How ASML, TSMC And Intel Dominate The Chip Market | CNBC Marathon

How ASML, TSMC And Intel Dominate The Chip Market | CNBC Marathon

Show Video

ASML has a monopoly on the fabrication of EUV lithography machines, the most advanced type of lithography equipment that's needed to make every single advanced processor chip that we use today. And this company is one of the most extraordinary organizations in the world. We are the only provider on the planet of this critical technology. The most advanced technology manufactured in the United States. This vision that started 50 years ago with bringing the digital world onto a chip.

And that pace of technology innovation is unstoppable. At the center of this big factory in the Netherlands, in the midst of a months long assembly process, there's a revolutionary machine that the whole world has come to rely on. You could see an EUV machine right behind me.

The size of a city bus but working with atomic level precision, these EUV lithography machines are the most expensive step in making every advanced microchip that powers the modern digital age. Data centers, cars, and every single iPhone. We are the only provider on the planet of this critical technology. These machines are the only way to print minuscule designs on these chips.

They cost up to $200 million, and they're only made by a single company, Advanced Semiconductor Materials Lithography, or ASML. Today, ASML has a monopoly on the fabrication of EUV lithography machines, the most advanced type of lithography equipment that's needed to make every single advanced processor chip that we use today. And this company is one of the most extraordinary organizations in the world. The machines that they produce, each one of them is among the most complicated devices ever made. In the midst of a chip shortage that's caused back orders of everything from PS5s to Teslas, the need for a ASML has never been higher. Its stock has skyrocketed since 2018, while its three main customers, chipmakers TSMC, Intel and Samsung vie to be front of line for ASML's next breakthrough technology. The price tag for this next machine, which

promises to push the boundaries of known physics, is more than $300 million. It's so expensive that most companies cannot afford it. While the chip wars rage on, we wanted to find out what's really going on inside the quiet company making the machines that print them all.

This is the optical part of the machine that makes EUV possible. We got a rare tour inside ASML's clean rooms in California and the Netherlands to see how these machines use precision lasers, exploding molten tin and the smoothest surface in the world to bring our digital age to life. ASML's crucial role on the chipmaking stage has brought it wild success over the past few years, making it even more valuable today than Intel, one of the biggest chip makers it supplies. It's double digit growth every year. And we're not a startup.

You know, we have now 32,000 people. Peter Wennink has been CEO since 2013, but he joined ASML back in 1999, just 15 years after its humble beginnings. It started as a subsidiary of Dutch electronics giant Philips in 1984, conducting research out of a leaky shed next to a Philips office building in Eindhoven in the Netherlands. They were in financial dire straits, so we had no money.

We were poor. And because the problems Philips had were so big, nobody looked at this little outfit out there that was trying to do something crazy, so they neglected us. Still, in its first year, the company successfully launched a first-of-its-kind machine that used precise rays of light to print tiny designs on silicon to make microchips, a technology known as lithography. The first lithography tool really looked like a projector. There is basically a reticle which all the image that you want to project. Then there is an optical system which is going to take this image and project it on the wafer.

Semiconductor lithography was invented in a U.S. military lab and for a long time up through the 1980s, the key lithography firms were American, based in New England. Chris Miller of Tufts University is writing a book called The Chip War: The Fight for the World's Most Critical Technology.

When the industry was getting ready to jump into the early stages of EUV research, none of the U.S. firms were ready to take the plunge on what would be an expensive and risky proposition, whereas ASML was. By 1988, ASML had five U.S. offices with 84 employees and a new Dutch office that eventually became its headquarters in Veldhoven, where CNBC took a tour earlier this month.

We're walking through the EUV factory, which is about 50,000 square meters of space with 1500 employees who are working in shifts seven by 24 to produce 100% of the EUV machines shipped worldwide from this facility. With a breakthrough machine, ASML started turning a profit and went public on the Amsterdam and New York Stock Exchange in 1995. By the 2000s, ASML was acquiring California tech companies like Silicon Valley Group and various key suppliers like Cymer in San Diego, where we also got an inside look at the cleanroom where ASML's light source is produced. So this is actually a nozzle manufacturing area where we actually build the nozzles.

