A few days ago, Elon Musk said his company xAI could eventually take artificial intelligence into space. Musk has now announced TeraFab, a semiconductor manufacturing facility roadmap, to build chips needed to power this future. The massive undertaking, which would bring together Tesla, SpaceX and xAI, has only one goal: create a galactic civilisation so humans could become immortal in a way, their memories and ideas being preserved and passed on for generations to come.
Musk unveiled the TeraFab project during a live event at the Seaholm Power Plant in downtown Austin, Texas. Speaking at the launch, Musk described the project as an attempt to build the computing infrastructure required for what he called a “multi-planetary” and eventually “galactic civilisation.” According to him, the scale of AI computing needed in the future is so large that current chip production simply cannot keep up. That is why he is bringing Tesla, xAI and SpaceX together to build a dedicated facility capable of producing an unprecedented number of advanced processors.
What is TeraFab?
TeraFab will be led primarily by Tesla, with major involvement from SpaceX and integration with xAI, Musk’s artificial intelligence company. The project is expected to cost between $20 billion and $25 billion, and the factory could eventually produce between 100 and 200 billion advanced 2-nanometre AI chips every year, a number far beyond what most current semiconductor fabs can deliver today. These chips are expected to power everything from self-driving Tesla cars and humanoid Optimus robots to massive AI data centres in space.
Musk explained that, unlike a conventional semiconductor plant, TeraFab is designed to bring almost the entire chip-making process under one roof. The facility will handle logic chips, memory, packaging, testing, mask design, and continuous redesign loops within the same complex.
The chips produced at TeraFab will serve two main purposes. One category will be designed for Tesla products, especially its autonomous vehicles and the Optimus humanoid robot. Musk said Optimus could eventually be produced in far greater numbers than Tesla cars, meaning the company will need a huge supply of custom AI processors to support it.
The second category of chips, known as D3, is being designed specifically for space environments. These processors are intended to power orbital AI data centres that SpaceX could launch using its Starship rockets. Musk believes that running AI in space could eventually become cheaper than running it on Earth, partly because satellites can use continuous solar energy without the limitations of ground-based power grids.
According to Musk, once the cost of sending hardware to space falls low enough, putting computing infrastructure in orbit will become the next logical step. “So as soon as the cost to orbit drops to a low number, it immediately makes extremely compelling sense to put AI in space,” he said. “It becomes a no-brainer, basically.” This will include building networks of AI satellites equipped with large solar arrays capable of generating around 100 kilowatts each, with future versions potentially delivering even higher power levels.
Another reason for moving computing into space is energy. Massive AI data centres on Earth require enormous amounts of electricity, cooling, and land, and often face resistance from local communities. Musk suggested that space-based computing will have “abundance” of energy though sun, allowing almost unlimited scaling.
