This is the entire machine. What do you think, James? It's absolutely huge. In a factory near the city of Eindhoven in the Netherlands, Mark Asink is showing me a machine. a gleaming hulk of pipes and metal the size of a double-decker bus, nicknamed the Beast. The technical name is the EXE 5000. So this is an EUV machine.
This is essentially the most advanced chip making machine in the world. ASINK works for ASML, the Dutch company that supplies machines to the global semiconductor industry. And the beast... represents the absolute cutting edge of chip making technology available today. Here we have the light source. So that's where the EUV light is generated. It has a huge...
Lens optical system and because the lens is getting bigger and bigger the machine is getting bigger and bigger The beast is a lithography machine It makes chips by tracing intricate patterns of transistors, tiny electrical switches, onto pieces of silicon using light. And the scale at which this machine does that is nothing short of a technological miracle. It creates microscopic transistors measuring just eight nanometers.
That's thousands of times smaller than a human cell, or only three or four times bigger than the width of a strand of human DNA. And this means it can fit tens of billions. of these transistors onto a piece of silicon no bigger than your thumbnail, creating the most powerful and advanced chips in the world. Everything is at the edge of what is
It's the immense accuracy of the system. We are measuring with the precision of a few atoms. We're really pushing the limits here, pushing the laws of physics. The result is that if you want to make... the most advanced computer chips in the world today, you need a machine like the Beast. a machine that weighs as much as two airliners, is made of tens of thousands of precision components, took nearly two decades to develop, costs...
almost $400 million to buy, and is only made by one company in the whole world, ASML, here in the Netherlands. Arguably, it's the most complex system. you can buy in series. You can debate about the Large Hadron Collider, which is extremely complex, but this is a box that you can essentially buy, and we believe it's the most complex system ever made. This is Tectonic from the Financial Times. I'm James King.
There's a chip war going on, a battle over who gets to use and make the advanced semiconductors that power our phones, computers, and the artificial intelligence systems of the future. In this episode, the miracle of modern chip manufacturing. How making computer chips became one of the most complex and sophisticated processes in the world. and why that has put chips at the heart of the tech war between the United States and China.
Today's computer chips are the product of more than 60 years of constant innovation. Back in the mid-60s, when semiconductors were in their infancy, Gordon Moore, the co-founder of Intel, noticed a trend. The first chips in the early 60s had featured just a handful of transistors linked together. But every year, engineers were fitting more and more transistors onto pieces of silicon, meaning that every year, the chips were getting more powerful.
Gordon Moore was really our technology leader from the beginning of this company. He was always viewed as the technology guru of the company. Sanjay Natarajan leads technology research at Intel today. 60 years ago at this point, he wrote a paper, a very short four-page paper, and he observed in that paper that what he called number of components on a chip were doubling every year. And that's really the gist of the paper. The observation became known as Moore's Law.
Side point, Gordon Moore never actually liked that. He never liked the law being named after him, but the name stuck. And then he kind of added more data as the years went by. And then he said, it seems to be doubling every two years. And that was kind of the final answer on Moore's Law. The number of transistors on a chip doubles every two years.
Moore couldn't have guessed it at the time, but his observation helped define the entire chip industry for the next 60 years. The whole industry has made it its mission. to keep Moore's Law going, producing chips that get progressively more complex and more powerful. It's probably important to point out that it's not a law in any physical sense or scientific sense. It really was just an observation. And really what that turned into.
is almost marching orders for the entire semiconductor industry to keep that. trend going. Not just Intel, but I have to say the entire industry worldwide has been marching to this drumbeat. And what it has created for the world is nothing short of amazing. The success of Moore's Law meant that every couple of years, the chip industry could deliver the same computing power in smaller and smaller packages. powering a revolution in computing. If you imagine 50 years ago, computers were
enormous. They were in air conditioned room. They were only available to a very small number of users, generally business or government users. And now everyone has that exact supercomputer from 50 years ago. Everyone has one of those in their pocket today. Those computers have gotten thousands of times faster. They're consuming thousands of times less energy.