This is actually the piece where the tin shoots out of, that's what's going to create your EUV. EUV refers to extreme ultraviolet, an incredibly short wavelength of light that ASML uses to print smaller, more complex chips. But developing this revolutionary technology was incredibly expensive. We didn't have the money, so we went out and we found partners, which actually was the basis of the way we built the company. So we were forced to be a system architect and a system integrator.

In 2012, ASML offered about a quarter of its shares to its biggest three customers: Intel, Samsung and Taiwan Semiconductor Manufacturing Co., or TSMC. They had to accelerate the R&D for EUV, and the only way they could do this is to get their largest customers involved. And one way you can make your commitment real is to make them a shareholder. ASML is a Dutch company, but it's also a Dutch company that relies very heavily on U.S.

components, in particular for its machines, and at this point relies also very heavily on one Taiwanese customer for its sales. TSMC made up nearly 40% of ASML's sales last year. In 2019, the Taiwanese chipmaker was the first to deliver high volume chips made with EUV, a milestone that's kept it at the head of the pack ever since, its chip technology at least one node ahead of Samsung and Intel. And it has been TSMC's customers that have gained a lot of benefit like AMD, Nvidia and others. And yeah, you can argue that this has come at the expense of Intel not executing. Intel is just now producing its first chips with EUV this year, three years behind TSMC, but it's made a bold move in hopes of catching up: an early investment to secure the first prototype of ASML's next machine: High Numerical Aperture.

To understand why the success of a giant like Intel hinges on ASML, let's take a look at how EUV lithography revolutionized chipmaking. When you start breaking down, what does it take to make an EUV lithography machine? It's sort of Nobel Prize winning in terms of the engineering involved. Chips are made from silicon, an abundant element found in rocks and sand that's purified, melted down, then sliced into circular wafers, the surface on which chips are built in a grid formation. Each wafer can have dozens of thin layers, making up billions of transistors that determine what the chips can do. These layers are printed using

lithography. Extremely precise rays of light are projected through a mask of the chip design. When the light hits the surface of the wafers, which have been coated with photoresist chemicals, it prints the minuscule designs on each layer at extremely high volumes. If you think of a typical processor chip, in an iPhone, for example, will have over 10 billion transistors on a chip, and Apple will sell 100 million or more iPhones for each model that's rolled out.

So you're already talking in numbers that are far bigger than you or I remember how to pronounce. As the wavelength of the light source in making chips gets narrower and narrower, it gives us the ability to make chips with smaller features, which means the chip is faster, the chip can be smaller, the power consumption of the chip can be lower. The smallest transistors are more than 10,000 times thinner than a human hair.

The designs have gotten so small, ASML had to develop new methods of printing at the very edge of known physics. With the help of customer investments and a consortium of scientists, ASML figured out a way to create large amounts of extreme ultraviolet light with a wavelength so short it's not only invisible to the human eye, it's absorbed by all natural substances, even air. So the entire process has to happen in a vacuum, a first for lithography. At 13.5 nanometers, ASML's EUV wavelength is the size

of just five DNA strands laid side by side. The previous generation machines used deep ultraviolet light or DUV, with a wavelength of 193 nanometers. The vast majority of ASML's business, 268 of the 309 machines sold in 2021, still use DUV technology, which is used to print the less advanced chips which are in shortest supply. DUV is for anything that is low technology, like a toaster or refrigerator or even some of the electronics in your car. Today's iPhone 13 is EUV. Both DUV and EUV lithography is so advanced it requires precision down to the atom. This is an EUV cabin of our clean room, which is 10,000 times cleaner than the outside air.

We're wearing these clothing not to protect ourselves from the environment, but we're protecting the machine from the contamination that's created by us. This tiny thread may look like the strand of a spiderweb, but it's actually molten tin being shot out at a pressure of 4000 PSI. And it's how the EUV light is created. This is continuous tin. It never, ever, ever stops.

The tin is streaming through a perfectly calibrated nozzle which we saw being built in San Diego at a rate of 50,000 droplets per second. A 30 kilowatt carbon dioxide laser hits each droplet twice per second, vaporizing them into plasma. These tiny explosions are what emit photons of EUV light. A huge number of explosions need to happen because only about 5% of the photons reach the actual wafer. The light particles are so short they get absorbed by mirrors, the typical method used to precisely aim light through a lens. So ASML partnered with German optics company Zeiss, which makes the flattest surface in the world.

