To reach his goals of scaling humanoid robots, self-driving cars and data centers in space, Elon Musk plans to undertake the “most epic chip-building exercise in history.” The project, called Terafab, is a joint initiative by SpaceX, Tesla and xAI to produce 1 terawatt, or 1 trillion watts, of computing power per year — a 50-fold increase from the 20 gigawatts of computing capacity currently produced by the entire semiconductor industry.
Musk believes the chips are essential to manufacturing a billion Optimus humanoid robots annually, deploying a million solar-powered data centers in orbit and ultimately transforming humanity into a “galactic civilization” that is so prosperous it no longer needs money.
What Is the Terafab Project?
Terafab is a joint initiative by SpaceX, Tesla and xAI to produce 1 terawatt of computing power per year to power autonomous robots, self-driving cars and data centers in space. Elon Musk said the chip industry produces only 2 percent of the computing power he believes his companies will eventually need.
The project is ambitious — even considering Musk’s penchant for overpromising. He will have to overcome numerous hurdles, including money, land, energy and the fact that none of his companies have any experience in the semiconductor industry. He has defied long odds before, though, as evidenced by Tesla commercializing the electric car and SpaceX reducing the cost to orbit through reusable rockets. Whether he can do the same in the semiconductor sector remains to be seen, but to hear him tell it, he has no choice but to try.
“We either build the Terafab or we don’t have the chips,” Musk said in an announcement on March 21. “We need the chips, so we’re going to build the Terafab.”
What Is the Terafab Project?
The Terafab project is a joint initiative by SpaceX, Tesla and xAI to produce 1 terawatt of computing power per year. Musk said he will continue to buy all of the available chips from his current suppliers, such as Samsung, TSMC and Micron, but he said the chip industry’s current production of 20 gigawatts per year is only 2 percent of the 1 terawatt he expects SpaceX, xAI and Tesla to use annually.
The Terafab project will start with the creation of an “advanced technology fab” in Austin, where chips will be designed, produced and tested under the same roof. By receiving instant feedback about a chip’s performance, Musk said technicians can make faster adjustments to the chip’s design template, known as a photomask, creating “an incredibly fast recursive loop for improving the chip design.”
“To the best of my knowledge, this doesn’t exist anywhere in the world where you’ve got everything necessary to build logic, memory and do packaging and test it and then do the masks, improve the masks and just keep looping it,” Musk said.
While most of the chip design will occur in the advanced technology fab north of the 10-million-square-foot Gigafactory in Austin, Musk clarified that the bulk of the fabrication will occur elsewhere, as it will require thousands of acres of land and 10 gigawatts of power at scale. He said the facility would be on the order of magnitude of roughly 100 million square feet. Neither Musk nor his companies have announced where the site will be or how it will be powered.
What Will the Chips Will Be Used For?
The Terafab project will produce two types of chips: one for robots and self-driving cars and another for data centers in space. Musk said he needs 1 terawatt of chips for orbital data centers and 100 to 200 gigawatts of chips for Optimus humanoid robots.
Chips for Optimus
The A15 and A16 chips will be optimized for edge inference for use in Tesla’s Optimus robots and full self-driving cars — but mostly the Optimus robots, which could hypothetically then work in the chip fabrication plants.
Eventually, Musk said he anticipates 1 to 10 billion humanoid robots will be manufactured every year, which is 10 to 100 times that of the roughly 100 million vehicles produced annually. Tesla will make a significant percentage of those robots, he said.
Chips for Orbital Data Centers
The second type of chip will be rugged, high-powered chips designed for space. Musk has long talked about scaling data center capabilities in outer space, and it was a main reason why SpaceX acquired xAI, Musk’s AI company, in February 2026. Other companies, most notably Blue Origin, Google and Nvidia-backed Starcloud have also explored orbital data centers.
To produce 1 terawatt of computing power from space, Musk said they would have to deploy 10 million tons of AI satellites into orbit per year on SpaceX’s Starship spacecraft. Each satellite would generate 100 kilowatts of computing power.
In space, the orbital data centers would receive five times more solar power than they would on Earth, Musk said, and they would not need battery storage because they would have constant access to the sun. As the cost of orbit drops, Musk said it will become a “no-brainer” to build data centers in space because Earth has limited power capabilities and room for data centers.
“Increasing power on Earth becomes harder over time and more expensive over time,” Musk said. “But in space, it becomes actually cheaper and easier over time.”
Cost and Timeline Remain Murky
The project will be expensive and may take some time to develop, as none of Musk’s companies have any experience in the semiconductor industry.
While Musk did not offer a cost estimate or timeline during his presentation, several media outlets have reported the project could cost $20 billion to $25 billion. Morgan Stanley estimates the price could be even higher: between $35 billion and $45 billion.