They've gotten significantly smaller and they've gotten far more affordable. And there's countless examples of the way that technology that Moore's law has guided us to has really revolutionized the world. Broadly speaking, Moore's law has continued to hold to this day. One of the first integrated circuits developed in the early 60s featured four transistors. Today, the thumbnail-sized chip in a high-end smartphone contains tens of billions of them.
So, let me... You've seen the chip. Have you seen the chip? Yes. I'll show you something. Back at ASML in the Netherlands, Mark Asink shows me what the result of decades of Moore's Law looks like. So this is the chip. out of my son's iPhone. Right, okay. This is an extremely complex thing, right? About, yeah, what is it? A fingernail, I would say. My wife's fingernail. Square. Square, yeah.
And so it looks like a very small black plastic thingy, but inside is like an apartment building. It has over 100 layers with tiny structures and the bottom layer holds transistors. switches i would say and this one has about 20 billion switches in this device so it's like 20 billion light switches And it's connecting these transistors such that it defines the functionality of the chip. And those 20 billion transistors would fit on the chip the size of a fingernail? Yes. And in fact...
Gordon Moore predicted that every two years it would basically double. So in two years we would go to 40 billion. In fact, there are already chips out there with far more transistors. For example, Apple's M2 Ultra chip, used to power desktop computers, has 134 billion transistors. And the latest AI chip from Nvidia harnesses the power of more than... 200 billion. But key to that relentless progress has been the ability of the industry to make transistors smaller and smaller.
And that's where companies like ASML, which supply the machines to every high-end chip factory in the world, come in. If we can make a device, machine, tool that can print smaller transistors, you can put more transistors in the same package, right? Smaller means more. And after 60 years of Moore's Law, the scientific and technological challenge of making microscopic transistors even smaller has become immense.
For your information Jos, we just went to the EXE. Oh, you saw the beast? Honestly, it just blew us away. Yeah, we saw the beast. I dare to say that the latest and greatest machine we have built, the high-end air EV machine, is the most complex machine ever built by mankind. More complex than the first moon lander. In an office above ASML's factory floor, I met Jos Benshop, the company's executive vice president for technology. He was working for the Dutch company Philips.
in the mid-1980s, when ASML was spun off as a separate company. initially housed in a few prefab buildings. Well there's a very nice picture of some wooden sheds in front of the Philips building and indeed this is where the ASML employee, just over 30 of them. were housed and there's a big dumpster with junk next to these portacabins. It really looked and felt like a startup at the time.
Advances in ASML's lithography machines have been central to keeping Moore's law going. Lithography, tracing intricate patterns of transistors using light, has been used in chip making for decades. To make transistors on a microscopic scale, you need very short wavelengths of light. And until a few years ago, the most advanced chips... were made using deep ultraviolet light, or DUV.
But around the turn of the millennium, the industry was worried that DUV light wouldn't be able to make the next generation of computer chips. The wavelength of the light was simply too big. to print transistors with enough detail. In fact, it's very simple, basic physics. The smaller the wavelengths, the smaller you can print the feature. So, Ben Shop and his team at ASML began developing lithography machines that used extreme ultraviolet light.
or EUV light, which had an even shorter wavelength so it could print with an even finer resolution. But it turned out that building an EUV machine was incredibly difficult, with several huge scientific and technological obstacles to overcome. The number one, top number one for many, many years was the EUV source to generate light of a wavelength of 13 nanometer, 13.5 nanometer. How do you make EUV light? They settled on blasting tiny droplets of tin with lasers. You shoot 30 micron.
thin droplet, which is the thickness of a human hair, roughly. You shoot him with 50 meters per second into a vacuum vessel and then you have to head it dead on with a high power laser. Then they needed a series of mirrors to control the beam of light. But these mirrors needed to be almost impossibly flat to make sure... the light could be focused with atomic accuracy. To give an idea, you have to make a mirror with the size of a pizza that is shaped with atomic precision.
