5:39
We're about to enter one of the most controlled factories on Earth. A place where a single wrong atom can ruin America's most advanced chips. And some of the rooms we're going into are so closely guarded, no media has ever set foot inside. There's like a secret code to unlock the door. Intel has to keep this factory 1,000 times cleaner than a surgical room. So, what if a hair, for example, got into one of the machines? A hair is huge. Today, it's hard to imagine a world without chips. They run chatbots, fighter jets, AI data centers, even this device I'm talking through. But, right now, the US has a big problem.
Even though America designs many of the world's most advanced chips, about 90% of them are made in Taiwan. But, building chip factories is brutally expensive. And inside, just one machine costs $400 million. A single facility needs rows upon rows of them. That's why the US government is investing in Intel to make more chips in America. We went to the factory at the heart of this comeback mission and asked the question at the center of the global chip race. If, for example, something happened with Taiwan and China, how vulnerable do you think the semiconductor supply chain is today?
Semiconductor factories are called fabrication plants, or fabs. And almost every inch of this sprawling space is engineered to protect what's going on inside. A process so sensitive, you can't wear just anything in. So, no lotion, no hairspray, no makeup, no spray deodorant either. So, I don't know how we're going to be smelling at the end of today, but we'll just have to see. All right, let's go. This is really exciting. Before we could even start shooting, our gear needed a total wipe down. Cleanliness is so important. Intel keeps a whole room full of thousand dollar suits to make sure no one gets past this point dirty.
Each little tiny speck can cause a defect, which would destroy the chip. Chris Ott is in charge of making sure that doesn't happen. So, what you're going to want to scrunch up your suit so that the sleeves don't touch the ground here. Just to keep any particles that may be on the floor from your booties off of your hand area. And then we tighten it up. Snap it up front. So, now we're going to put some uh latex gloves over our previous gloves. One more layer of cleanliness. Almost ready to go. Make sure none of my hand particles escape. But, if one rogue particle does, these holes in the floor would suck it out. So, if I were to sneeze, for example, how fast would that exit this room?
It would move very quickly. It'd probably in uh less than uh a minute. There's 12 football fields of clean room space here. And so, keeping that all clean is one of the most expensive things that you'll do. The room I'm sitting in right now has millions of particles floating in every cubic meter of air. But, the room you're about to see us walk into can't have more than eight particles. As soon as you step inside, you realize that even the wrong kind of light can destroy a chip. Some of the materials that we use are very sensitive to white light. Which is why they use yellow lighting throughout their factories. And it totally messes with your sense of reality.
So, I got really excited when we first walked in here because there are so many things that look bright pink. But in reality, we're here in this factory under yellow lighting, everything that is actually red looks very pink to our eyes. So, we're going to show you that. We're going to change our white balance on our cameras to normal and this is what it would actually look like if these lights weren't yellow. So, I'm devastated. Sorry. The whole complicated chip making process starts with something deceptively simple. Yeah, so this is one of the very first steps that it is. So, the wafer is blank. You can see how it's shiny it is up at the top.
The wafer is basically a thin slice of silicon, a material that comes from sand. It's the base layer the chip gets built on. But for the chip to work, this silicon has to be 99.99999999% pure. What would happen if it wasn't? If it was like 98% pure? So, what the contaminants can do is that they can interfere with the electrical connections in the transistors and it'll cause that chip to fail. And we would we would have to throw that away. That might not sound like a big deal until you realize how expensive that mistake can be. You're somewhere in the 50 to 500,000 dollars just for one wafer. Now you think of it's in a whole box of 25 wafers. So, now you're into the millions for
just one box. So, yes, mistakes are very costly. It takes about 3 months and everything going exactly right for this wafer to become a chip. And in all that time, no human will ever touch it. Robots move the wafers along the ceiling to keep them safe. You have some of the largest factories that are making some of the world's smallest features. From start to finish, the wafer will travel hundreds of miles on these tracks as it turns into a chip. It'll move along, it'll stop at a tool, do some processing, and then when that's done, it'll go back up and then it'll move to the next step.