The flatness is really just incredible. If you took a mirror element that is maybe this big and you blew it up to the size of the country that we're in. The biggest bump would only be about one millimeter across the entire surface of a mirror the size of this country.

EUV light bounces off these groundbreaking Zeiss mirrors until it hits photoresist chemicals on the surface of the silicon wafer to print minuscule designs that make up the chips. The aim needs to be so precise TSMC says it's equivalent to shining a laser from the moon to hit a coin on the earth. So your tin is inside of reservoir here and then you're firing out this way. Pete Mayol has been running this clean room for six years. If any kind of defect particle whatsoever is even on the tip of that capillary, it's a fail.

We'll move and start all over again. And the speed and scale at which this has to happen is staggering. ASML says an EUV machine churns out about 3000 wafers a day.

There can be hundreds of chips on a 300 millimeter wafer and up to 10 billion transistors per chip. They take extraordinary achievements of engineering and physics, and they're able to replicate these on a mass production scale and at a low enough cost where these machines can be used in chip fabs to churn out thousands and millions of chips for the companies that buy them. A completed EUV machine is actually made up of seven different modules, each built at one of ASML's six manufacturing sites among its 60 total locations around the world, then shipped to and reassembled in Veldhoven for testing. Then it's disassembled again for shipment, which takes 20 trucks and three fully loaded 747s. In 2021, ASML sold 42 EUV machines, bringing the grand total it's ever shipped to just about 140. With each machine costing up to $200 million, only five customers can afford to buy EUV systems: Micron, SK Hynix, Samsung, Intel and TSMC, the last three making up nearly 84% of ASML's business.

It certainly has eliminated a lot of players out of that market. So we saw GlobalFoundries back five years ago or more say that they weren't going to pursue a seven nanometer chip. The handful of huge customers it does have are furiously adding capacity to try to ease the global chip shortage, which is impacting ASML too.

We got a lot of messages from our suppliers that said, Hey, we might be late in delivering our modules to you guys because we cannot get the chips. And we said, if we cannot get the chips, we cannot make the machines to make more chips. So there's a catch-22. We're still managing, keep our fingers crossed, but it's a daily struggle.

The question is, can ASML keep up with demand? I think the answer is probably yes. Maybe the growth will exceed even their targets, that's possible, but they're certainly preparing to ramp up their production, which is, I think, good news if you're worried about a chip shortage. The world needs more chips, so we need to make more machines, which, by the way, will keep growing an average selling price as long as we can drive the cost per transistor down, which is exactly what we've been doing for the last 38 years, and we will keep doing for the next couple of decades. Before, EUV chip makers had three companies they could choose from for their photolithography tools: ASML, Nikon and Canon.

Nikon in Japan is still a competitor for DUV, but ASML is the only option for EUV. Experts say it could take decades for any other company to catch up, not only because of ASML's proprietary tech, but because it's built complex, often exclusive deals with nearly 800 suppliers. And we're unique to our customers, like some of our suppliers are unique to us, and those almost symbiotic relationships some people say are worse than being married because you cannot divorce. It takes ten years to not only get the technology, but then be accepted.

So the buyers for semiconductor manufacturing fabs are very risk averse. One of the ways ASML has insulated itself against supply chain risks is by purchasing some of its suppliers, like Berliner Glas in 2020. A fire broke out there in January, but Wennink says it won't significantly impact system output in 2022. Instead, ASML projects a 20% sales growth this year and an annual revenue growth rate of 11% until the end of the decade. It's actually driven by you. You're asking for more solutions that will help you to have a better life, to make your life easier or your life more productive.

We're changing into a sensing world. There are sensors everywhere: they're in your car, they're in your fridge, they're in your PC, they're everywhere. Sensors, they need semiconductors. All of the world's most advanced semiconductors are made in Asia by two of ASML's biggest customers: TSMC and Samsung. But the chip shortage has raised concerns about overseas dependency.

This is why you see all these initiatives around the globe: the U.S. Chips Act, the EU Chips Act, the Korean Chips Act, the Japanese Chips Act, the Chinese Chips Act. It's now a very strategic commodity. Intel just announced a $20 Billion chip fab in Ohio, and it's also building one in Arizona just down the road from a massive new fab where TSMC will make advanced chips in the U.S. for the first time. And Samsung is building a $17 billion fab in Texas. All this came after President Joe Biden proposed the Chips Act with $52 billion in subsidies for chip companies to manufacture on U.S.

soil. It means that we need to ship our machines sooner, earlier and at higher volume. So it means we need to hire more people in the U.S. It's talent. It's people.