0.1 nanometer accuracy, which is less than an atom. So to make a macroscopic object with atomic precision was always considered to be near to impossible. After years of delay, we finally cracked it through a combination of hard work, science and luck. It took ASML 17 years of work and more than 6 billion euros in research and development to build the EUV machines.
But the new machines meant that the semiconductor industry could create chips with a higher density of transistors than ever before. Today... All of the most advanced chips in the world are made with an EUV machine. And the only company in the world that can make those machines is ASML. There are no other competitors in the market. I asked Jos Benshop how it could happen that the whole advanced chip-making industry today...
could be entirely dependent on ASML for such a crucial piece of manufacturing technology. We are very focused on one thing. We build these machines. The second thing is our commitment to technology. There's not that many companies that have the stamina to stand behind a decision for two decades while it continues to be delayed. But there's also an element, as a physicist, I have to come back to the economics.
There's been consolidation in the industry for economical reasons, if nothing else. There's the demand for EV machines, but the investments are enormous. So you can have one company investing 10 billion. and enjoy the majority of the market, or you have three companies, each investing 10 billion, competing for the same market. It simply doesn't add up. So it's a little bit winner-takes-all. We came to the finish line first.
So it's a race where there's ongoing consolidation also at our customers. You see the semiconductor makers are also consolidating. The reason to consolidate is economics. Benchop is right that there is a consolidation across the semiconductor industry, particularly when it comes to making the most advanced chips. ASML is the only company in the world that makes EUV machines.
Just a handful of companies know how to use the machines, including Intel in the US, Samsung in Korea, and most importantly, TSMC. which uses ASML machines to make 90% of the world's most advanced chips in Taiwan. This consolidation with the whole of advanced chip-making reliant on just a few crucial companies is a result of the success of Moore's Law. Making advanced chips with billions of microscopic transistors is so difficult, only a few companies in the world can do it.
And this highly consolidated industry is what's being fought over in the chip war. Hello everyone, my name is Jamie Lang and I'm the host of Great Company Podcast and today we are very kindly sponsored by Sage. So as a business owner, which you clearly are, there are tasks that need doing, right? Yeah, 100%.
Well then, you're going to enjoy this advert. Sage's new AI productivity assistant, co-pilot, can help you work faster, analyse how your business is performing and help you make smarter decisions. Yes, Jemima, and it's powered by the latest innovations in Gen AI. My technology is an X, which means all those repetitive tasks I hate can be done so much quicker. More time.
For scrolling on TikTok, obviously. And making TikToks. And making TikToks. Exactly. Where do we visit, Jemima? Visit sage.com forward slash copilot to find out more. The money, time, effort and expertise you now need to make the world's most advanced chips. has created a vast global network of companies, each with a vital role to play in the chip-making process. And this consolidated supply chain is at the heart of the battle.
between the US and China. When you look at the chip industry, there are a series of choke points. Chris Miller is the author of Chip War, the fight for the world's most critical technology. In lithography, for example, there's just one firm that can produce the most advanced lithography system. The same is true for many of the types of tools that are used in chip making, where one or two firms are capable of cutting-edge production. If you want to design an advanced...
chip. There are three firms which you are almost inevitably going to buy chip-making software and intellectual property from. 30 years ago, there used to be a couple dozen chip manufacturers who could produce cutting-edge chips. And today, it's just a handful of firms that are capable of producing the most advanced chips. And of those, the vast majority are produced by just one firm, Taiwan's TSMC.
These choke points are behind a lot of the anxiety that the US in particular has about chips. It makes the industry vulnerable. If you lose a crucial company in the chain the whole supply of chips is disrupted. The fact that more than 90% of advanced chips are made by TSMC in a handful of plants in Taiwan is the main reason that the US government is currently spending tens of billions of dollars to encourage advanced semiconductor factories to set up shop on US soil instead.
