← All videos

Elon Musk’s $119 Billion Factory That Has TSMC Worried

aTech EN Published Sep 24, 2026 Added 1mo ago 12:21 9 views Open on YouTube ↗

Chapters

Topic clips curated from this video. Click to jump in.

Description

$119 billion. That’s the price tag of Elon Musk’s brand-new factory, announced in March 2026: TeraFab. Also known as the world’s largest semiconductor manufacturing facility.

This gigafactory can produce more chips than NVIDIA, Intel, and AMD… combined.

It could have looked like just another factory. But we’re talking about Elon Musk, the man who has a thing for the letter X and dreams of conquering space.

And that’s exactly where this factory comes in. TeraFab is at the heart of Elon Musk’s ambitious plans for space.

We investigated TeraFab, from its mind-blowing specs and promises of galactic exploration to the many controversies surrounding the project.

This project is simply fascinating.

▪️ ▫️ ▪️ ▫️ ▪️ ▫️ ▪️ ▫️ ▪️ ▫️ ▪️ ▫️ ▪️ ▫️ ▪️

00:00 TeraFab: Elon Musk’s Giant Chip Factory

00:57 TeraFab Aims to Revolutionize Semiconductor Manufacturing

02:23 Why Elon Musk Wants to Make His Own Chips

02:52 Data Centers and Chips Built in Space

03:18 Is TeraFab Really Already Under Construction

Transcript

Read auto-generated transcript (2216 words)

