Tesla and SpaceX are no longer simply talking about making more of their own computer chips.
They are planning semiconductor manufacturing on a scale that could dwarf almost anything the industry has built before.
In August 2026, Tesla and SpaceX confirmed plans for the first major phase of Terafab, a semiconductor manufacturing complex in Grimes County, Texas. The initial investment is expected to reach approximately $16.8 billion, while the companies envision a completed manufacturing campus containing more than 100 million square feet of space.
If that number sounds difficult to visualize, that is because it is enormous even by megafactory standards.
The ultimate goal is not merely to build another chip fabrication plant. Elon Musk wants Terafab to vertically integrate logic chips, memory, advanced packaging, testing, and even lithography-mask production while eventually generating enough chips to support Tesla’s autonomous vehicles and Optimus robots alongside SpaceX’s increasingly ambitious artificial-intelligence infrastructure.
And according to Musk, existing semiconductor suppliers simply will not be able to make enough chips for what his companies are planning.
Why Tesla and SpaceX Suddenly Want Their Own Chip Factory
Tesla already depends heavily on custom silicon.
Its vehicles process huge quantities of camera and sensor information, while the company’s Full Self-Driving software relies increasingly on large neural networks. The planned Cybercab could make that requirement even greater because Tesla’s vision assumes autonomous cars will continuously perform AI inference as they navigate roads.
Then there is Optimus.
Tesla wants to manufacture humanoid robots at enormous scale. Every robot would require substantial computing hardware to interpret cameras, process its environment, make decisions, control movement, and interact with humans.
SpaceX has an entirely different requirement.
The company increasingly views orbital computing as part of its long-term infrastructure strategy. That could eventually involve enormous constellations of satellites carrying AI compute hardware specifically designed to operate in space.
Musk has argued that the combined future demand from Tesla, SpaceX, and xAI could exceed what existing semiconductor manufacturers can supply. Reuters reported that he characterized the problem unusually simply: either the companies build Terafab or they will not have enough chips.
That turns semiconductor manufacturing from a supplier relationship into a strategic bottleneck.
Terafab Could Eventually Cover 100 Million Square Feet
The planned scale is where Terafab starts sounding less like a conventional semiconductor facility and more like another Musk industrial megaproject.
SpaceX and Tesla have said the completed Grimes County complex could contain more than 100 million square feet of manufacturing space. Musk has described it as potentially becoming the largest and most valuable building on Earth.
For comparison, Fortune noted that such a structure would be more than five times larger than the building currently recognized among the world’s largest by floor area.
The project is expected to create at least 3,000 jobs in Grimes County and neighboring Brazos County, although a facility of this scale would almost certainly create a much larger economic ecosystem involving equipment manufacturers, construction companies, utilities, chemical suppliers, logistics operators, and semiconductor specialists.
The first $16.8 billion phase would therefore be only the beginning.
Documents connected with the project have indicated that future expansion could potentially push total investment far higher, with figures approaching $119 billion appearing in development plans if the campus ultimately reaches its envisioned scale.
This Is Actually More Than One Chip Factory
The name Terafab can make the project sound like one enormous manufacturing line.
The strategy is more complicated.
When Musk first publicly described the project in March 2026, he said Tesla and SpaceX planned two advanced semiconductor fabs, each optimized around different kinds of computing requirements.
One family of chips is intended primarily for terrestrial applications.
These processors would support Tesla vehicles, autonomous driving, and Optimus humanoid robots, where extremely efficient real-time AI inference matters.
The second would be designed for SpaceX’s orbital computing ambitions.
Semiconductors operating in space face different challenges from those inside a car or conventional data center. Radiation, extreme temperature variation, power limitations, and reliability requirements can all influence chip design.
SpaceX itself has described Terafab as supporting one chip family optimized for terrestrial edge inference and another optimized for the space environment and future orbital compute infrastructure.
In other words, Tesla needs brains for robots and autonomous machines on Earth.
SpaceX wants brains that can operate above it.
Tesla Wants Almost the Entire Chip Process Under One Roof
One of the most unusual parts of Terafab may not be its physical size.
It is the amount of semiconductor production that Musk wants to combine inside it.
Modern semiconductor manufacturing is extraordinarily fragmented. A chip may be designed by one company, fabricated by another, packaged elsewhere, combined with memory made by another manufacturer, and tested through additional suppliers.
Terafab aims to compress much more of that process into one vertically integrated operation.
Tesla’s own Terafab recruiting material says the project is being designed to house logic, memory, packaging, testing, and lithography-mask production under one roof.
Why make masks there as well?
Speed.
When engineers design a new processor and discover something that could be improved, conventional supply chains can make each experimental iteration expensive and slow.
If chip design, mask creation, fabrication, memory integration, packaging, and testing occur inside the same organization, Tesla believes engineers can run that cycle much faster.
SpaceX has made the same argument, saying vertical integration could allow rapid testing and manufacturing improvements while giving the companies more control over the foundational computing layer.
That philosophy should sound familiar.
Tesla spent years bringing battery packs, electric motors, power electronics, software, and manufacturing engineering closer together.
SpaceX did something similar with rockets.
Now Musk wants to apply the strategy to semiconductors.
Intel Is Part of the Plan
Tesla and SpaceX are not attempting to learn advanced semiconductor manufacturing entirely alone.