I think that's where the biggest challenge will be. But this movement toward domestic production has another side that poses a challenge for ASML. A desire to stop sharing chipmaking technology with China. China has wanted to get into that race, but there's been politically generated reasons why China has not had access to the same type of technology as other companies. As far back as 2018, the Trump administration reportedly pressed ASML not to sell EUV systems to China. ASML still hasn't sold a single EUV machine to

China. 43, 42 countries around the globe have agreed to put export control measures on it because it's so critical. So it's not our choice, it's the choice of governments. ASML also refurbishes older lithography systems and sends many of those to China. More recent DUV machines all the way back to its early systems from the nineties.

96% of all the machines we ever sold, we ever shipped, are still working. There's a lot of debate about whether selling additional DUV equipment to China is also a national security risk by letting China increase its ability to manufacture close to cutting edge semiconductors. So I think there's some chance that in the coming years there are new restrictions that are imposed on ASML's ability to sell DUV equipment to China as well. If export controls were expanded to include DUV machines, it could greatly impact ASML's bottom line.

This is where the biggest demand is. This is where the exponential curve is. So trust me, we need every manufacturing capability on the planet, whether it's in Korea or in China, to just keep adding capacity.

Let's go look at the big boy. And then there's the question of whether demand for the most advanced chips will remain high enough to support continued development of ASML's next generation EUV machine, High-NA. This is the machine Intel announced it'll have first by 2025 and ASML has already sold four other units. This is the EXE:5000. So this is what we'll be testing for High-NA. This will be what makes our next generations even better.

But even now, before the bigger, better machines, the whole world's reliance on a ASML is only growing, no matter what gets in the way. What can really get in the way is the geopolitics, like the Russia and the Ukraine war right now, those are big geopolitical friction points that can, of course, not only hurt us but hurt the world economy. But apart from that, let's hope and let's pray that can be controlled, then it's all about execution.

And we will keep shrinking the cost per transistor and we will provide the world with ever more powerful semiconductors. That's not going to stop. Chips are in everything and they've been in short supply since just a few months into the pandemic last year. That's why it's been hard to buy everything from cars to PS5s. Turns out one company makes 24% of all the world's chips and more than 90% of the most advanced ones. The smallest, fastest chips used in today's

iPhones, supercomputers and automotive. We even have product that's landed on the last Mars launch that are taking pictures of Mars. Taiwan Semiconductor Manufacturing Company, or TSMC, is not a household name, but it's quietly making chips for every new iPhone, U.S. fighter jets, the highest end processors, you name it.

And now it's investing $100 billion over three years to ramp up production amid the shortage. The combined output of what we're doing is in excess of 12 million wafers a year. But the world's massive reliance on TSMC may also leave the global chip supply vulnerable to earthquakes, drought and geopolitical tensions with China. It's become almost a monopoly at the leading edge, and all of those manufacturing operations, for the most part, are out of Taiwan. That becomes a matter of national importance for the United States, but not only the United States, but the Western world. TSMC almost always keeps its production sites closed to U.S. video crews.

Until now. The total for space for this fab is around a 2.3 million square foot. The U.S. was the birthplace of advanced silicon, but for decades now it's been losing market share to Asia, where 75% of chip production happens now. TSMC is now bringing the world's most advanced chip making back to the U.S.

with a $12 billion fabrication plant, or fab, in the middle of the Arizona desert. It's going to be, when it gets introduced to production in 2024, the most advanced technology manufactured in the United States. We got an exclusive tour of the fab site in northern Phoenix to get the truth about the secretive Taiwanese company and why the world's largest contract chip maker is bringing bleeding edge chip manufacturing back to U.S. soil.

When Morris Chang first proposed the idea for TSMC in the mid eighties, investors were skeptical. Born in China and educated at Harvard, MIT and Stanford, Chang moved to Taiwan after 25 years at Texas Instruments. There, the government asked him to create a Taiwanese semiconductor company that would become a world leader. His idea focus only on manufacturing, what's known now as a pure play foundry.

