Inside Elon Musk’s $119 Billion TERAFAB — Shocks China & Detroit!

Kind: captions Language: en 3 miles. That is how far one building in Texas is planned to stretch. It would hold more than one 700 football fields of floor space on the edge of a lake that once cooled a coal power plant. The closest town, Anderson, has fewer than 300 people. Nearly 900 residents signed a petition against it. The man behind the project is Elon Musk. The number in this video's title, 119 billion, comes from county paperwork, and it is a ceiling. The amount the companies have actually put their names to is 16.8 billion. That is still one of the largest industrial bets in Texas history, but it is a very different promise. So, which number is real? And can a team that has never run a chip factory build one this big? Stay for the middle of this video because I want to show you what a SpaceX representative told worried neighbors about water and what that answer left out. And near the end, there is a state deadline on October 19th that could shape what happens next. I put the sources for every number in the description. If you want stories like this checked against filings instead of headlines, subscribe to Auto Pulse. Start with the money because the money tells this story better than any headline. In March of 2026, Musk announced a project called Terapab at a shuttered power plant in Austin. The first price tag in circulation was between 20 billion and 25 billion, depending on who was reporting. Then came May, county filings in Texas put the first phase at about 55 billion and the full build at up to 119 billion. The register covered the same figures that week and set them next to Intel's Arizona expansion, which cost roughly 30 billion. Wall Street had been guessing far lower. An earlier Morgan Stanley estimate for the entire project sat between 34 billion and 45 billion. So in about 2 months, the number jumped from tens of billions to more than 100. Then came August. Tesla and SpaceX confirmed the site and attached an official figure to phase 1,6.8 billion. The state of Texas added a 30 million grant and the companies promised 3,000 jobs. Reports on the job count are not uniform, by the way. Some put the minimum closer to 1,800. The 3,000 figure is the one the state and the companies used. Notice what happened there. The filing said 55 billion for the first phase. The confirmation said 16.8 billion. I could not find anyone from either company explaining that gap in public. Most reporting treats 16.8 billion as the committed money and the larger figures as what the site could grow into. That is how I read it too. There is a second reason for caution. SpaceX's IPO paperwork in May described Terapab as a general framework between the companies with the financial terms and the ownership of the technology still to be negotiated. And according to Sherwood's reading of the filing, the arrangement needs sign off from both boards, Tesla's and SpaceX's. So here is a fair way to hold the title in your head. 119 billion is what it could cost. 16.8 billion is what is on the table. And not every signature that connects the two has been written. But why spend even 16.8 billion on a factory when you could simply buy chips? Musk has an answer and it is not really about cars. His argument fits in one sentence. Both Tesla and SpaceX, he wrote, will need far more chips than current and future global production can supply. Run the numbers behind that sentence. The goal he has given is more than 1 terowatt of computing capacity every year. A terowatt is a trillion watts. It is a measure of electrical power and it is how Musk counts the computing he wants to build. Who gets it? Reporting from the next web puts about 75% with SpaceX for data centers in orbit and about 25% with Tesla for Optimus robots and cyber cab. Musk has called Terraab the most epic chip building effort anywhere and described the vision in a phrase of his own. Sci-fi city is what we're aiming for. He also wants it built fast. In his words, the company has to move at light speed, which is an unusual pace for an industry where a single new factory normally takes years. That split matters. 3/4 of the chips are meant for data centers in space, and that plan depends on SpaceX getting launch costs down with Starship. As the register pointed out, the biggest customer for this factory is a product that has to succeed first. Tesla is not starting from zero on chips either. It signed a 16.5 billion deal with Samsung in 2025 for its AI6 chip. Samsung has even been reported to suggest that Tesla use spare capacity at its plant in Taylor, Texas instead of building its own. So, the question is a fair one. If a Texas chip factory already exists, why pour concrete for another? Scale is the answer. TSMC says its annual capacity passed 17 million wafers in 2025, counted as 12in equivalents. A wafer is a dis of silicon about the size of a dinner plate patterned layer by layer and then cut into chips. The long-term goal reported for Terapab is 1 million wafer starts a month. That is 12 million a year, roughly 70% of what the entire TSMC network can make. Treat that as a rough comparison. The chips differ and so do the plants. And it is a goal, not a schedule. But it shows what Musk is asking you to believe. One site in one Texas county approaching the output of the biggest chip maker on Earth. A goal that size needs somewhere to live. And the place he picked says a lot. To size it properly, the Pentagon holds about 6.6 million square ft of floor space. Terraab's full footprint is planned at up to 100 million. That is roughly 15 pentagons of floor space on one site. Gibbons Creek Reservoir sits in Grimes County near College Station. It was built to cool a coal plant that