Elon Musk wants to build his own chips, and he wants to do it at a scale no one has tried. Tesla and SpaceX have confirmed plans for Terafab, a semiconductor plant in Grimes County, Texas, that the companies bill as the largest chip manufacturing site on the planet. At more than 100 million square feet, it would dwarf today’s biggest factories, and its purpose is simple to state and staggering to picture, to make the chips that will run Musk’s robots, cars, and satellites.
The first phase alone is priced at about 16.8 billion dollars and is expected to create at least 3,000 jobs, most of them local, according to the companies, and that is only the beginning, since filings suggest the full build out across several phases could eventually cost as much as 119 billion dollars. Texas Governor Greg Abbott welcomed the project as a boost to advanced manufacturing in the state, and Musk said Texas offered fewer hurdles than California, where much of his business used to sit.
The reason for a factory this big is demand. Tesla and SpaceX say their combined need for chips will exceed one terawatt of compute a year, a figure that outstrips what the entire global chip industry can supply today. Rather than wait in line behind everyone else for that much silicon, Musk wants to make it himself, aiming, in his words, to “push the limits of physics in compute.” Terafab is meant to churn out the specialized chips his companies need at a volume the open market simply cannot promise.
The output is aimed squarely at Musk’s other bets. The plant would build edge and inference chips for Tesla’s Optimus humanoid robots and its self driving Cybercab, along with higher power silicon for the data centers SpaceX plans to fly in orbit. In other words, one Texas factory feeding a fleet of robots on the ground, robotaxis on the road, and AI computers in space. It is the kind of vertically integrated plan, chips, hardware, and finished products all under one owner, that few companies could even attempt.
Terafab is also unusual in how it would work. Instead of a standard fab that makes wafers and ships them elsewhere to be packaged and tested, the companies describe a single campus that houses logic chips, memory, packaging, and testing together. The pitch is speed. Keeping every step in one place, they argue, shortens the production cycle and lets them improve yields faster by iterating quickly. Whether that theory holds at this scale is one of the big unknowns.
And the unknowns are large. Chip fabs are among the hardest and slowest things in all of industry to build, often taking years and running over budget, and a 119 billion dollar campus making custom chips for products that are themselves still ramping, from Optimus to orbital data centers, stacks ambition on ambition. Musk has a long record of unveiling enormous plans on aggressive timelines that then slip. The renderings look spectacular and the strategy is coherent, but Terafab is an announcement and a first phase, not a running factory. The real test will be whether the largest chip plant ever proposed can actually turn out a single working chip on schedule.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →