Musk’s $119 Billion Chip Factory Has a Giant Circle Inside

Kind: captions Language: en Somebody asked Elon Musk a simple question online, and he replied with six letters: F E L F T W. Free electron laser for the win. It reads like a throwaway joke. It isn't. Those six letters may explain the strangest feature in the blueprints for Terafab, a $119 billion chip factory so large it could swallow the Pentagon 50 times over. At the heart of the layout sits a giant circle. No chip plant anywhere has anything like it. And if Musk's hint is real, that circle could change how the world's most [music] advanced chips are made. Let's be clear about what that money buys. Not a rocket, not a constellation of satellites, not another car plant. One building. A footprint of 100 million square feet built to turn out as many as 1 million wafers of silicon every month. Output on that level would drop Terafab straight into the same arena as the giants of the chip world, the likes of Samsung, TSMC, and Nvidia. So, the obvious question is this: Why would anyone spend that kind of money on a single structure? And what exactly is supposed to sit inside that circle? By the end of this video, you are going to know both answers. You are going to understand what the circle is, why it could rewrite the way a chip factory is laid out, and why even a perfect version of it would not be enough on its own. Because Terafab has three more problems to beat, and one of them cannot be bought at any price. To get there, we have to start with something Musk has been running into across every company he controls. A shortage so deep that the whole planet does not have the factories to fill it. On paper, Musk's companies barely belong in the same sentence. One builds electric cars. One launches rockets. One trains artificial intelligence. One is is thousands of satellites into orbit. They make different things for different customers in completely different markets. But open any one of them up and you hit the same wall. All four live or die on advanced chips and all four want more of them than the world's chip supply can possibly deliver. Start with the car that has no driver. Tesla is building a vehicle called the cyber cab. It has no steering wheel. It has no pedals. It is meant to move through real traffic entirely on its own. Now think about what you do without noticing every time you drive. You spot a dog drifting toward the curb. You sense that a delivery van is about to pull out. You ease off the gas before you have even put the thought into words. A machine gets none of that for free. It has to take in every frame from every camera, label every object in view, guess where each one is heading, and pick a safe move. All inside a few thousandths of a second. That is heavy computing and it has to happen inside every single car. Now look at Optimus, Tesla's humanoid robot. Standing upright on two legs is something you stopped thinking about as a toddler. Carrying a box, pulling open a door, climbing a flight of stairs, it all just happens. For a robot, nothing just happens. Every joint angle has to be worked out. Every shift of weight, every step, every placement of every foot calculated over and over many times a second. The amount of processing behind one simple walk across a room is huge. And once again, each robot carries its own chip. This is where the scale stops making sense. Musk is not thinking in thousands, not for the robots and not for the driverless cars. His numbers run into the millions, even the tens of millions. Should even a slice of American homes end up with an Optimus in the hallway, the number of chips required would be bigger than what the entire smartphone business consumes today. Pile on the giant computing clusters XAI uses to train its models. Pile on the smarter satellites SpaceX has in mind. Stack all of that together and you reach a figure no factory on the planet was ever built to handle. Musk has put a number on it himself. By his estimate, today's entire global output covers roughly 2% of what his companies will one day need in AI chips. Not 90%, not half, 2%. Sit with that for a second. It means this is not a money problem. Musk could have an unlimited budget, phone every chip maker on Earth, and offer to pay whatever they asked, and he still would not get what he needs. The capacity is not sitting somewhere waiting to be purchased. It has never been built. Picture a small bakery that can turn out 200 loaves a day. You can walk in with a briefcase full of cash and ask for 10,000 loaves by tonight. It will not happen. There are not enough ovens. There is not enough flour on the shelves. There are not enough bakers in the back. Your money is perfectly good, but it cannot conjure up equipment that does not exist. The chip industry runs on the same hard limit. Every fab operating today already has its output promised to someone, often years ahead of time. That left Musk with two roads. He could keep lining up at other companies' doors asking for a bigger share of production he knew would never cover his needs, or he could build a factory of his own, one big enough to make chips in the volumes his products actually demand. He took the second road, and that decision has a name, Terafab. It is also not meant to be a one-trick factory. The plan is to keep the whole process together. Computing chips sit alongside memory, advanced packaging, and testing. Every stage on the same site. That is a sharp break from how the industry normally operates, where each of those steps tends to belong to a different company in in location. Terafab would fold that entire chain into one complex. The cyber cabin Optimus are the real reason this is happening. Those two will eat through chips faster than anyone in history has produced them. Without a factory dedicated to feeding them, both would spend years waiting on parts. We spend a lot of time making sure every detail in these videos holds up. If you care about that, too, a like and a subscribe make sure the next one finds you. So, that is the why. Now comes the how. Because if the problem is measured in tens of millions of chips, the building that answers it has to be measured in something almost absurd. This is the part where most people lose the thread. 