But it's also given the US an opportunity to stop its chief technology rival, China, from getting hold of advanced chips. There's a fundamental irony behind all of the progress the chip industry has delivered, which is that it's required the concentration that today makes the United States very nervous. But it's also used this fact, concentration of the industry as well as the link.
lengthening supply chains to limit the types of technologies that can be transferred to China, even for companies that aren't based in the United States. The US has successfully pressured TSMC to stop making high-end chips for Chinese companies, effectively cutting China off from the advanced chip industry. And in order to stop China making its own chips, it's lent on the Dutch government to stop ASML from selling its most advanced lithography machines to Chinese factories.
Combine that with a ban on companies like NVIDIA selling advanced AI chips in the Chinese market, and the US has succeeded in almost completely starving China of the chips it needs. to develop the latest technology. If you can't produce 2024 cutting-edge ships, or at least something very, very close, your ability to develop expensive products or, more importantly, your...
ability to train cutting-edge AI systems is going to be extraordinarily limited. And so the U.S. and its allies are betting that they'll be able to stay meaningfully ahead of China, thanks to Moore's law, and that its restrictions... will therefore limit the ability of China to develop some of the capabilities that U.S. security officials are most worried about.
There is a lot of debate over whether China is overtaking the US as a tech superpower, but access to advanced chips is the one area where the US still has a clear lead. and it may be decisive in who emerges as the leader in technologies like artificial intelligence. China would like to catch up, but the US embargo is just half the problem. The fact... That the industry is relentlessly driven forward by Moore's law makes it seem much more difficult for China to gain ground.
If China threw the state's huge resources at making its own AI chip, for example, By the time they'd achieved this, the rest of the industry would have moved on to even more advanced chips. The cutting edge of chip technology is a constantly moving finish line. In this sense, the US has a particular interest in keeping Moore's Law going. As long as chips keep getting more and more advanced, the US can maintain its chip lead over China.
But the extremes that the industry now has to go to to make more and more advanced chips begs the question, for how long can Moore's Law keep going? For years, researchers have worried that we may be fast approaching some kind of physical limit to the number of transistors you can squeeze onto a chip. himself warned that no exponential change continues forever.
Today, transistors are so small that they are made up of only a few dozen atoms of silicon. If they keep shrinking them, what happens when you reach one atom? And even before you get to that physical barrier, strange quantum effects will start to disrupt the working of chips when they're constructed on such a tiny atomic scale. Does that worry Jos Benshop at ASML? I mean, generally speaking, you've fulfilled Moore's law over a number of years. But how small can you go?
There's always the question, when do I get smaller than an anthem? I'm going to show you something and the listeners can't read it, but I took the... International Technology Roadmap for semiconductors. And I just extrapolated it. I will be 105 by the time you hit an anthem, which is 2065. So I'm very calm. We have tens of... billion transistors on an advanced chip. And in fact, the expectation of the industry is to reach a trillion transistors in a single package by 2030.
In reality, though, the chip industry is already focusing less on the further shrinking of transistors and more. on other ways to increase the power of chips and keep delivering the benefits of Moore's Law. The death of Moore's Law has been predicted for almost as long as it's been around. Sanjay Natarajan at Intel sees his mission as working to continue Moore's Law.
But he says these days that's less about the transistor density of the chip and more about the computing power the chip delivers, however it's made. If we redefine the notion of Moore's Law as continue to deliver energy efficient and cost efficient computing to the world, we see that the future remains bright. We may move away from the number of transistors on a chip.
more towards, for example, advanced packaging, like making a product out of several smaller chips, we call them chiplets, each sort of purpose optimized for its function. That may be one of the next generations of Moore's Law evolution is we move into a very advanced packaging realm and we deliver that Moore's Law value to the end user that way.
