What Is Terafab? Elon Musk's Plan to Build the World's Biggest Chip Factory

Kind: captions Language: en March 21st, 2026. Austin, Texas. An old power plant that once lit up an entire city stands silent in the dark. Then, light beams explode into the Texas sky. Inside, a crowd of engineers, investors, and government officials wait. The governor of Texas is in the audience. Cameras from around the world are watching a live stream. Elon Musk walks onto the stage. What he announces over the next hour will either go down in history as the greatest industrial achievement of the 21st century or the most spectacular overreach of a man who has already defied the odds more times than anyone thought possible. He calls it terraab and he says it will change everything. The chip problem nobody is talking about. To understand why Terrafab exists, you need to understand one uncomfortable truth about the modern world. Everything runs on chips. your phone, your car, your television, the traffic lights outside your window, every hospital monitor, every military system, every streaming platform, every AI chatbot you have ever used. All of it depends on tiny silicone wafers packed with billions of microscopic transistors. Each one smaller than a strand of human DNA. And the world is running out of room to make more of them fast enough. Right now, the global chip industry is dominated by a handful of companies. TSMC in Taiwan, Samsung in South Korea, Intel in the United States. These companies have spent decades and hundreds of billions of dollars building the infrastructure to manufacture the most advanced chips on Earth. The problem is even they have limits. At Tesla's Q4 2025 earnings call, Musk warned investors that external chip capacity from TSMC, Samsung, and Micron would hit a ceiling within 3 to 4 years. He was not being dramatic. He was doing math. Think about what Musk is building right now across his companies. Tesla is making electric cars that drive themselves. SpaceX is launching hundreds of satellites every year. XAI is building artificial intelligence systems that compete with the best in the world. And then there's Optimus, the humanoid robot that Tesla is preparing to manufacture at a scale that has never been attempted in human history. Gigate Texas alone is expected to have capacity for 10 million humanoid robots per year, which would require 20 million chips, approximately six times Tesla's current chip demand across its entire auto business. And if Tesla hits its stated long-term target of 100 million Optimus robots annually, it would require more than 200 million chips, over 50 times its current demand across automotive and robo taxi combined. Read that again. 50 times. The chip suppliers that exist today cannot come close to that number. Nobody's can. So Musk made a decision that only Elon Musk would make. If the world cannot supply the chips, build a factory that can supply the world. What terraab actually is? Terraab is a planned semiconductor fabrication facility jointly developed by Tesla, SpaceX, and XAI designed to produce more than 1 terowatt of artificial intelligence computing capacity per year. 1 terowatt, 1 trillion watts of computing power every single year. That is approximately double the amount of power currently generated across the entire United States per year. The word terra in terapab is chosen not casually. It is the target. It is the mission statement and it is a number so large that most experts in the semiconductor industry say it borders on fantasy. Terrafab will consolidate every stage of semiconductor production under one roof, including chip design, fabrication, memory production, and packaging. It will be built on the north campus of Giga Texas in a building planned to dwarf Giga Texas itself, which is already one of the biggest buildings on Earth. To understand how radical this is, consider what it normally takes to build chips at this level. Chipm is not one process. It is hundreds of processes happening in a precise sequence, each requiring different machines, different chemicals, different levels of cleanliness that make a hospital operating room look dirty by comparison. Normally, these processes are spread across multiple facilities, multiple countries, multiple companies working in a carefully coordinated supply chain built over decades. Terrafab wants to do all of it in one building from scratch. The project targets 2 nanometer process technology, the most advanced in the world, with an initial output of 100,000 wafer starts per month. The long-term goal is 1 million wafer starts per month, producing between 100 and 200 billion custom AI and memory chips per year. Two chips to rule them both. Inside Terraab, only two chip designs will be made, but those two chips are designed to power two very different frontiers of human ambition. The first chip is for Earth. It is a terrestrial inference chip for Tesla's full self-driving system, its CyberCab robo taxi program, and the Optimus humanoid robot line. Tesla's fifth generation AI chip, the AI5, is among the first products the facility is designed to produce with small batch production targeted for late 2026 and volume production projected for 2027. Every self-driving Tesla on the road, every robo taxi picking up passengers in cities, every humanoid robot loading shelves in a warehouse or working on a factory floor. All of them will run on this chip. designed in-house, built in-house, controlled entirely by Musk's companies from the first line of code to the final packaged product. The second chip is for space. The D3 is a high-powered radiation hardened processor built for operation in space. Temperature swings hundreds of degrees between sunlight and shadow. Cosmic rays constantly bombard any electronics that dare to operate beyond the protection of Earth's atmosphere. Normal chips fail within days. The D3 is built to survive it all and still run artificial intelligence workloads that would challenge the most powerful data centers on the ground. And