When you're just focused on one thing, you do one thing really well. Rick Cassidy is TSMC's top executive in the U.S. He's been with the company for 23 years. The slice we spun out was foundry, and that's what we do. And we put all of our resources into doing that

one thing. Chang bet big on a need that didn't exist in the eighties . When he founded TSMC in 1987, giants like Intel and Texas Instruments took pride in designing and making their own chips. A legendary saying in the industry back then was, Real men have fabs. When Morris went out to get funding, he went to many named companies and they told him, Morris, your idea won't get off the ground. If you get it off the ground, it can't scale. But as chips got more complex, making them became an enormous undertaking.

Building a fab today takes at least two years and $10 billion. It's become nearly impossible for even the biggest chip companies, Intel, Nvidia, Broadcom, Qualcomm, AMD, to do it all and keep up with the most advanced tech. Intel, for example, still designs and makes its own chips, but it's fallen behind Samsung and TSMC in recent years, even relying on TSMC to make some of its chips. So if you were a smart designer, you didn't have to have billions of dollars in a fab behind you, for the first time, with the emergence of TSM. Now, each major step of chip making is often handled by a separate company.

Some, like Arm and MIPS, focus on IP and architecture, providing the core building blocks to design chips. Then there's electronic design automation, EDA companies, like Cadence and Synopsys, that write the software used to design chips. Only one company, ASML, makes the $180 million extreme ultraviolet light machines required to etch designs into the most advanced chips. And then, of course, there are the wildly successful fab-less companies designing the chips. Think Apple, Qualcomm, Nvidia, and many more. As these fab-less companies took off, TSMC found itself on a flywheel, making more and more of the world's chips.

And this has allowed TSMC not only catch up but, in my opinion, surpass Intel to become the world's greatest manufacturing technology on the planet and responsible for becoming one of the top ten most valuable companies in terms of market cap in the globe. TSMC was first listed on the Taiwan Stock Exchange in 1994. In 1997, it became the first Taiwan company listed on the New York Stock Exchange. By the 2000's, it had caught up with the 20 or so other companies making the most advanced chips at the time. As the tech kept advancing, more and more fell behind until today, only two manufacturers remain that can make the most advanced five nanometer chips: TSMC and Samsung.

In 2013, Apple started relying on TSMC to make its A-series chips for the iPhone as it moved away from reliance on Samsung, a direct competitor in mobile phones. Today, there's a TSMC chip inside every iPhone on the market, and Apple has moved away from Intel too, now relying on TSMC to make the chips inside most Macs. But they remain sort of in the background.

So Apple gets all the accolades when a new phone comes out. We let our products speak for themselves. Their success brings all the business that we could ever hope for. As to why TSMC hasn't allowed U.S. media into its sites before now, does part of the secrecy have to do with IP? Sure, because this IP protection is very important for in this industry, not only the TSMC but also for the other company in the industry.

In 2018, at age 86, Chang retired as chairman of TSMC. His radical, pure play foundry idea continues to pay off. With the opening of a new fab in Taiwan next year, TSMC is in a race with Samsung to make the world's first three nanometer chips, with Intel planning to get there by 2025. Along with cutting edge three and five nanometer, TSMC also makes far larger chips for everything from cars to coffee makers. To understand the different kinds of chips and why nanometers matter, let's look at how they're made.

Silicon, an abundant element found in rocks and sand, is purified and melted down, then sliced into circular wafers. These wafers are the surface on which chips are built in a grid formation. Each chip on the wafer can have hundreds of tiny layers, each made up of transistors and electrical circuits which determine what the chip can do. The minuscule circuitry is printed on each layer using lithography, extremely precise rays of light. The smaller the width of the transistor gate, five nanometers, three nanometer, the more processing power can fit in a given space with less power needed. The smallest transistors are more than

10,000 times thinner than a human hair. Most of the chips are probably about the size, a large one, of my thumbnail . On there you might have something like 50 billion plus transistors and they all have to work. These are parts that are going to be used in lots of different places: CPUs, GPUs, IPUs, etc.