shut down in 2018. Chip factories burned through huge amounts of power and water. So, an old power plant lake is a logical place to look. As of late September, according to Spectrum News, crews were already clearing land with some homes only about 2 miles from the site. SpaceX's team told residents that construction could begin within months, and local reporting expects limited operations by the end of 2027. According to county judge Joe FA, the plant will use recycled surface water from the reservoir and its own power supply, separate from Ericot, the main Texas grid. Now, the water. At a meeting with residents in August, neighbors asked about the Navasota River, their groundwater, and their electric bills. Riley Trenel, who speaks for the project, answered, "We definitely did not pick this location for groundwater. We picked this location for its ability to store water on the river." He said, "We can't dry up the Navasota River. The amount of water we can take on an annual basis is limited." Listen to those answers closely. They are reassuring. They also contain no number. A limit exists, he says, but in the coverage I read, nobody states what it is. That is the question I would put to them. How many gallons a year and who checks? The worries of the neighbors are practical. Traffic is already building on local roads. One resident, MJ Labor, said it plainly, "It's already very, very busy and it's going to be busier." Another Evelyn Shimchack was more philosophical. I feel like we were going to grow anyway but faster. Whether those worries change anything is another matter. On August 6th, SpaceX told residents face to face that the project was a done deal. Once the state incentive agreement was signed, Trenel said, "We are going to be moving nearly immediately to get going on the project." The petition has nearly 900 names on it. The bulldozers do not appear to be waiting on a vote. Local leaders see the other side. Judge FA called it a generational change for Grimes County. The project's backers say 7.2 million people live within an hour of the site, a deep pool of potential workers. Tom Johnson of the Greater Braz Partnership said the region stands ready to deliver the workforce and infrastructure that projects of this scale require. Then there is the tax break. Commissioners approved it and many reports said the abatement was total. SpaceX's own representatives put it at 78% with about 710 million paid to the county over 35 years according to KBTX. And Jordan Bus, the project's environment and safety lead, framed the stakes far beyond one county. This is something that is strategically vital for the United States and for Western civilization. Quick pause. If this kind of breakdown is useful to you with the money, the engineering, and the fine print behind EVs and the companies that build them, subscribe and turn on notifications. It is free and it is how this channel grows. Back to the plant. Land, water, and a tax deal are the problems money can solve. Making the chips is where plans go to die. Here is the technical plan. Tesla builds a pilot factory at Giga, Texas first. It costs around 3 billion and turns out a few thousand wafers a month, enough to test designs and fix problems. SpaceX would then run the big plant. The manufacturing process comes from Intel, which announced in April that Tesla would use its 14A technology. Now, the catch, 14A is not ready. It is still in development at Intel, a step beyond the process Intel calls 18A. Musk himself said, "By the time Terraab scales up, 14A will probably be fairly mature or ready for prime time, probably fairly mature." That is the foundation under a plan worth up to 119 billion. Both companies are betting that something unfinished will be finished by the time they need it. Then there is yield. Each wafer is patterned and cut into chips, and a speck of dust can ruin one. Yield is the share that works. Raising it takes years of chasing small failures and the best chipmakers guard what they learn. Trend forc's analysis said rivals rely on long accumulated process integration defect databases and experience curves. Starting at the most advanced node, it said would be extremely challenging for Terrafab. It listed two more hurdles. The machines that print the finest chip features called EUV lithography tools come from very few suppliers with long weights and high prices. And the United States, the report said, still lags behind Asia in semiconductor engineering talent, fab construction experience, and supply chain maturity. Because of all that, Trendforce suggested a more realistic weigh-in. Tesla's most likely entry point, it said, is not a cuttingedge wafer at all. It is advanced packaging, the step where finished chips are stacked and connected into one module. Packaging capacity is tight right now and it limits today's AI chips. A packaging line in Texas could ease that bottleneck years before a full factory is humming. TSMC's chairman CCI was blunderter. There are no shortcuts in the foundry business. He said by his math building a fab takes 2 to 3 years and ramping production takes another 1 to two. Of course, Weey runs the company with the most to lose, so weigh his words with that in mind. He also called Intel a formidable competitor that should not be underestimated and noted that both Intel and Tesla remain TSMC customers today. The register also noted a fact that deserves its own sentence. SpaceX, XAI, and Tesla have no previous experience manufacturing semiconductors. Tesla does design its own AI chips, and that counts for something. But designing a chip and building the factory that prints it are different jobs. The other side of the argument is Musk's record. Experts called reusable rockets impractical for years