Tesla is developing its own AI inference chips, with AI5 slated for production in 2027 and AI6 in 2028, and the company’s Musk notes the AI5 chip will be optimized for edge computing in Robotaxi and Optimus deployments.

Terafab is part of a broader strategy to secure semiconductor supply, seen as a strategic constraint for Tesla and related ventures like SpaceX and xAI.

Tesla’s early adoption of custom inference computers across its fleet could prove crucial as demand for AI compute grows, a view shared by Musk and a Tesla engineer.

Engineer Yun-Ta Tsai warns that the current compute shortage is only the beginning, and autonomous driving will intensify the scarcity; Musk has publicly echoed this concern on X.

Tesla continues to design its own neural-network inference hardware, prioritizing chips that offer strong performance per watt for mass deployment in autonomous driving.

Terafab represents a long-term chip supply strategy to scale AI chips for Optimus, with the Austin development fab carrying an estimated cost of about $3 billion and aimed at speeding iteration across logic, memory, packaging, and testing.