here is where the vision gets truly extraordinary. Musk said 80% of Terrafab's compute output would be directed towards space. His argument is that solar energy in space is roughly five times stronger than Earth's surface and managing heat in the vacuum of space makes it easier to cool large computing systems. The vision bets that within 2 to 3 years, running AI workloads in orbit will be cheaper than doing so on the ground. The satellites powered by constant solar energy would effectively turn low Earth orbit into the world's largest data center. SpaceX has already filed to launch 1 million data center satellites into low Earth orbit. Not 10, not 100, 1 million. And the D3 chip is what will power every single one of them. The wall of doubt. Now, let us be honest about something. This is the part of the story where we have to ask the hard questions because Elon Musk has a complicated relationship with timelines. He is a man who genuinely changes the world, but rarely on the schedule he announces. Musk has no background in semiconductor manufacturing. TSMC spent $165 billion over years to build six fabs in Arizona, and those will not even reach 2nm production until 2029. A single 2nm fab with 50,000 wafer starts per month costs roughly 28 billion, and it takes about 38 months just to build in the United States. Morgan Stanley projected that even under an aggressive buildout, initial chip output from Terapab would not occur until mid 2028 at the earliest. And then there is the history. In 2020, Musk stood on a stage and promised a revolution in battery manufacturing. The 4680 battery cell was going to cut costs in half. Tesla was going to ramp production to 10 gatt hours within a year. 5 and 1/2 years later, the program has fallen far short of every target he set. The Dojo supercomputing team, meant to train Tesla's autonomous driving AI at massive scale, was quietly dissolved in 2025 after failing to meet expectations. Tesla's auto business itself is under pressure with sales declining for the second consecutive year in 2025, including a significant drop in Europe and its first ever annual decline in China. Critics look at all of this and see a pattern. A man who announces the future loudly, then quietly delivers a smaller, later version of it while moving on to the next announcement. These are fair criticisms, and anyone investing money or placing bets based on Musk's stated timelines should keep them in mind. Why the doubters might be wrong. But here's the other side of that story. Every single time Elon Musk was told something was impossible, there was a long list of serious, credentialed, experienced people making that argument. When he said SpaceX would build reusable rockets that could land themselves back on a pad, aerospace engineers laughed. The physics seemed to work against it. The cost seemed prohibitive. The timeline seemed absurd. SpaceX now does it routinely. When he said Tesla would build electric cars that could compete with luxury vehicles and eventually outsell combustion engines, the automotive industry dismissed him. Electric vehicles were a niche market for enthusiasts, they said. The range was too limited. The charging network did not exist. The consumer appetite was not there. Tesla is now one of the most valuable car companies in history. The pattern with Musk is not that he's wrong about the destination. It is that the journey almost always takes longer and costs more than he says. But the destination, he has a remarkable track record of eventually arriving there. And Terrafab, stripped of its dramatic presentation, is solving a real problem. The world genuinely does not have enough chips for the AI future being built right now. Not just by Musk's companies, but by every major technology player on Earth. The combined output of every advanced semiconductor foundry currently operating on Earth represents roughly 2% of the compute Musk says his companies will need. Even if that number is inflated, even if you cut it by 90%, the gap between chip supply and AI demand is very real and growing every year. Terrafab is one man's answer to that gap. An answer that's audacious, expensive, technically brutal, and years away from full realization. But so was every other thing he has built. The bigger picture. Step back from the headlines and the skepticism and the laser light show in Austin for a moment. What Terrafab actually represents is something deeper than one factory or one man's ambition. This project threads together every major Musk initiative of the past 2 years. the SpaceX and XAI merger, Tesla's solar manufacturing push, the Optimus humanoid robot program, and Starship's development. For the first time, there is a single piece of infrastructure that ties all of it together. Cars that drive themselves, robots that work alongside humans, satellites that blanket the planet in intelligence, rockets that carry civilization to other worlds. They all need chips. And Terapab is the factory that Musk believes will feed every single one of those ambitions. Whether it delivers on schedule or not, whether it reaches 1 terowatt or onetenth of that, the direction it points towards is the direction the entire world is moving. The next era of human technology will be defined by artificial intelligence, robotics, and space infrastructure. and the country, company, or person who controls the chips that power those systems will hold more leverage over the future than anyone holding oil fields or shipping lanes today. Elon Musk just announced he wants to be that person. And if history has taught us anything, the smart bet is to not assume he fails. That's Terraab. That's the story. And this is just the beginning. If you want to stay ahead of where technology is actually going, not where the headlines say it's going, but where the infrastructure is quietly being built right now, subscribe to Pi Uncut because we're going to keep following this story until we find out exactly what Musk is really building out there in the Texas sky.