Intel joined the Terafab initiative in April 2026.
Reuters reported that Intel would support the project as the companies pursue the enormous goal of eventually producing one terawatt of computing capacity annually.
Intel’s involvement is particularly important because leading-edge semiconductor fabrication is one of the most technically difficult manufacturing processes humans have created.
A modern processor contains features measured in nanometers. Manufacturing requires enormous cleanrooms, extreme chemical purity, sophisticated plasma processes, advanced deposition equipment, precision metrology, and lithography systems capable of repeatedly producing structures approaching atomic scales.
Tesla’s current job postings provide a glimpse of that ambition. The company is recruiting engineers experienced in sub-7-nanometer semiconductor manufacturing and describes work on 2nm-class dimensions, including advanced Gate-All-Around transistor architectures.
Tesla cannot simply build a huge warehouse, install some machines, and start producing competitive AI chips.
Yield, process control, reliability, and manufacturing economics will determine whether Terafab becomes revolutionary or enormously expensive.
The Project Has Already Changed From Its Original Austin Plan
When Musk first revealed Terafab in March, the project was closely associated with Austin.
Reuters reported that the initial plan involved two advanced chip factories in Austin, near Tesla’s existing Texas operations.
Tesla is still developing a Research Fab at Gigafactory Texas, and the company has advertised multiple Terafab semiconductor engineering positions in Austin. Tesla’s first-quarter 2026 materials described that research operation as the starting point for its semiconductor partnership with SpaceX.
But a full-scale manufacturing campus needs enormous amounts of land, electricity, water, industrial infrastructure, and room for future expansion.
That led to the much larger Grimes County project.
The new site sits northwest of Houston and outside the dense Austin metropolitan area, offering the kind of physical footprint required for a 100-million-square-foot industrial campus.
So the Austin operation appears increasingly positioned as the research and development side of the equation, while Grimes County becomes the potential home of large-scale manufacturing.
Why Does Musk Want One Terawatt of Compute?
Terafab’s most extraordinary target may not be its cost or building size.
It is one terawatt of computing capacity per year.
That figure reflects just how large Musk expects AI hardware demand to become.
Tesla could eventually need chips for millions of vehicles and potentially millions of Optimus robots. SpaceX’s ambitions could involve large-scale orbital AI infrastructure, while xAI continues expanding increasingly powerful training and inference systems.
Each of those businesses consumes computing hardware.
Put them together, and Musk believes traditional semiconductor supply will become one of the biggest physical constraints on growth.
SpaceX’s 2026 prospectus makes that strategy explicit. The company identifies chip manufacturing, data-center infrastructure, and power generation as fundamental physical constraints on AI expansion and argues that controlling more of that stack could create a long-term advantage.
That is why Terafab matters far beyond Tesla cars.
It is an attempt to control the hardware underneath an entire ecosystem of AI products.
Terafab Could Be Transformational—or Extremely Difficult
Musk’s companies have a history of entering industries dominated by established players and challenging assumptions about how quickly hardware can be developed.
SpaceX made reusable orbital rockets practical.
Tesla helped turn electric cars into mainstream performance vehicles.
But semiconductor fabrication presents a different kind of challenge.
The industry has spent decades accumulating specialized knowledge that cannot be purchased simply by constructing a larger factory. Leading-edge fabs require experienced process engineers, highly specialized equipment, extremely stable utilities, sophisticated materials, and enormous capital expenditure.
Even Intel, Samsung, and TSMC routinely spend billions bringing new processes into reliable high-volume production.
Terafab therefore remains an ambitious industrial project, not a guaranteed technological revolution.
SpaceX’s own disclosures also acknowledge that the companies expect to continue buying significant amounts of computing hardware from outside suppliers even as Terafab expands.
The objective is not necessarily to replace TSMC, Samsung, Intel, Nvidia, or memory suppliers overnight.
It is to prevent those suppliers from becoming the ceiling on Musk’s ambitions.
Tesla Is No Longer Just Becoming a Car Company With Robots
Perhaps that is the biggest implication of Terafab.
Tesla once primarily manufactured electric vehicles.
Then it became a battery manufacturer, charging-network operator, energy-storage company, AI developer, autonomous-driving company, and robotics business.
If Terafab succeeds, semiconductor manufacturing joins that list.
SpaceX is following a similar path. A rocket company expanded into satellites and internet service, then began discussing orbital data centers and now wants chips designed specifically for computing in space.
The initial commitment is already enormous: approximately $16.8 billion for the first phase in Grimes County. The long-term plan is vastly larger, with more than 100 million square feet of manufacturing space and a target of one terawatt of annual compute production.
That does not mean every part of Musk’s vision will happen on schedule—or happen at all.
But Terafab has moved well beyond a casual idea.
Tesla is already recruiting semiconductor engineers. Intel has joined the effort. SpaceX and Tesla have identified the large-scale Texas site, and the companies have publicly outlined the first phase of investment.
The question is therefore changing.
It is no longer whether Tesla and SpaceX want to manufacture their own advanced chips.
It is whether they can build semiconductor infrastructure on the same extraordinary scale that Musk once brought to electric cars, rockets, and giant factories.
If they can, the largest building in the world may eventually be filled not with cars or rockets—but with machines manufacturing the tiny pieces of silicon needed to make all of them intelligent.