The capital-intensive project may be driving SpaceX's decision to go public, which could bring in $40 billion to $80 billion from investors, according to The Wall Street Journal. The company, which recently acquired xAI, announced several days after Musk’s Terafab announcement that the company could file paperwork to go public within the week, with an initial public offering expected sometime in June.
Gaining Control Over Chip Supply
By producing its own chips, Musk recently told Tesla investors that the company could solve its chip supply issues while protecting itself against geopolitical risks. While Musk’s companies’ demand for chips is extraordinary, chip shortages are a common pain point for tech companies that see more computing power as the key to winning the AI arms race.
Many hyperscalers and tech companies find themselves reliant on Nvidia, so several companies have sought to mitigate that reliance and experiment with different chip designs by developing their own chips. Google led the way with its Tensor Processing Unit (TPU), followed by Microsoft’s Maia chip, Amazon’s Trainium series and Meta’s Training and Inference Accelerator (MTIA). But still, most of these chips are fabricated by Taiwan Semiconductor Manufacturing Company, which manufactures roughly 90 percent of the world’s advanced chips. The entire industry’s reliance on this one company is a supply chain bottleneck in itself, but that vulnerability is exacerbated by China’s interest in gaining control over Taiwan.
In 2022, the U.S. sought to address this supply chain vulnerability with the CHIPS Act, which allocated more than $52 billion to support semiconductor manufacturing, research and development in the U.S. That funding has incentivized TSMC to build three fabrication plants in Phoenix, where the first is operational, a second was recently constructed and a third is under construction. The funding also helped U.S.-based Intel build a fabrication plant in Arizona, even as plans for an Ohio facility have been pushed back to 2030. Samsung plans to open a fab in Texas in 2026, and Micron is constructing fabs in Idaho and New York.
All of these facilities would pale in comparison to the size and scope of Musk’s proposed Terafab. To turn his vision into reality, Musk will have to overcome numerous hurdles, including a lack of skilled labor, high labor costs and access to specialized fabrication equipment like ASML’s extreme ultraviolet lithography machines.
None of that seems to deter Musk, who has built a career out of attempting things that were thought to be impossible. He appears to be going all-in, with the belief that the productivity gains of AI, robotics and space innovation will “unlock an age of amazing abundance” that will eliminate the need for money.
“I think things will just be free in the future,” Musk said. “It sounds nuts, but if you’ve got an AI robotics economy that is anywhere close to a million times the size of the current Earth economy, literally any need you possibly want can be met. If you can think of it, you can have it.”
Major Terafab Developments
The Terafab project is an ambitious initiative that will take years to execute, but could reshape the national chip manufacturing industry. These are some of the major developments as Tesla and SpaceX work to advance chip production and computing capabilities.
Intel Joins Terafab Project (April 2026)
Intel announced it will join Tesla and SpaceX in its project to build chip facilities and data centers in space. Although it is a lofty plan, neither Tesla nor SpaceX have experience in the semiconductors space, whereas Intel is one of the most prominent companies in the industry. Because of its experience, Intel may play a crucial role helping Terafab design and manufacture advanced chips.
Tesla Begins Hiring Chip Engineers (March 2026)
Days after announcing the project, Tesla began hiring Terfab engineers. According to Business Insider, Tesla posted listings for process, integration and silicon engineers in Palo Alto and Austin. The roles posted would directly contribute to designing silicon wafer chips.
Tesla and SpaceX Announce Terafab Project (March 2026)
SpaceX and Tesla, two companies led by Elon Musk, announced they will build two chip facilities in Austin, Texas that will produce chips for automobiles and humanoid robots. Additionally, the project will build data centers in space to handle the computing demands for AI platforms. When operational, Terafab will reportedly produce one terawatt of computing power and could cost between $35 and $45 billion.
Frequently Asked Questions
Where will Terafab be built?
The project will begin with an advanced technology fab just north of Tesla's 10-million-square-foot Gigafactory in Austin. The bulk of fabrication, however, will occur elsewhere, as it will require thousands of acres of land and 10 gigawatts of power — a facility on the order of roughly 100 million square feet.
How much will Terafab cost?
Elon Musk did not provide a cost estimate when announcing the Terafab project, but several media outlets have reported the project could cost $20 billion to $25 billion. Morgan Stanley's estimate is higher, putting the price between $35 billion and $45 billion.
What kinds of chips will Terafab produce?
Terafab will produce two types of chips: the A15 and A16 chips, optimized for edge inference in Tesla's Optimus robots and self-driving cars, and a second type of rugged, high-powered chip designed for use in orbital data centers in space.