So would it be an oversimplification then to say that the next generation of Moore's law would be a shift away from just increasing the number of transistors on a chip to using different methods, such as designing it differently? I do think the number of transistors on a chip are going to continue to increase.
where lithography roadmap continues to have very good options. So I think at the core of Moore's law and miniaturization, we will continue to see improvements. But in a sense, you're absolutely right, James. We're going to also see... Part of the heavy lifting done by other aspects of design, advanced packaging, these are all going to be core elements of that. My visibility as leading technology research for Intel is about 10 years out. And in that 10-year horizon, I see...
more than enough options at our disposal to continue Moore's Law. Should we care if Moore's Law ends? We should care if Moore's Law ends because it implies that we're not going to continue to be able to deliver that kind of innovation, world-changing, you know. health outcomes, lifestyle, lifetime outcomes, human connection outcomes, we're not going to be able to continue to deliver on that that we have been for the past 50 years. As stewards of Moore's Law, we really do...
feel the responsibility and take the responsibility seriously. We're determined to continue and we believe that that's something that is going to change the lives of the 8 billion people on this planet for the better. Chris Miller, the author of Chip War, is confident in the ingenuity of the industry to keep pushing the cutting edge of semiconductors forward. This is particularly true given the chips that will be needed to power the coming AI revolution.
studying the history of the chip industry, I learned that it's never been the case that we know how we'll produce better chips more than a decade out. The time horizon for companies is only five or at most 10 years. And so whether you're standing in 1960 or 1980 or 2000 or the present, you never know how it will be possible to deliver ongoing exponential. increases for more than about a decade into the future.
Which is why throughout the history of the chip industry, the most eminent computer experts and industry leaders have on a regular basis predicted the end of Moore's law in just about a decade. Gordon Moore himself in 2003 said he couldn't envision how Moore's Law could persist for more than a decade. And here we are two decades later, and from my perspective, the rate has barely changed.
And when I look at all of the money going into produce better chips, better designs, better manufacturing processes for the next generation of AI chips, it seems to me foolish to bet against this extraordinary sum of money. and all of these brilliant engineers trying to produce the next generation of advances. The chip industry keeps driving forward in the relentless pursuit of better chips.
This is good news for the future development of technology, but it's also good news for the US too in its efforts to maintain its advantage in chips over China. For as long as the chip industry keeps making advances, And for as long as the US can keep control over who has access to the cutting-edge processes needed to make the latest chips, the US can keep its chip lead. and its position as a global tech superpower. The United States has been...
betting on computing now for 70 years. And so the U.S. has been betting on Moore's Law. Its companies have been betting on Moore's Law and they've been delivering Moore's Law. I think for... economic reasons for technological reasons but also for strategic reasons the u.s is still betting on advances in computing undergirding its economic position as well as its strategic position
In the next episode in this season of Tectonic, I visit the most important choke point in the whole global chip industry, the island at the front line of the chip war. Taiwan. We're providing the chip manufacturing for everybody in the whole. Everybody's cell phone, everybody's TV, everybody's car using the chips from the factory that we are running over here.
In Taiwan, they make more than 90% of the world's most advanced chips. They call the chip industry the holy mountain that protects the nation. But can it protect Taiwan from the near constant threat of a Chinese invasion? It's our most important industry in Taiwan. Some of my foreign friends told me, if not for the semiconductor, most of the foreign companies will not care if...
Taiwan get taken over by China or not. And it is true. Tectonic is presented by me, James King. Our senior producer is Edwin Lane. And our producer is Josh Gabbert-Doyon. Executive producer is Manuela Saragossa. Sound design by Breen Turner and Sam Giovinco. Music by Metaphor Music. Our global head of audio is Cheryl Brumley. And the impressive IONIQ 5N shows how far the brand has progressed. Smart speaker. Search Hyundai. No, Hyundai. No, the car company.
Ah, you mean Hyundai, maker of the award-winning Ioniq 5N and the all-new Insta, their electric city car. If you think you know our cars and how to say our name... Where have you been? Okay, search Hyundai. It's the dawn of a new Hyundai. No home. No address. No address. No bank account. No bank account. No job. No job. No home. No home. No address. No address. No bank account. No bank account. No job. No job. No home. It's time to break the vicious circle.
We're working with charities like Shelter to provide a bank account for people who are homeless.