Tesla has unveiled Terafab, a large-scale semiconductor fabrication facility developed jointly with SpaceX and xAI, marking a significant strategic push by Elon Musk to bring advanced chip manufacturing in-house across his network of companies.
The facility, formally announced on March 21, 2026, at the historic Seaholm Power Plant in Austin, Texas, is designed to produce custom AI chips for Tesla’s autonomous driving systems, Optimus humanoid robots, and broader data center workloads. Musk described the project as a gigafactory-scale initiative aimed at securing chip supply that he argues existing foundry partners cannot reliably deliver at the volumes he requires.
Terafab is targeting cutting-edge 2-nanometer process technology and is projected to cost between $20 billion and $25 billion across its development phases. The facility’s stated annual output goal is more than one terawatt of AI compute capacity, a figure that underscores the scale of Musk’s broader infrastructure ambitions.
The fab is engineered to produce Tesla’s next-generation AI5 chips, which are intended to serve both edge inference applications in vehicles and robots, and higher-performance variants for data center and space-based computing environments. Musk has indicated that approximately 80 percent of Terafab’s compute output is earmarked for orbital AI satellites, part of a longer-term vision to make space-based AI infrastructure cost-competitive with ground-based alternatives.
The announcement positions Terafab as a direct response to supply chain constraints that have increasingly challenged AI hardware development at scale. By internalizing semiconductor production, Tesla, SpaceX, and xAI aim to reduce dependence on third-party foundries such as TSMC and Samsung for their most critical compute needs.
Terafab represents one of the most capital-intensive bets Musk has made on vertical integration, extending a strategy that has defined Tesla’s manufacturing approach since its early years.
Read Article: Nvidia Faces Rising Competition From Hyperscalers and AI Chip Rivals