The work has gotten more intricate over the years. Chips these days are barely bigger than a fingernail. That's why there's literally no room for error. You have about 2,000 steps and you need every step to be perfect. A bump or ridge in the silicon can interfere with the patterns they have to make, which is why the wafer visits this machine to get flatter and flatter until it's quite literally one of the flattest things ever made. And so any vibration, you and I walking by, the tool next to it vibrating, can impact the process. And if you look at all of these tools, they actually sit on top of a pedestal.
The whole factory is engineered to barely move. Its foundation uses twice as much concrete as the Burj Khalifa, the world's tallest building. Because if this place shakes, it's not just a few dozen wafers that get destroyed. Tens of thousands of wafers are at risk. Yeah, the automation that you have keeps track of where each lot is and what it needs to do next is a whole art form in itself. When they add a new layer to a wafer, they often have to send it back to machines like this one, which polishes the surface and clears away anything that doesn't belong. The layer they're building next holds one of the most important parts of the chip, the transistor. So what are we actually able to see here?
These are the wires that connect each of the transistors to each other. Just one of these squares contains 10 billion transistors. They are microscopic and work like an electric switch that turns on and off. The closer you pack the transistors together, the faster the current can travel throughout the chip and the less power it uses, which makes the chip more valuable. Uh very much how you would build a city, you know, with very tall skyscrapers. We will put layer upon layer of these skyscrapers up. Uh if you stacked 10,000 of them, one on top of another, it'd be thinner than the sheet of a piece of paper.
There's only one type of machine on Earth that can build structures that small efficiently and only one company that can make it, ASML, headquartered in the Netherlands. They run somewhere in the range of 200 million to upwards of 400 million dollars just for one tool. You're over a billion dollars worth of tools just in this short walk here. Inside this machine, lasers create one of the world's rarest forms of ultraviolet light. That draws tiny patterns into the wafer. And then you need tens of these tools in order to really have an operation that you can scale and create chips at a
high level. It's a very, very expensive game. We had to blur the names on the other equipment Intel uses because that can reveal sensitive details about its process. Today, only a handful of chipmakers have these kinds of tools. And Intel started using them at scale years after its biggest rivals, TSMC and Samsung. Without these tools, a chipmaker needs more steps to make the smallest patterns, slowing production and making their factory less competitive. That's why this race is not just about having tools like these.
It's about how fast you can upgrade them. Every 2 to 3 years, you probably replace 20 30% of the tools. Yeah, you're well into the uh billions of dollars each generation in order to be able to keep the factory at the leading edge. This time, Intel was the first to buy the newest version of this machine, worth about $400 million. That means it can be the first in the business to print even smaller patterns quickly. But here's the catch. That alone won't necessarily put Intel ahead. The secret sauce is how you integrate this machine with all the other thousand machines that are in here and create one process flow. That process helps determine what kind of chip you can make. And you need to make sure it will be the
one everyone wants. Because once you optimize a factory for one type of chip, pivoting is hard. How do you ensure that the process that's going on here is aligned with where the market is going? That's always a tricky one because if you go back say even three four years ago, did we expect the explosion of AI as it was? No. And it and AI is going to have different characteristics that they want than say a cell phone manufacturer. So they want the process to be a little different. And so you do kind of try to look ahead and say, "Hey, how do we want to design this process?" That's the gamble chip makers have faced from the beginning.
