Tech • AI • Robotics • Game
Incredible Inside Elon Musk's New $119 Billion Terafab Chip Factory
9.4/10TL;DR
Terafab is an extraordinarily ambitious Texas chip project tied to Tesla, SpaceX and xAI, designed to secure semiconductors at a scale Elon Musk argues existing global supply cannot provide.
KEY POINTS
A chip factory built for internal survival
Terafab is framed as a strategic response to a looming semiconductor shortage across Musk-linked companies rather than a conventional merchant foundry. The project is aimed at supplying chips for Tesla vehicles, Optimus humanoid robots, Cybercabs, SpaceX systems and expanding xAI computing needs. Musk has said current chip supply may cover only about 2% of expected future demand for these plans.
Huge site and investment claims
The project is described as a facility on a roughly 100 million-square-foot site with investment estimates reaching $119 billion. Musk’s stated goal is chip output measured in terawatts per year, a scale he argues would justify the name Terafab. If achieved, the annual output value has been compared to the size of Switzerland’s economy, underscoring how far the concept stretches beyond a normal factory expansion.
Three chip families at the core
Terafab is expected to target three chip lines: AI5, AI6 and D3. AI5 is intended for AI workloads in Tesla vehicles and Optimus, while AI6 would support later generations of products with greater computing needs. D3 is aimed at more specialized applications linked to SpaceX, including computing systems that may eventually operate in space.
Optimus and robot-scale demand
The scale of the chip requirement becomes clearer in the robot program alone. Musk has discussed production of tens of millions of Optimus robots annually, with some projections reaching 100 million units a year. Even at 10 million robots a year, one dedicated processor per robot would mean at least 10 million AI chips before counting cars, robotaxis or data-center hardware.
Space-based computing raises the bar
SpaceX is exploring the idea of AI data centers in orbit, which would demand processors able to withstand radiation, severe temperature swings and stricter reliability thresholds than terrestrial servers. That is why D3 matters to the broader strategy. The effort is not only about keeping assembly lines moving on Earth, but also about preparing hardware for future off-planet computing infrastructure.
A manufacturing rethink, not just more capacity
The project is pitched as an attempt to redesign how chips are made. Modern fabs require hundreds of steps including deposition, lithography, etching, implantation, cleaning, inspection and packaging, and wafers often spend much of their total cycle waiting or being transported. Terafab aims to pull logic, memory, advanced packaging and testing closer together to cut the time from design to finished chip.
Austin research hub before full build-out
Before the main production lines rise in Grimes County, Texas, a research operation is planned inside Gigafactory Texas in Austin. The idea is to test process changes and production methods there first, then transfer successful improvements into mass manufacturing. Musk’s goal is to shrink experimentation cycles from about one year to roughly one month, a radical target in an industry known for slow validation.
ASML, Intel and TSMC are central to feasibility
The plan depends heavily on outside expertise and equipment. ASML is crucial because advanced production requires EUV lithography systems using 13.5-nanometer wavelength light to print tiny circuit patterns, and each machine costs hundreds of millions of dollars. Reports also point to involvement from Intel, discussions with TSMC, and outreach to suppliers such as Lam Research and Applied Materials, highlighting that success will require a vast ecosystem, not just funding.
Equipment bottlenecks and yield risk remain severe
Securing EUV tools may be one of the hardest near-term hurdles, with supply expected to be tight in the 2027-2028 period as chipmakers compete for limited deliveries. Even if the machines arrive on time, building a stable line with strong manufacturing yield is another challenge entirely. In semiconductors, small process errors can turn entire wafers into scrap, and experience from established manufacturers often matters as much as capital.
Power and water needs resemble an industrial city
Publicly discussed infrastructure plans suggest more than 40 gas turbines at roughly 50 megawatts each, for about 2 gigawatts of generation capacity. Tesla Megapack batteries could be used to stabilize supply because even short voltage disturbances can disrupt sensitive chip production. Water may come from Gibbons Creek Reservoir, with an on-site wastewater treatment system, indicating the project requires its own utility backbone before high-value tools can run.
CONCLUSION
Terafab is shaping up as a bid to vertically secure the semiconductor foundation for Tesla, SpaceX and xAI at a scale rarely attempted by a nontraditional chipmaker. Its outcome will depend less on rhetoric than on whether equipment, infrastructure and manufacturing yield can match the project’s immense ambition.
Ask a question