humanoid robotics, creating potential cost and performance advantages for commercial AI deployments (source: Sawyer Merritt).