The US invented semiconductor chips in the 1950s and for a long time it dominated this industry. Intel, founded in 1968 in California, was at the forefront. It was the first to put the most important parts of a computer onto a single chip and sell it commercially. One result, computers that once filled rooms now occupy cabinet space. For a time, Intel was better at making those chips than almost anyone. It is not every day that we introduce a new microprocessor generation. It just seems like it. By the 1990s, Intel had become the biggest semiconductor manufacturer in the world. But, that was about to change. Because Intel's factories were mostly
optimized for building chips for PCs. This symbol outside means you have the standard inside that an entire library of software has been written to. Behind the scenes, Apple was building what would become the biggest device in decades, the iPhone. But, it needed different chips. In the mid-2000s, Apple gave Intel a shot at building them, but Intel walked away from the deal, betting on PCs instead, right as the market was about to shift away from them. Intel was reportedly hesitant to invest in the massive factory pivot needed to make Apple's chips.
It was splitting its budget between designing chips, which is super expensive, and manufacturing them, too. While many chip companies had already chosen to specialize in one or the other. Most manufacturers were based in Asia, where labor was cheaper. These factories could dump their entire budgets into upgrading tools and pivoting if a chip designer in the US needed them to. So, they ended up with more deals. That's how Taiwan Semiconductor Manufacturing Company, or TSMC, rose to the top. Today, chip companies turn to TSMC more than any other manufacturer, giving it control of about 70% of the foundry market. But, the US government now sees this as a major geopolitical risk. Taiwan governs
itself, but China claims it as its own and has repeatedly threatened to take it by force. If conflict breaks out on the island, the US could lose access to many of its most advanced chips almost overnight. So, whoever controls that supply would hold enormous leverage over the technology that runs the world. There is vulnerability in every part of the supply chain. It could be materials related. It could be equipment manufacturing related. That's why Washington is pushing for more chip making in the US, sometimes deploying tariffs as a threat. It also offered TSMC financial incentives, including tax credits, to build new plants in America. But the company is headquartered in Taiwan. And a lot of its most advanced
manufacturing is still based there. That's why Intel's comeback matters. Intel's invested heavily in improving that supply chain resiliency. It's more local for local. Made in America versus American made is something that we should really think much harder on. So far, Washington has pledged about $11 billion to help Intel build back up, even taking a 10% stake in the company. But even with government help, building a fab in the US costs about 10% more than it does in Taiwan. And running it is about 35% more expensive. And while Intel itself has set aside $100 billion to build plants across America in the coming years, its rival TSMC has gone even bigger, budgeting $165 billion into its US build-out. So, what would you say to The that say
Intel foundry can't compete. I don't see it as a competition with TSMC or Samsung. When we look at the future, the demand is going to continue to grow for semiconductors. And there's going to be a need for continued innovation. When we talk about semiconductors in United States, everyone thinks manufacturing. Because yes, manufacturing brings more jobs. Manufacturing has more scale. Manufacturing builds huge facilities. But the engine behind manufacturing is technology development. We are staying ahead of technology development. With that strategy, Intel hopes it will be ready for whatever innovation is
coming next. And that could attract customers like Apple and Nvidia. That's why they're putting a lot of money and thought into R&D. Okay, so right now we're in Intel's research lab and this place is wild because researchers here are basically looking at the periodic table of elements and trying to figure out which ones will be using in the chips 5-10 years from now. So this is like um well, America's a test kitchen. That's Myunghee Na. She leads the team that's figuring out what the next big chip making process needs to look like. And a lot of that work happens inside these tools.
I'm going to get to put my hands through one of these things. Okay, so why do we need arm thingies? Oh, great question. So a lot of the things that we work on are sensitive to oxygen and water. These materials aren't even used in chips yet. Secret sauce. Yeah. This is your secret sauce. though rudimentary it looks, it's actually our most advanced materials. So the reason why researchers are looking at new materials is because in chip making, it's not just about how the machines work. It's also what you put into them. Every layer of the chip needs its own mix of chemicals and elements, and figuring out what those are is especially important right now, because
the ones the industry relies on today are starting to reach their atomic limits. This is a bare silicon. The Holy Grail in my industry, and everybody is a dream about it. As you go scaling down and down, silicon hits the limit. In other words, silicon got the industry incredibly far. But if chips are going to keep advancing, researchers need something better. Imagination is the limit, and we are looking for new materials to enable the next generation of channel devices. So what are we looking at?
