Terafab 6 Months Later: the Truth About Musk's Mega-Factory

Kind: captions Language: en On September 3rd, 2026 in Austin, Tesla holds the official launch of the Cyber Cab. The car with no steering wheel, no pedals, the one that's supposed to become a self-driving taxi. Four engineers walk on stage. They talk for 15 minutes. Elon Musk isn't there. It isn't a falling out. It isn't a scandal. But for a man who spent 15 years being the show, that absence says something. Musk is looking somewhere else. He's looking 80 miles northwest of Houston at an empty stretch of land where until recently a demolished coal power plant stood. That's where he wants to build the largest building in the world. Back in March on this channel, we told you about an announcement. Seven words on X and a project the CEO of Nvidia called virtually impossible. 6 months have passed. The site [music] is chosen. The money is committed. Intel has signed on and the construction schedule is set. Here's what you're going to understand. Where the Terra Fab project actually stands. What's been confirmed and what has changed. And [music] why there's one obstacle that Musk's billions, the governor of Texas, and Intel combined cannot remove. Everything you're about to discover in this video is the result of hours of research. [music] If you appreciate that work, subscribe to the channel. Your support truly means a lot. Now, let's start with the only number that really matters. Part one. 100 million square ft. That number [music] means nothing on its own. So let's translate it. 100 million square ft is about 9.3 million m of manufacturing space. 9 1/2 km. First benchmark. The largest building in the world today is called the New Century Global Center. It's in Chungdu, China. It contains a water park, a shopping mall, hotels, an Olympic ice rink, and an artificial beach with a giant screen that simulates a sunset. You could fit five of them inside Terraab. Second benchmark for the Americans watching. The Pentagon is the symbol of the outsized government building. Terraab [music] is 15 Pentagons. Third benchmark, an industrial one. Boeing's Everett plant in Washington State is the largest building on Earth by volume. It's where they assemble the 747 and the 777. Terraab is 23 Boeing Everett plants and one last one for human scale. Terraab would employ at least 3,000 people. Divide the floor space by the headcount. That's roughly 3,000 square meters per person. Statistically, every employee would have half a football field to themselves. This is obviously not a factory built for humans. It's a factory built for machines supervised by humans. Musk summed it up his own way on X. Terrafab, Texas, will be the largest and most valuable building on Earth by far. And it will be stunningly beautiful, right? A 9 and 1/2 square kilometer building. That's stunningly beautiful. We'll see. Part two, what was actually signed. Now, the official numbers, and they're more modest than what you'll read almost everywhere. On August 6th, 2026, SpaceX and Tesla confirmed [music] the site, Grimes County, Texas, about 15 mi east of College Station. The announced investment for the first phase is $16.8 billion, not 20, not 120. 16.8 for phase 1. At least 3,000 jobs. Construction starting December 1st, 2026, targeting completion by the end of 2028. The tax arrangement deserves a second because it's been reported wrong almost everywhere. You'll read that the county granted a full exemption. In reality, SpaceX's attorney, Buck Brandon, said it himself in a public session. The effective abatement works out to roughly 78%, not 100. In exchange, SpaceX pays $10 million upfront, then 20 million a year for 35 years. 710 million total to the county and the location writes itself. Terrafab will be built on the site of the demolished Gibbons Creek coal power plant. The water will come from the nearby reservoir rather than local groundwater. That's an explicit SpaceX commitment. As for electricity, it will come mainly from natural gas. Sit with that for a second. The largest AI chip factory ever designed is going to be built on the ashes of a coal plant and powered by fossil fuel. That isn't a moral contradiction. It's a physical constraint. There is no renewable source in Texas capable of delivering that kind of continuous power by 2028. But it deserved to be said. So much for the ground. Now the real question. Does anyone actually know how to do this? Part three. Intel steps in. Because building cars and rockets is one thing. Manufacturing leading edge chips is another. And neither Tesla nor SpaceX has ever etched a single transistor. Which brings us to the partner and this may be the most important development of the past 6 months. In April 2026, Intel joined the project. Terrafab will use Intel's 14A process, its most advanced node, the successor to 18A. On the reported figures, it delivers 15 to 20% more performance and 30% more transistor density. You have to measure what that contract means for Intel. For a year, the company had been saying it might abandon 14A if it couldn't find a major external customer. Intel's CEO said he could think of no better partner than Elon Musk to explore unconventional ways to improve chip manufacturing. It's the largest external foundry contract in Intel's history. The stock rose 3.6% after the announcement. Tesla also hired Gary Jong, 17 years at Intel, a specialist in starting up multi-billion dollar fabs. a Tingua graduate. He probably won't run Terrafab as a whole, but he's exactly the profile you hire when you're about to run an intel process inside your own walls. And the plant would produce [music] two families of chips. One for Earth, the brains of the Cyber Cab, and