They'll be used in smartphones. Bigger chips are used in most household devices, things like a TV remote or electric toothbrush. Cars often use less advanced 28 to 40 nanometer chips, and all types of chips have been impacted by the shortage. Carmakers like GM and Toyota have paused production at some plants, and Apple is cutting its 2021 production targets for the iPhone 13, with orders for the 13 Pro Max delayed by more than a month. Right now, no fab in the U.S. can make five nanometer chips, but TSMC is changing that. The F-35 Strike Fighter to these consumer products, their customer base is wide.

500 plus companies are their customers in the United States. And so as a byproduct of that, we knew they were going to need to be in the United States at some point. Chris Camacho of the Greater Phoenix Economic Council got to visit TSMC's fabs in Taiwan during the five years he was helping negotiate the deal that brought the project to Arizona. The robotics, the automation, the mechanization occurring before your eyes.

And so you can see how these things not only are so capital intensive, but also their output is so significant. TSMC is six months into building this massive five nanometer fab outside Phoenix that will pump out 20,000 wafers per month starting in 2024. The chips from the wafers will end up in iPhones, high end processors and much more. Arizona project leader Tony Chen has led 17 other fab construction projects in his 23 years with TSMC. This approach is designed for five nanometers fab.

That's a copy from the fab we have in Taiwan. Just down the road, Intel is in the midst of building two new fabs, spending $20 billion. These massive buildings, used to make minuscule chips, have brought some of the world's largest equipment to Arizona. This is the biggest crane that Manitowoc makes.

There's only two of them in existence, and it's a 2300 ton crane. Since we've started, our dirt contractor has moved over 3,731,000 cubic yards of dirt. We've also used over 260 million gallons of water. Indeed, building a fab and making chips takes an incredible amount of water, something that's not easy to find in the middle of the desert. Arizona's biggest water source is groundwater, but deep wells at big farms are using up groundwater faster than it's naturally replenished. We do need around 4.7 million gallons per day in

water to support the production. TSMC is no stranger to water shortages. Taiwan is facing its worst drought in 56 years, something that TSMC says has not impacted production . In Arizona, TSMC says an onsite water treatment center will recycle up to 90% of water used at the fab.

And then ultimately that water will be re-injected into the aquifer in partnership with City of Phoenix after reverse osmosis and other technology solutions are provided. Another challenge of producing the most advanced chips stateside? The current specialists are all in Asia. TSMC's best engineers right now are in Taiwan.

They're likely going to stay in Taiwan. The most cutting edge R&D is going to be done in Taiwan. To solve this, recruiter Roxanna Vega says TSMC is bringing over some of its top experts from Taiwan. They're seen as subject matter experts in what they do in our fabs over there. And so it'll be a temporary assignment depending.

Two maybe three years. TSMC has already sent some 300 new U.S. hires to Taiwan for 12 to 18 months to get up to speed. And the opportunity to train in our five nanometer giga fab in Taiwan is going to give them that insight of how immense and how state of the art our tools, machinery and everything is going to be here in Arizona.

Taiwan is not very good when it comes to analog semiconductor design and by moving to the United States, we'll be able to tap into a much larger number of analog designers. This diversification is a key reason for TSMC to bring advanced manufacturing to the U.S. And then there's proximity to its huge, fab-less customers based in the U.S .

like Apple, Nvidia and Qualcomm. If you want more capacity, you have to build more fabs. And that's one of the reasons that we're moving to the U.S. Our customers want us in the U.S. The U.S. government wants us here. Over 60% of their customer base is still U.S. companies. So some of these companies, like Apple, had

hinted that they want their supplier to be closer to home just in case. TSMC has 12 fabs, almost all of them in Taiwan and China. They account for nearly 54% of all global foundry revenue. And this heavy reliance on TSMC in Taiwan leaves the world vulnerable to potential slowdowns, from earthquakes, the current drought there, or the geopolitical tensions swirling around the U.S., China and Taiwan. But some refer to TSMC as Taiwan's silicon shield.

The silicon shield, TSMC, is extremely, extremely important. And I think people depend on us. The media paints a very bleak picture of this situation, but I'm actually much more optimistic in part because of this idea, the semiconductor shield. China, as of right now, needs them for their leading edge manufacturing. The U.S. also depends heavily on the chips coming out of Taiwan. A key reason the government worked hard

to convince TSMC to bring its tech here. We're not going to have to worry about geopolitical conflict. We're not going to have to worry about another major pandemic. We will have these kind of manufacturing capacities on U.S. soil. Today, only 12% of the world's semiconductors are made in the U.S. That's down from 37% in 1990.