and then SpaceX landed boosters and flew them again. A fair person can look at that and decide never to bet against him. Another fair person can look at a field where experience piles up slowly and decide this bet is about time more than will. So when does this happen? It depends on who you ask. Local reporting says limited operations by the end of 2019. Seven. Texas filings point to the end of 2028 for completion. Tesla itself has talked about silicon around 2029. We's math 3 to 5 years from a construction start points to the later end of that range or past it. I owe you the second half of the title. You may have clicked expecting to hear that China and Detroit are shaken. I looked for a reaction. A statement from a Chinese chipmaker from GM from Ford. I found none. What I did find is a real connection and it is more interesting than a reaction. Start with Detroit. In 2021, Alex Partners estimated that the chip shortage would cost the auto industry 210 billion in revenue that year. The chips at the center of it were mostly older, inexpensive parts, not advanced AI processors. It happened again in October 2025. The Dutch government took control of the chipmaker Nexperia, which makes simple switches and logic chips that cars use in headlight controllers, battery management, and anti-lock brakes. China answered by blocking exports from Nexperia's plant in Donguan, which handles about 70% of its end product capacity. Honda stopped production at its Mexican plant that builds the HRV. GM, Ford, Nissan, Mercedes-Benz, BMW, Renault, Volkswagen, and Volvo all reported production risks or dipped into their stock piles. Nissan set aside 25 billion yen, about 163 million for the fallout. By late November, China eased its export procedures and shipments started to resume. Chipmakers have a name for that category. Anything built on 28 nanometer technology or larger is called a mature node and that is where cars, display panels and sensors live. These chips are not glamorous. They are cheap. They are everywhere and a car cannot leave the factory without them. China's weight in that world is real. Trend force data shows China held 34% of global capacity for mature chips in 2024 behind Taiwan at 43%. and projects China passing Taiwan by 2027. Now put Terapab beside that. As described, it is built for AI chips, the logic, memory, and packaging for robots, cars, and data centers. It is not aimed at the $2 parts that stopped Honda's line. As far as the published plans show, a $16.8 billion plant in Texas would not have changed what happened to Detroit in 2021 or in 2025. So, where is the connection? Exposure. Detroit has learned twice in 5 years that a part costing a few dollars can idle a factory that builds a $50,000 truck. Musk's answer is to own as much of the chain as possible, from chip design to the factory to the rocket. Jordan Bus's line about Western civilization is the same instinct with bigger words. TSMC sees the threat differently. In the short term, Trendforce reported TSMC says Terafab is no meaningful threat to its lead at the most advanced nodes. The risk it names for the medium-term is customer bargaining power. A giant customer with its own factory can negotiate harder even if the factory never matches TSMC's. For Tesla's rivals, my read and this is opinion is that the effect arrives later and more quietly. If Terrafab works, Tesla stops competing with everyone else for the same few factories. If it does not, Tesla stays in the same line as everybody else and the 16.8 billion becomes a very expensive lesson. So, what should you watch? First, October 19th, Governor Abbott paused new data center permits and ordered an audit of what they take from the power grid and the water supply. The report is due that day. Terrafab will have a data center on its site and county leaders argue it falls outside the pause because the plant is a chip manufacturer. Judge FA put it this way. It is a chip manufacturing operation. They will have a data center on their site, but it's only to aid their particular business. Neither the governor's office nor SpaceX responded to KBTX when asked to clarify. Second, the paper between Tesla and SpaceX. A general framework either becomes a binding agreement or it does not. Watch who pays for what and who owns the technology. Third, the pilot fab. A few thousand wafers a month is small, but the yield numbers from it will say more than any announcement if Tesla ever shares them. Fourth, Intel. Everything at Terafab leans on 14A. The day Intel says it is ready for volume production and shows customers beyond Tesla, a big piece of the risk disappears. If the date slips, the whole schedule slips with it. Fifth, the water one number in gallons written down and published. So, can a team with no chip factory experience build the biggest fab in America? This is what the evidence supports. The bulldozers are real. The 16.8 billion is real. The 3,000 jobs and the state grant are real. The 119 billion is a ceiling. And the chips are a promise with a date somewhere between late 2027 and 2029 running on a process that is not finished yet. Nobody can call it a hoax with machines clearing land. And nobody can call it a done deal with 14A still in development. It is a very large bet with the hardest part still ahead. Which way do you think it goes? Does Musk build a fab that rivals TSMC? Or does the first wafer teach him what TSMC's chairman meant? Tell me in the comments. If this video was useful, subscribe to Auto Pulse and turn on notifications. The chips that stop Detroit's factories cost a few dollars. And the company that learned to beat Detroit on cost is BYD. The video on your screen takes apart its $10,000 seagull and shows what a team of Detroit engineers found