Key Points:
- Elon Musk announced “Terafab,” a massive semiconductor production facility.
- It’s a joint venture by Tesla, SpaceX, and xAI, located at Giga Texas.
- Musk believes current chipmakers can’t meet his companies’ massive AI needs.
- Terafab aims for 2-nanometer chips and over a terawatt of AI computing power annually.
Over the weekend in Austin, Texas, Elon Musk, CEO of Tesla and SpaceX, shared what might be his biggest vision yet for his companies: “Terafab.” This huge project is a partnership between Tesla, SpaceX, and xAI, aiming to bring every step of semiconductor production under one roof. This includes designing chips, making them, producing memory, and packaging them. Terafab will be built on the North Campus of Giga Texas, in a building expected to be even bigger than Giga Texas itself, which is already one of the largest buildings on Earth.
The initial cost for Terafab is estimated to be between $20 billion and $25 billion. Musk noted that Tesla’s planned spending for 2026 doesn’t even include these Terafab costs. The timing is also interesting, especially with SpaceX possibly going public this spring.
Musk called Terafab “the most epic chip-building exercise in history by far.” He’s pursuing this project because he believes chipmakers like TSMC and Samsung simply aren’t making chips fast enough to keep up with his companies’ enormous needs for AI and robotics.
“We either build the Terafab or we don’t have the chips, and we need the chips, so we build the Terafab,” he stated. He claimed that current global AI compute output is about 20 gigawatts per year, but his companies alone need about 2% of that total.
Terafab will focus on two main types of chips. One is an “edge-inference processor” specifically designed for Tesla’s Full Self-Driving systems, Optimus humanoid robots, and Robotaxi fleets. The other is a high-power chip built for space environments, supporting SpaceX satellites, orbital data centers, and xAI projects.
The Optimus robot program is the biggest driver of this demand. An analyst from Morgan Stanley suggested that Giga Texas alone could produce 10 million humanoid robots per year. This would require 20 million chips, which is roughly six times what Tesla currently needs for its entire car business. If Tesla hits its long-term goal of 100 million Optimus robots annually, it would need over 200 million chips, more than 50 times its current demand across cars and Robotaxis combined.
Interestingly, Terafab plans to use 2-nanometer process technology, which is the most advanced in the world. Even established chipmakers like TSMC are only just starting to produce chips at this level.
Musk’s ultimate goal is to produce more than a terawatt (a trillion watts) of AI computing power each year, with most of it eventually going into space. He didn’t give exact dates for construction or when it would be fully operational but said the project would start with prototyping and testing the infrastructure.