100 million square feet is a number your brain simply refuses to picture. So, let's climb up to it one step at a time. Begin with Gigafactory Texas. It is already one of the largest buildings anywhere on Earth. Thousands of electric trucks leave its lines every week. Crossing it from the front doors to the far wall on foot is a genuine trek. Terafab is planned at around 10 times that size. If Giga Texas makes you feel tiny, take that feeling and multiply it by 10. Still too abstract? Use the Pentagon. It is so big it was given its own postal code. Visitors can spend hours working their way through its corridors. By floor space, Terafab would be roughly 50 times larger. Set them side by side and one of the most recognizable government buildings in the world would look like a garden shed. Put it in metric terms and you get about 9.3 million square meters. Layout 1,300 soccer fields edge to edge and you are in the right territory. Then, rip up the grass on every one of those fields and fill them with testing labs, printing machines, and clean rooms. Seen from the air, this would not read as a factory. It would read as a city that only does one thing. At that size, the inside of the building has to work differently, too. Imagine someone on shift in the memory section who needs to reach the testing section on the opposite side. Walking there is out of the question. The plans point to an internal transport network. Small electric vehicles running along roads that are entirely indoors. Less factory floor, more town. Split into districts, each devoted to its own stage of making a chip. Those districts would not even share the same climate. The packaging zone might be kept at a set temperature, while the printing hall right beside it runs under entirely different conditions. Crossing from one to the next could feel like changing seasons in the space of a doorway. The building would develop its own rush hours, its own flow of traffic, its own daily rhythm, the way any small city does. And a city needs people. Terafab is expected to create upward of 3,000 jobs. That figure spans construction crews, the people running the machines, staff moving shipments, technicians handling repairs, and security guards. All of those people will need somewhere to eat, somewhere to take a break, and quite possibly somewhere to live close by. This is not only a factory being built, it is a community under a single roof. But the size is not about showing off. It is a plan. Terafab is laid out so that Tesla can grow it in phases. Instead of starting a brand new construction project every time demand jumps, the company can open up the next section and switch on the equipment inside. The first phase may run at only a fraction of full output. The long-term goal stays the same. 1 million wafers a month. Reaching that goal means clean rooms on an enormous scale. Inside a clean room, the air gets scrubbed to an almost unbelievable standard, because one speck of dust landing in the wrong spot can kill a chip. Think of trying to engrave a signature onto a grain of rice while someone shakes sand over your hands. That is why everyone inside is sealed in a full body suit, so not a single flake of skin or loose hair ever reaches the air. Then, there are the printing machines themselves. Forget anything that fits on a desk. Each one is about as big as a school bus, and a single unit runs into the hundreds of millions of dollars. Terafab would need row after row of them, running day and night without a break. Once you picture hundreds of bus-sized machines, everyone carrying a price tag like that, the $119 billion figure stops sounding like fantasy. It starts sounding like arithmetic. This building was never meant to wow visitors. It is meant to grow, section by section, keeping pace as the appetite for AI chips keeps rising. But, however huge the structure gets, its size is just the outer shell of this story. The thing that could actually set Terafab apart is sitting right in the middle, inside that circle. Pull up the plans, and the circle is impossible to miss. It sits at the center of the entire complex. Judging by the scale of everything around it, it may well rank among the costliest and most technically demanding pieces of the whole project, and yet hardly anyone has paid it any attention. Here is where those six letters come back in. One person online put the question to Musk directly. Was there a particle accelerator hiding under that circle? He did not write a long answer. He wrote f e l f t w. Free electron laser for the win. Read that the obvious way, and the circle could be aimed squarely at one of the hardest problems in chip making today. To see why, you need to know how a chip is made in the first place. Chip makers chasing faster, smaller chips use light to print incredibly fine patterns onto silicon wafers. It works a lot like old film photography. Light shines through a pattern and casts that pattern onto a surface below. The The is the scale. The features on a modern chip are so small that your fingertip could hold thousands of them. And the smaller the feature you want to print, the shorter the wavelength of light you need. Today's cutting-edge chips depend on a technique called EUV, which stands for extreme ultraviolet lithography. Current EUV machines produce light with a wavelength of around 13.5 nanometers. Getting there is close to violent. A high-powered laser hits tiny drops of molten tin 50,000 times every second. Each strike turns a droplet into a flash of plasma, which in turn releases exactly the kind of light the process needs. From there, the light travels across a chain of mirrors made to a standard that