the Optimus robots, one for space, radiation hardened chips for SpaceX's orbital data centers. On paper, it all holds together. The money, the land, the process, the people. Except one thing is missing, and that thing cannot be bought. Part four, the machine that doesn't exist. To etch chips at that level of precision, you need extreme ultraviolet lithography machines. One company on Earth knows how to build them. ASML in the Netherlands. We devoted an entire video to it, and the summary is one sentence. Without ASML, nobody etches anything. Now, the numbers. Adrienne Sanchez, an analyst at Yol Group, told EEE Times, "ASML shipped 48 EUV systems worldwide in 2025. Its production capacity runs a little over 50 a year with long-term projections slightly above 100, and its order book is fully allocated to TSMC, Samsung, and Intel through 2027. Terraab, by most estimates, would need more than 300. Do the math." One analysis did exactly that. Equipping Terapabs Logic Fabs alone would take ASML roughly 70 years at its current output rate. 70 years for a building announced for the end of 2028. This isn't a money problem. Musk can write a check for any amount. The machines don't exist. It's a limit of global physical capacity and it doesn't depend on him. So what do you do when the machine you need doesn't exist in sufficient quantity? You try to build a different one. Part five, the secret bet and the market. This is where the project gets [music] genuinely interesting and it's the part almost nobody talks about. Terrafab is reportedly exploring an alternative light source technology, the free electron laser. Here's the principle. Simplified. An ASML machine generates its light by firing 50,000 tin droplets a second and hitting them with a laser. Every machine carries its own light source, which is what makes it enormous, expensive, and fragile. A free electron laser works differently. You accelerate a beam of electrons. You make it oscillate through a magnetic field and it emits light. A single central installation could feed several etching machines at once. If it works, [music] everything changes. It routes around the monopoly. It cuts energy consumption. And it would explain why Musk is building at this scale. Because a central particle accelerator feeding dozens of scanners needs room. If it doesn't work, Terrafab is as dependent on ASML as everyone else. And the timeline collapses. Nobody knows. It's a technology bet, not a purchase order. There's one last thing that changed since March, and it alters the nature of the project itself. SpaceX went public in June 2026. Until now, Musk funded his bets with private money with no quarterly reckoning. That's over. On the Terafab announcement, SpaceX stock gained 6% to $115, which is 17% below its IPO price of $135. Terafab isn't just an industrial bet anymore. It's a line in a quarterly report read by shareholders who can sell. And the analysts aren't gentle. Barclays warned back in April that the excitement around Terraab was masking a capital burden layered on top of a Tesla budget already exceeding $20 billion for 2026. And Bernstein put a number on the full-scale project. Their range between 5 trillion and 13 trillion. For scale, 13 trillion is roughly half the annual GDP of the United States. One note, this script was finished on September 26th, 2026. If something has broken since then, you know more than I do, and I'd like to hear it. Conclusion: two readings. So, 6 months later, where does this actually stand? What's confirmed is solid. The site is chosen, and the agreements are signed. Phase 1 money is committed. Intel brought its most advanced process and its know-how. A manufacturing veteran has been hired. Construction starts in December. What isn't confirmed is just as substantial. No confirmed ASML orders. No definitive licensing agreement made public with Intel for now. It's a framework arrangement, not a manufacturing contract and a breakthrough technology that has never [music] worked at industrial scale. So there are two readings and both stand up. The first, it's the Tesla playbook run again. When batteries were the bottleneck, Tesla partnered with Panasonic, absorbed the knowhow, then built its own cell. When the Model 3 nearly sank the company, Musk iterated under pressure until he had one of the most efficient plants in the world. Find the bottleneck partner, integrate, scale. He's done it twice. Intel is playing the Panasonic role here. The second, this time, the bottleneck isn't internal. It isn't about Will or Tesla engineering. It's a Dutch company that produces 50 machines a year with a full order book. No amount of determination builds an EUV machine faster. And there's one thing 6 months has made visible. In March, this was a project. In September, there's land, a permit, a construction date, a partner, and a FAB director on payroll. It isn't a slide deck anymore, but there's also that cyber cab launch on September 3rd where four engineers spoke for 15 minutes to an invitationonly room without him. Musk launched the most outsized project of his career, and he wasn't on stage to present [music] the product that project is supposed to feed. You can read that as a man who has finally learned to delegate. You can also read it as a man whose attention has already moved to the next construction site. So tell me in the comments, will Terraab get built or are we looking at the most beautiful slide deck in industrial history? I read every message and they decide what I dig into next. And if you liked this one, there's another waiting on your screen. But if you really want to know who ends up controlling this technology, please subscribe. See you in the next video.