Elon Musk unveils Terafab in an effort to unlock an AI supercomputer in space
Elon Musk unveiled Terafab, a hyper-scale chip manufacturing plant aimed at unlocking supercomputer AI and eventually supporting space-based infrastructure and human expansion beyond Earth.
Tesla, SpaceX, and xAI, now part of Musk’s aerospace company, are developing the project together. Their goal is to produce one terawatt of computing each year, about 50 times the current global output of AI chips.
“The way to advance civilization is to increase energy in space…because we’re actually capturing a small amount of the sun’s energy on Earth,” Musk said in a recent SpaceX broadcast.
“We want to be a civilization that reaches the galaxy with spaceships and anyone can go wherever they want at any time,” he added. “To do that, we need to use solar energy. Terafab, although it’s huge, a terawatt of computing power per year is huge by our civilization’s standards, it’s still just one step away from being a Kardashev.”
Production center
Terafab will integrate the full cycle of chip development in one place, according to Musk. The system will include capabilities for lithography mask creation, chip fabrication, testing, and redesign, allowing a rapid feedback loop to iterate on chip designs.
Musk suggested that this approach could significantly speed up development cycles compared to the fragmented structure of today’s chip supply chain.
The project is expected to begin with an advanced manufacturing facility in Texas, supported by federal support.
Two categories of chips
The step looks at two different categories of chips. The first will be developed to accommodate the kind of onboard processing required by Tesla’s Optimus humanoid robots and its fleet of autonomous vehicles, including the upcoming Cybercab.
Musk projects that the production of humanoid robots could eventually reach 10 billion to 10 billion annually, which would reduce the roughly 100 million cars produced worldwide each year.
A variant of the second chip will be purpose-built for space conditions, designed to withstand high-energy particle blasts, and designed to operate at higher temperatures to reduce the weight of thermal radiators on orbiting platforms.
Why space, not Earth
Musk argued that the world’s energy constraints make it practically impossible to use a terawatt of computing on Earth, where US electricity production hovers around 0.5 terawatts. Instead, much of the computing infrastructure will orbit the planet via solar-powered AI satellites.
Specifications for the “small satellite” prototype call for 100 kilowatts of output, with future iterations rising to the megawatt range. Achieving the full terawatt target would require launching about 10 million tons of material into orbit each year with an efficiency of 100 kilowatts per ton.
The current variant of the Starship V3 can deliver around 100 tons per payload, a figure the upcoming V4 version could double to 200 tons. SpaceX has completed more than 500 successful booster launches and launched launch costs from over $65,000 per kilogram during the Space Shuttle era to roughly $1,000 to $2,000 per kilogram today.
The company’s stated desire is to push that figure to between $100 and $200 per kilogram for Starship operations, a range that Musk believes will make space-based AI deployments cheaper than ground-based alternatives within two to three years.
In support of this change, Musk pointed to the development of reusable launch systems such as Starship, which he said will be very important for transporting the large equipment needed.
“A starship is a critical piece of the puzzle, because in order to measure computing power, you have to go to space, which means you need a large payload in space. And Starship will do that,” Musk said.
He also outlined long-term concepts, including the use of moon-based manufacturing and mass propulsion to further reduce the cost of infrastructure deployment in orbit.
Gap scale
Global AI computing capacity is about 20 gigawatts per year. All the world’s semiconductor manufacturing plants combined account for about 2% of what Terafab would need to reach its one terawatt annual goal.
Musk emphasized that existing suppliers, including the world’s largest foundations, remain critical, but noted that their comfortable growth rate falls far short of his needs.
“Either we build Terafab or we don’t have the chips,” he said. “And we need chips, so we built Terafab.”