is almost impossible to grasp. They rank among the flattest surfaces humanity has ever produced. Scale one up until it is as wide as the Earth, and its highest bump would still be shorter than you are. And whenever one of those mirrors wears down, the whole machine has to stop for servicing. In a plant chasing a million wafers every month, each hour of downtime costs real money. It is an extraordinary piece of engineering. It also comes with real problems. It eats huge amounts of electricity. Keeping it running is hard work. And there is always the danger that fragments from all that exploding tin will scratch or dirty the mirrors. And every one of those mirrors carries a price in the hundreds of millions. A free electron laser goes about it in an entirely different way. There is no tin and no explosion. Instead, it pushes electrons until they move at close to light speed, then fires them through a row of magnets known as an undulator. The magnets force the electrons to swerve back and forth, and as they swerve, they release light. That light can be tuned with great precision to whatever wavelength is required. In principle, a strong enough free electron laser could deliver clean EUV light with no molten tin, no flying debris, and no ongoing threat to those priceless mirrors. But, a cleaner light source is only half the idea. The bigger shift is how Terafab would use it. In a normal chip plant, every EUV machine carries its own light source inside it. Picture an apartment block where every single unit has its own small boiler for heat and hot water. It works. It is also costly. It eats up space, and every one of those boilers needs looking after. Now, picture the same block with one large central boiler in the basement, piping heat to every apartment. That is the thinking behind the circle. A single free-electron laser in the middle could generate a huge supply of EUV light and route it through pipes into dozens of machines simultaneously. So, instead of 50 machines that each come with their own light source, Terafab could put one powerful laser at the center and feed everything around it. That one decision would rewrite the design of a chip factory from its foundations up. And that is the answer to the riddle. The circle is not decoration. It is not a bold architectural statement. It is the beating heart of the entire site, the single feature that would make Terafab unlike any other chip plant on Earth. Of course, building a laser that powerful is a huge gamble. And even if it works flawlessly, Terafab still faces a problem with no connection to light at all. It has to do with keeping the place switched on. Say the chips are designed and the light source is solved, there is still a monster waiting. Picture 100 million square feet of floor crammed with enormous machines that never get a night off. A place like that needs electricity that never so much as flickers. A chip factory can ride out plenty of setbacks. A power cut in the middle of production is not one of them. Lose power for just a few seconds and every wafer mid-process can be ruined. Tools can be damaged. The losses start stacking up immediately. Think of a surgeon partway through a delicate operation when the lights and the monitors go dark. A few seconds is all it takes for careful work to turn into a disaster. Inside a fab, every outage is that moment multiplied across thousands of wafers at once. So, how many chips Terafab can make is only half the question. The other half is where the power comes from to keep 119 billion dollars worth of factory awake around the clock. And the reported answer is not the one most people would guess. According to reports, SpaceX is preparing a plan that keeps Terafab independent of the Texas public grid. Rather than plugging in, the idea is to build natural gas power plants right beside the factory, backed up by large battery storage systems. In effect, Terafab would make and manage its own electricity. That can sound strange at first. Musk's world is closely tied to solar energy. Tesla puts solar roofs on homes and sells the Powerwall home battery and Megapack systems for utilities. So, why not blanket Terafab with panels, add batteries, and call it done? The answer is a single word, reliability. Consider what a site like this would pull from the wall. The clean rooms on their own use massive amounts of energy keeping the air clean enough for chip making. The printing machines draw heavily every time their lasers fire. Cooling units, vehicles, lighting, all of it piles onto the total. Terafab could consume as much electricity as a medium-sized city, and that supply has to arrive every second of every day without a single gap. Power engineers have a name for that kind of supply. They call it firm power. It means electricity that keeps flowing no matter the weather, the hour, or the season. Solar is excellent while the sun is up. After dark, it delivers nothing. Running a plant this big purely on solar would mean a battery bank capable of carrying a small city through 12 hours each and every night. The price would be enormous. Natural gas brings something solar cannot offer at this scale, control. With its own plants, Terafab is not stuck waiting for local utilities to expand. It also steps around a fight that keeps getting uglier. Tech companies all over the country are racing to build AI data centers and every one of them pulls the electricity of a small town. Grid capacity is already stretched. If Terafab had to queue for its share, the wait could run to years. With gas plants on site, it can run day and night without asking anyone's permission. The batteries have a different job. A factory does not pull a perfectly steady load. When machines power up or wind down, demand jumps and falls. The batteries soak up those swings and keep the supply smooth. And if the gas plant stumbles, the batteries step in instantly and carry the load until it is back. Look at the full picture and something