Back in the days of Bell Labs, in the early days of Silicon Valley, we were probably 100%. Both state and federal officials are eager to entice TSMC to bring advanced silicon back to the country where it first took off. The state of Arizona has a number of programs, including the qualified facilities tax credit and the quality jobs tax credit, that's really an incentive to help lower the cost of operations. In addition to that, the city of Phoenix put together a $200 million infrastructure package that helps TSMC access water and additional infrastructure needed. The Biden administration has proposed $52 billion in subsidies for chip companies like TSMC to manufacture on U.S.

soil. It's been nicknamed the Chips Act. This is infrastructure. So look, we need to build the infrastructure of today, not repair the one of yesterday. And things like the Chips Act are absolutely critical for the success of our country, not only to compete but to recruit these kind of firms to operate in the U.S. Otherwise we're going to be importing chips for

the rest of our lifetime. Over the last 20, 30, 40 years, we've slowly slipped in that manufacturing element, especially as we have seen the decreasing cost in other countries. It's somewhere between 20% to 25% cheaper for American firms to produce their semiconductors outside of the United States. TSMC's Rick Cassidy took part in discussions that led to the Chips Act. We don't want anything more than to create a level playing field so that it doesn't cost more to make chips in the U.S. than it does in other locations.

Industry reports estimate a $50 billion investment from the U.S. government would enable the construction of 19 new fabs in the U.S. over the next ten years, more than doubling domestic chip manufacturing capability. As the shortage continues, similar investments are happening around the world.

Industry association SEMI projects 72 new fabs or major expansions will come online by 2024, ten of them located in North and South America. I heard more announcements of investments in last two or three years than my entire life. Korea will invest $450 billion in the next ten years. EU has announced roughly $150 billion in investments, and based on that, we feel that by the end of next year, we should start seeing some relief on the chip shortage.

But until then, as demand continues to soar, TSMC is raising chip prices as much as 20%, a cost that could trickle down to the price of consumer electronics. TSMC has always been able to charge a premium if it was necessary, and most of their customers recognize that if there's a good reason, they're willing to pay for it. Meanwhile, TSMC will certainly continue investing in ramping up production capacity, including in the U.S., where the 1100 acre Arizona site certainly has room for a second phase and more. So we've got a lot of land and we have the ability to do more there. It will take time, but it's not just the chip in the foundries. It's going to be the entirety of the supply

chain. So it's packaging companies. It's companies that produce the chemicals and the gases required that go into the manufacturing process. So I see this as an entirety of a shift in the semiconductor sector for the United States. As you can see, we can get into a lot of trouble when everything is in one area alone.

So I think it would be a great victory, in fact, to see the United States reverse the declines that we've had over the last few decades. Intel was once synonymous with the world's most advanced chips. It's responsible for inventing the very building blocks of modern computing from memory chips to microprocessors. Business models that have come about, Internet being one of them, is all as a result of this vision that started 50 years ago with bringing the digital world onto a chip and that pace of technology innovation is unstoppable.

Chip technology is indeed advancing at roughly the same relentless pace predicted in 1965 by Intel co-founder Gordon Moore, doubling every two years. But Intel has failed to keep up. The chips being made inside Intel's massive fabrication plants, or fabs, are no longer at the cutting edge. Intel was the Moore's Law company and the undisputed leader, and something that was supposed to take them two years instead took them more than five.

And they still struggle to get back on Moore's Law today. Now, only two companies in Asia, Taiwan Semiconductor Manufacturing Company and Samsung, make all of the smallest, most advanced chips that power next-gen iPhones, supercomputers and automotive AI. The new Alder Lake CPUs just released are packed with competitive features, but its chip technology is behind the most advanced chips made by TSMC and Samsung. They got fat, dumb and happy and they took their eye off the ball. Once you fall off the treadmill, it's really,

really difficult to get back on. It's a very dynamic and fast moving industry. But Intel's new CEO has a bold plan to catch up and help the global chip shortage. I think I have more concrete trucks working for me today than any other human on the planet. They have construction in Oregon, New Mexico, Arizona, Ireland and Israel and we expect to plant our next major fabs in the U.S. and Europe before the end of this year.