This is the new materials we are talking about. We are really doing the experiment to see whether there is a value proposition of the next generation channel devices. This is our playground for a lot of things, yes. But sometimes, it takes a lot of failed experiments to find the perfect material. Many, many, many millions. I always believe that fail is not failure in R&D. Over in Taiwan, TSMC is doing this research, too. And that's one reason why speed matters so much. If Intel can find the next big materials and start developing the whole manufacturing process around them first, then it could become the foundry for big chip designers, like Nvidia and Apple. But a breakthrough like that also takes the right mix of minds.
Only time you actually see, "Wow, I never thought about it." It's when you actually collect people together, think very different ways, different background, and different cultures, and then actually comes from the different answers to the problem set. I think you cannot underestimate diversity. A lot of that diversity comes from other countries. But now, big US tech companies, including Meta, Amazon, Microsoft, and Google, say it could get harder to bring in that global talent. The Trump administration says it wants to prioritize American workers and has imposed new restrictions on the H-1B visa that allows employers to hire skilled foreign workers. A majority of these visas go to tech-related roles. But now, companies need to dish out a $100,000 fee for
certain new petitions. No more will be big tech companies or other big companies training foreign workers. As a result, some tech companies have been training up Americans. Some of this talent is not readily available and they might not be exactly suited to the semiconductor technology, but we can mold them through the training process that Intel has. We are doing it as an upskilling for talent in US, in America. And inside Intel's labs, keeping talent interested means giving people space to test ideas. We can test a lot of different things in this lab, and then we once we down select, we move a few elements to the fab to test the a little larger scale.
But scaling up takes time. And that's why this work starts long before the technology ever exists. You're looking at about 5 to 10 years down the road devices, yes. How are you making sure that researchers are developing the thing that people will want 5 10 years into the future. That's very good question. So, success rate is not always 100% and that's the risk we take. About 10 20% 25% what is going to make the product. My job is making sure we have enough options. That is where Intel's strategy looks different now. Instead of betting mainly on one kind of chip the way it did 20 years ago, it's trying to cast a wider net.
Finding materials and processes for lots of different chips. So, it can pivot more easily when the next big market shift happens. I think our industry's biggest challenge is going much faster speed than we ever done before. The AI is really changing the world, but suddenly it feels like to me that we got to go speed 10 and [clears throat] that's going to be not just the Intel Foundry's challenge. I think that's the actually everywhere's challenge in this industry. Now, the company seems to be gaining ground. It's stock has gone up more than 400% over the past year, largely because Intel is starting to win big deals with customers
like Tesla, which announced it would use Intel's newest chip manufacturing process. Intel also struck a preliminary deal with Apple to manufacture some of its chips and now it's partnering with Nvidia by producing components that support the AI giants chips. But, Nvidia still relies on Intel's biggest competitor, TSMC, to make its most advanced technology. In fact, Nvidia recently announced plans to design its own PC chip and have it manufactured by TSMC. A direct challenge to the market Intel has dominated for decades. What does Intel need to prove to get Nvidia to commit more fully to Intel's foundry.
We have to execute and we have to be trusted by the customer to ensure we are looking at their needs, how they want to reshape the world, and how Intel can be part of their solutions. Before any advanced chip can work, it has to be assembled with other parts. Assembly is its own highly specialized process that often needs its own plants. That's where we're about to go next. Intel has never allowed the media to visit this part of its Oregon plant in person, but it plays a really important role in how the company turns its chips into usable technology. So, we're getting exclusive access to where Intel basically takes the chips and packages them onto a final product that ultimately goes into a device.
Okay, so Olivia, this is for us to do electrostatic discharge testing for you. Tyler Osborne oversees what goes on beyond these doors. We're here to make sure that you don't accidentally shock any of our wafers. Okay, so what would happen if I shocked a wafer? Uh electrostatic discharge can damage semiconductor devices so that they no longer properly function. Okay, so we're not shocking any wafers today. We're going to be stepping on them.