bigger comes into focus. Terafab is a chip factory, yes, but it doubles as a power station, a field of batteries and an industrial utility rolled together. Winning in AI used to look like a shopping contest for processors. Not anymore. Every processor is a chip. Every chip has to come from a factory and every factory runs on an unbroken stream of electricity. Musk's $119 billion plan comes close to a factory that powers itself. So now there is a building, a light source, and an energy supply. One ingredient is still missing and it is the hardest one of all because it cannot be ordered, shipped, or copied from a set of drawings. A car plant and a chip fab live in different worlds. Tesla is expert at pressing steel, putting vehicles together, and shipping them out at remarkable speed. SpaceX is expert at designing rockets and solving problems on the fly. XAI is expert at training models. Not one of those skills gets you ready for what happens inside an advanced chip factory. Chip making happens at dimensions that start to approach single atoms. A fab's air must beat a hospital operating room for cleanliness. Its chemicals must be pure enough that the faintest trace of contamination can alter a circuit's behavior. Printing is so sensitive that a truck rumbling down a nearby road can ruin a chip. Some fabs sit on giant shock absorbers just to cancel out the smallest tremor coming up through the ground. One stray dust particle landing in the wrong spot gets handled like a genuine emergency. And then there is the figure that makes or breaks a fab. It is called yield. Yield is the percentage of chips that come off the line actually working. When a new fab first switches on, its yield is usually awful. Most of the chips fail and get thrown away. Years of experimenting can pass before yield rises far enough for the plant to turn a profit. Sitting at 50% means throwing away half your expensive silicon and half your machine time with it. To make financial [music] sense, a factory needs to be above 90%. Reaching that level ranks among the hardest feats in modern industry. Imagine a safe with thousands of dials and the door only opens when every one of them is set exactly right at the same moment. Now imagine that every wrong guess costs a fortune and burns weeks of time. That is what pushing up yield feels like. The industry's leaders needed decades to master it. And that knowledge is not written down in some textbook. Instead, it lives in the memories of engineers who stood next to these machines for years, saw what went wrong when a temperature slipped by the smallest amount, and figured out the cure. You cannot spend your way to that. Decades of hard-won experience will not compress into a few months, whatever the budget, which is why Terafab has to recruit people who have already won this battle at the highest level. According to reports, the project is set to be led by a former Intel executive who spent 17 years at the company. That person ran the 18A effort at Intel, which ranks among the most advanced chip-making work anywhere [music] in the world. Intel itself has been making chips for more than 50 years. It has lived through yield collapses, equipment failures, and problems that dragged on for months. That kind of knowledge is not something a consultant can explain to you over a slide deck. It is instinct built up over years. Look at what each piece brings. Tesla adds raw manufacturing muscle and the ability to go huge. SpaceX adds speed, the habit of cracking engineering problems quickly. XAI adds an appetite for chips with no bottom, so whatever comes off the line will always have somewhere to go. And the former Intel team adds what none of the others has ever had. They have taken a cutting-edge chip process and actually gotten it running in volume production. Inside the industry, raising yield is treated as closer to a craft than a science. Chemicals, temperatures, and machine settings all have to agree with the silicon. No manual can teach that. It takes being on the floor, watching the machines, chasing down problems one by one. A fraction of a degree can decide whether a chip works or ends up worthless. 17 years at Intel means 17 years of absorbing exactly those lessons. Hiring a leader who has already been down that road gives Terafab a real chance of skipping the long, painful years most new fabs have to suffer through. And with that, every piece is on the board. The demand is real. A building of 100 million square feet gives it room. The laser in that circle gives it light. Gas plants and batteries give it power. And a veteran from Intel gives it the know-how that turns every other piece into chips that actually work. What Musk is attempting has no precedent. A single site in charge of the whole chain, from the electricity at one end to the AI model at the other, running on a chip it just made. Should it succeed, the ripples will travel far past Tesla and SpaceX. The global chip business could be redrawn with proof that one huge centralized complex can outperform the scattered model the industry relies on today. And remember, it all comes back to six letters typed in reply to a stranger's question, a circle at the center of a blueprint, and a laser that could replace dozens of light sources with one. Strip it all down and Terafab is a $119 billion wager. A wager that AI, robots, and driverless cars are going to demand more computing power than today's factories can produce. The scale is almost too big to picture, but piece by piece, it is starting to come together. If this deep dive was worth your time, hit like, subscribe, and turn on notifications so you never miss the next story. Still hungry for more? A few more videos are waiting on screen now. One last question before you go. Is Elon Musk really capable of building something this vast, or has he finally found a bet too expensive even for him? Drop your answer below along with the part of Terafab that surprised you most.