Kind: captions Language: en $119 billion. That's the price tag on the brand new factory Elon Musk announced in March 2026. Terafab, otherwise known as the largest semiconductor fabrication plant on Earth. This Gigafactory can pump out more chips than Nvidia, Intel, and AMD combined. It could have looked like just another factory, but we're talking about Elon Musk here, the man with a thing for the letter X who dreams of conquering space. And it turns out this factory's key to Musk's most ambitious space project yet. We dug into Terafab. Between its off-the-charts specs, its promises of galactic exploration, and the pile of controversy around it, this project's flat-out fascinating. But to understand any of that, you have to first understand what Terafab actually is. Terafab is the factory of every superlative. Within a few years, Musk wants it producing 70% of the output of the world's number one chipmaker, TSMC. The plant covers roughly 10 million square meters. In terms of footprint, that's the size of the Swiss city of Fribourg. Its sheer scale makes it the largest factory [music] ever built in human history. For reference, the biggest semiconductor plant operating today, Fab 18, is 10 times smaller at 10.2 million square feet. At peak, Terafab is meant to produce 1 terawatt of AI compute capacity per year. It's slated for Texas in Grimes County on the site of a former power plant. Musk is targeting 1 million wafers a month, plus 100 to 200 billion chips a year. But that colossal scale is only the tip of the iceberg. This factory has another feature that isn't obvious at first glance. It's the first plant in [music] the world that wants to bring the entire chip making process under a single roof. Right now, making a single chip takes a lineup of hyper specialized players. Each one handles a part of the big three stages: design and architecture, etching in a clean room, and dicing. No manufacturer has ever pulled off all those stages inside one factory. And that's exactly what Musk wants to do. But, that's not even the [music] wildest part of the project. Because on top of producing chips, Terafab also wants to build data centers. And right now, you're wondering, "Where's Musk going to get that much power?" Because this is a factory that's going to draw more than 10 gigawatts. For scale, that's roughly the combined annual consumption of Berlin, Paris, and London. All of it to make chips. Forget the usual playbook, nuclear gas plants, all that. No, Elon is looking higher up. His solution, solar energy. And since Earth doesn't capture enough of it, the head of SpaceX wants to build his data centers directly in space. 80% of the chips Terafab produces will be mounted in space-based data centers orbiting Earth. And if you think that's science fiction or a slick move to pump SpaceX's stock, no, it's already under construction as we're putting this documentary together. Well, it's more of an R&D pilot line going up at Giga Texas in Austin. This prototype fab is a mere $3 billion investment. An R&D pilot line that still represents [music] $3 billion in investment. On top of that, Musk can offset his lack of experience in the field thanks to Intel. The American foundry has joined the project and is handing Terafab its most advanced etching process, Intel 14A. Okay, all this sounds great, but when's factory actually coming out? Because Elon Musk and outlandish announcements that never see daylight, that's a love story. Who remembers the human colonies on Mars by 2025? Or the Hyperloop? Across the 25 minutes of his keynote, he gave zero start date for construction. Listening to all this, one question comes to mind. Why aren't you subscribed to the channel yet? It's the best way to support us and never miss our next videos. More seriously, why is Elon Musk throwing himself headfirst into semiconductor manufacturing? There isn't one reason behind this choice, there are two. Either we build Terafab or we don't have the chips. According to Musk, all the world's foundries combined only produce about 2% of what Tesla and SpaceX need. And he's not exaggerating. His empire's ravenous for AI chips. Whether it's Tesla and his self-driving cars, Optimus and its intelligent robots, Neuralink, SpaceX, Starlink, xAI, every one of his companies has the same problem. They run on AI and therefore on semiconductors. And right now, the market is broken. Only [music] a handful of giants like TSMC and Samsung can actually produce chips, and you have to book years in advance. That puts Musk in a state of technological dependence. And on top of that, these companies aren't scaling fast enough. So, he didn't have a thousand [music] options to pull off his dreams juiced up on raw machine compute. He had built his own chip factory, one capable of running the entire creation process end to end. And when you look at how Musk's empire operates, this isn't the first time he's done something like that. Take Tesla. Where other automakers just design the car and outsource the manufacturing of most components, Tesla made the opposite call. It designs its cars, develops its own software, produces its own batteries and motors and assembles them all in-house. That's what you call vertical integration. It lets you control production costs, cut the time it takes to innovate, and keep total control over your products. It's the key to [music] SpaceX's dominance over its competitors. With Terrafab, Musk removes his [music] last bottleneck and no longer is capped by the capacity of TSMC, Intel, and Samsung. Terrafab lets him control the entire value chain from making the semiconductor all the way to the real-world applications in autonomous driving and his Optimus robots. Right there, we've already got a solid handle on the idea behind the Terrafab factory. Except, this explanation has a blind spot. If his only goal is to build a semiconductor factory, why did he sink the equivalent of Venezuela's GDP into it? And why does he want them in Earth orbit when it's more profitable to do it from the ground? The answer to both questions comes from a pretty surprising place. SpaceX's stock market debut. What do you think SpaceX sells? Drop it in the comments. The answer is in 3 SpaceX sells AI services, exactly [music] like Anthropic and OpenAI. You thought SpaceX sold rockets or space transport? >> [music] >> It still does, but that's not priority. And we're not making this up. Musk said it himself half under his breath in the filing [music] required for SpaceX's IPO. In case you missed it, on June 12th, 2026, SpaceX raised $86 billion and was valued at $1.77 trillion. That's the largest IPO in history. [music] But back to that filing, on page nine, SpaceX reveals the markets it's targeting, its total addressable market. And where you'd expect space travel to be at the top of the list, it's actually the smaller market SpaceX is going after. 370 billion. That's a lot, but it's tiny next to the 26.5 trillion earned for selling AI services. That might seem surprising until you look at Musk's holy trinity. Tesla, SpaceX, Optimus AI. Tesla serves as Terafab's commercial outlet and will absorb a big chunk of the chips the factory produces. SpaceX for its part is the carrier. The V3 of the Starship rocket aims to drop the cost of reaching orbit to $10 a kilogram versus 6,000 for competitors. Optimus AI finally supplies the workforce because [music] humans aren't remotely suited to the conditions in space or on other planets. Pair that with Starlink's network infrastructure and you've got every building block to make humanity the first intergalactic species, and that's Musk himself saying it. So, Terafab [music] isn't just a factory. It's the missing blueprint for Musk's space dreams. It supplies the brain his dreamed of interplanetary empire desperately needed. That's why he says every semiconductor maker out there can only supply 2% of the chips he needs. The project's web page spelled it out plainly. 1 billion Tesla Optimus robots will work to spread our civilization across different galaxies. It's also why his data centers will sit in space to break free of Earth's limited resources, namely the power and water needed to cool these chips. And also because building data [music] centers is getting more and more unpopular. For what it's worth, Google has its own space data center project in the works, Project Sun [music] Catcher. Okay, so all that sounds great, but is it doable? To answer that, [music] you have to reckon with the puzzle this project represents. Earth orbit has two advantages for data centers. Solar power up there is abundant and continuous, unlike on Earth where it vanishes at night or during storms. And there's no need to drink rivers dry to cool the chips, which means you can run them hot for longer. Except both come with problems. First problem, capturing solar energy isn't easy in space. To capture solar energy, you have to keep the panels pointed at the sun. Put that way, it sounds like a detail. It's actually extremely hard. And to make matters worse, that's not the only alignment that matters. You also have to aim the radiators at interstellar vacuum and the antennas at Earth. All at the same time. That demands high compute power, complex parts that can fail at any moment, and stratospheric maintenance costs. Second problem, cooling chips in the vacuum of space is complicated. Space means no atmosphere. So, the only way to cool the data centers is to use thermal radiation. And that creates a massive logistics problem. To understand it, we have to talk about a physical law that governs the ability to cool an object using infrared radiation alone. The Stefan-Boltzmann law. Radiated power depends on three factors. The material's ability to emit radiation, the emitting surface [music] area, the temperature of the object being cooled. In orbit, the only variable humanity can control is the cooling area. So, you have to send up very and I mean very large panels. And that's exactly the problem because shipping costs a fortune. And we haven't even touched the third challenge to space everyone conveniently ignores. Third problem, space is hostile to human-made hardware. The list of threats to the equipment is a mile long. Ionizing rays, ultraviolet rays, atomic oxygen, space debris. All of them are dangerous to the equipment and they cut its lifespan to 5 years on average. Factoring all that in, you have to wonder whether this project will ever be commercially viable. Musk admits as much several times in SpaceX's IPO filing. While we expect to construct Terafab to address such supply constraints, Terafab may not be successful. In which case, we may not have other sources of sufficient AI chips to meet our orbital AI compute demands. Either way, we'll be following the story closely. Subscribe now and turn on the notification bell so you don't miss our next documentaries on the subject. And if you want to dig deeper into Musk's empire, click the image on your screen to go behind the scenes at Tesla.

Related coverage