Yes. All right, but there's there's there's a little And then you're going to grab this handle to validate that you conduct electricity. Oh, wow. And now the door opens so you can go in. unlocks. There's like a secret code to unlock the door. But not a lot of people actually pass through these doors. So, we're in robot territory right now. territory. And they have a special way of announcing themselves. We hear the music, so there's robots around us. So, they have right of way. Robots here Yes. Okay. That's right.
All right, that's good to know. They're smart enough to know if you're here, but it's easier if we just stay out of their space. They are carrying precious cargo. Wafers that made it through thousands of steps without a scratch. But the chips on the wafer can't work until more silicon gets stacked on top. So what we're watching here is actually the heart of the operation. This is where we take individual pieces of silicon and bond them together so that they can communicate and talk with each other to function as a single device.
Now the wafer is ready to be sliced up into individual chips. So the water is there to help remove debris as we're cutting. Getting this far doesn't mean the chips will work. It just means they've made it to the moment of truth. Testing one by one. This tool is actually an acoustic microscope. And what it's doing is using the water to conduct sounds into the wafer. And then actually listen to see what that sound is coming back out of the wafer to see if things are actually assembled properly.
If those sounds come through right, then the chip moves to its next step. What you're seeing with flashing lights is actually cameras taking a picture almost faster than your eye can see it stop. It would look for defects perhaps happened somewhere along the way. Because no matter how much work and money goes into protecting these chips, some of them still won't make it. Those get left behind at each round of testing. This arm skips over chips that are meant for a different kind of product. Thousands and thousands of units run through each tool every single day. Every single day all day 24 by 7. And then whatever is put on that platform, those are perfect and they're moved on.
Yes. Even if the chip looks perfect, we still don't know if it will turn on until it gets tested again. So they're actually turning the chips on right here. Yes. How do you do that? Ah, so they have uh micro probes inside each of these test cells that are very, very fine. Um almost as fine as a human hair. They actually come down and contact the individual bumps on the chip and make electrical contact to turn it on. So, that's the real test. Not just how many chips you start with, but how many actually work in the end. That's what the industry calls yield.
It determines how valuable your factory is and whether customers like Nvidia or Apple trust you to make their chips. For years, yield has been TSMC's major edge over practically every other advanced chip maker. So, I had to ask Naga what Intel plans to do about that. I will never be satisfied with the yield unless we are at 100%. In the past, we have been a little bit non-inclusive when it comes to inviting more of the industry partners to come teach us how to be better. And now we have opened the doors more. The chips you are seeing are built for AI laptops. But Intel says its factories can also make chips for servers and data centers. And that its new manufacturing process can build them to be faster and more power efficient than
much of what the industry can make today. And if that proves to be true long-term, it could be Intel's ticket back to the top. If you look over here, you can actually see several of the die in the spool that are being put into a pocket and then covered with a cover tape and then put into that roll. So, it's being wound up just like old-fashioned movie projector film. So, how many, I don't know, laptops could you make from that one roll? Hundreds or potentially even a few thousand. So, we have a technology engine that's proven. That's not only manufactured in US, but it's actually developed and manufactured in US. If we do our basics right, if we focus on our controllables, I'm sure the
results will follow. But, even as Intel grows stronger, no one company can erase how vulnerable the semiconductor supply chain really is. The entire world and the semiconductor industry requires a broad supply base than what the current industry is able to provide because the demand is very high. And if the world ever had to shift away from Taiwan fast, it still would not be simple. If, for example, TSMC were to go offline completely tomorrow, how set up is Intel's foundry to be able to handle Nvidia's designs, for example?
Intel foundry is ready from the technology that we have and the supply chain that we have. And we will have to work with our customers to ensure they can port over their designs. But, every one of these designs take time for it to move over to a new technology and start ramp into manufacturing. So, it won't be immediate. It will take some time and work to do.
