Musk details Terafab chip fab plans in Austin

Elon Musk has formally launched Terafab, a new advanced chipmaking complex in Austin, Texas, intended to supply Tesla (TSLA), SpaceX and xAI with high‑performance AI chips. At an event in Austin on March 21, Musk described Terafab as “the most epic chip building exercise in history by far.” The project is structured as a joint venture between Tesla, SpaceX and xAI and is intended to manufacture AI chips at scale in the United States.

The facility will be located near Tesla’s Giga Texas campus and is designed to bring logic processing, memory production and advanced packaging under one roof. Musk framed the initiative as a response to anticipated shortages in third‑party chip supply for his companies’ autonomous driving systems, Cybercab robotaxis, Optimus humanoid robots and AI infrastructure.

Musk said Terafab will in practice consist of two separate fabs, each dedicated to a single chip design. One chip will be used in Tesla vehicles and Optimus humanoid robots, while the second will be designed for AI satellites and data centers in space. He said the space‑focused chip must withstand a harsher environment and operate at higher temperatures.

According to Musk, Terafab aims to eventually produce one terawatt of computing capacity per year. He compared this to about half a terawatt of computing capacity currently generated across the United States. Musk said global chip production as it stands would meet only a small fraction of the combined future needs of Tesla, SpaceX and xAI.

The Terafab complex is targeting 2‑nanometer process technology, described as the most advanced node currently in commercial production. Tesla’s AI5 chip is among the first products slated for the facility. Small‑batch production is expected in 2026, with volume production projected for 2027.

Initial capacity is planned at 100,000 wafer starts per month, with a longer‑term ambition to scale toward one million. Musk said the project is intended to produce enough computing power to support 100 to 200 gigawatts of AI infrastructure on Earth and ultimately a terawatt in space. He emphasized that Tesla will continue buying Nvidia (NVDA) chips in the meantime, positioning Terafab as a long‑horizon solution rather than an immediate shift.

Musk first highlighted chip supply as a growth constraint for Tesla on its January 2026 earnings call, saying that even a best‑case scenario with existing partners would not meet demand in three to four years. With millions of Optimus robots and Cybercab fleets on Tesla’s roadmap, he argued that external foundries are not prepared to commit to the required volumes on Tesla’s timeline.

During the Austin presentation, Musk said, “We either build the Terafab or we don’t have the chips,” adding that demand from his companies would eventually exceed total global chip output. He expressed gratitude to current suppliers Samsung, TSMC and Micron (MU), but said Terafab is needed to avoid a production ceiling tied to external capacity.

Tesla estimates the Terafab facility itself will cost between $20 billion and $25 billion. This outlay is in addition to Tesla’s existing 2026 capital expenditure guidance of more than $20 billion. Tesla’s CFO has said the full Terafab cost is not yet incorporated into that capex figure.

SpaceX’s role in Terafab marks an expansion beyond launch and satellite operations into chip manufacturing for space‑based AI data centers. The company has recently merged with Musk’s social media and artificial intelligence firm xAI and is preparing for a potential public listing that could value it at around $1.75 trillion, according to Reuters. Musk did not provide a detailed timeline beyond early production targets and has a record of announcing highly ambitious projects that have sometimes faced delays or not been completed as initially outlined.