Tesla is the proof of concept for physical AI at scale.
The stack includes:
• AI4 inference chips inside the vehicle
• Nvidia GPUs and Dojo chips training the self-driving AI model in data centers
• A robot on wheels embodying the inference computer running the FSD model
• FSD model generating tokens that instruct actuator movements — steering, braking, and acceleration
• FSD Subscriptions and Robotaxi services to capture the value created