CNBC got an exclusive tour of Intel's massive factory site outside Portland, Oregon, where it's building a huge new fab set to open early next year. And so what's inside of here and what are we about to see? So what's inside of this truck here and what's just been offloaded onto our dock is one of our next generation tools. It's going to be installed in our D1X-Mod3 factory. And it's spending another $20 billion on two new fabs in Arizona, where it will make not only its own chips, but those designed by Amazon, Qualcomm and others. And it also starts building up that base within the United States so that the United States can become more self-sufficient. We asked Intel's top executives and semiconductor analysts about how Intel fell behind and whether its aggressive plans for more U.S.

manufacturing could catapult it back to the front of the pack again by 2025. The story of Intel's founding is also the story of how Silicon Valley got its name. William Shockley, the inventor of the transistor, the most basic building block of computing, moved to Mountain View to start Shockley Semiconductor Labs in 1956. A year later, the so-called traitorous eight quit to start Fairchild Semiconductor, which quickly became the world's premiere chip company.

A decade later, two of these founding fathers of Silicon Valley, Bob Noyce and Gordon Moore, left to start their own company. They first called it an N .M. Electronics, then quickly switched to the name Intel for Integrated Electronics . At Intel's founding in 1968, short term memory or RAM didn't exist.

Neither did microprocessors or CPUs, today's brains of every computer. Those are both Intel innovations. These transistors are doing most advanced computational capabilities, never thought possible, and it's also enabled the ecosystem all around us. A short three years after raising an initial funding round of just 2.5 million, Intel went public with a

market cap of 58 million. Making chips with memory capabilities was great business. So much so that well established Japanese electronics companies like Hitachi and Fujitsu wanted in. A dozen years later, their massive factories,

which had been operating 30 plus years longer than Intel's, were making memory chips far faster and more affordably than Intel could. In 1974, Intel's global market share of the memory business was nearly 83%. But by 1984, it was down to just 1.3%. So in 1985, Moore and then-president Andy Grove famously fired themselves, walked out the door, then walked back in and made a drastic pivot away from memory chips and toward microprocessors. This was just one year after the first Mac computer came out.

They kind of made a decision, it was a huge one, to get out of that business and to bet the company effectively on this new market. Remember, there was no PC industry, there was no personal computer industry back then. Last year, Intel announced it's selling most of what remains of its memory business to South Korean rival SK Hynix for $9 billion. In 1971, Intel released the 4004, the world's first central processing unit, or CPU.

For the first time, engineers could purchase these building blocks to use in all kinds of electronic devices. Intel processors were in the world's first personal computer in 1974, and its groundbreaking x86 architecture processors were in the first IBM personal computers by 1981. It revolutionized transistor density and speed with the first 32 bit processor in 1985. It took competitor AMD six years to reverse engineer a similar product. Suddenly, personal computers had to have an intel processor to be competitive. Andy Grove took over from Moore as CEO in 1987, and Time magazine named him Man of the Year in 1997.

The market for personal computers continued to grow through the first decade of the 2000, and Intel reigned supreme in making the chips that powered them. In 2011, global shipments of smartphones started sneaking past PCs. And that's about the same time Intel turned down an early offer from Apple to make crucial chips for its first iPhones. And that was a massive mistake because what they actually missed, and that was the start of it, was the entire shift from PC to Mobile. The chip world was also in the midst of another trend.

Back when Intel was first shipping out its revolutionary processors, chip companies took great pride in designing and making their own chips. "Real men have fabs" was a common saying at the time. But as Moore's Law proved true, decade after decade, chips got so complex that making them became an enormous undertaking.

Building a fab today takes at least two years and $10 billion. So huge companies like Apple, Qualcomm and Nvidia decided not to build fabs, but rather to outsource the expensive, highly specialized manufacturing process to companies like TSMC, which focuses only on its foundry business making chips for others. And this is allowed TSMC to not only catch up, but in my opinion surpass Intel to become the world's greatest manufacturing technology on the planet. Despite the wild success of companies that decided to focus on designing chips, like Apple, or only making chips, like TSMC, Intel still does it all.

That makes it an integrated device manufacturer or IDM. Keyvan Esfarjani joined Intel in 1996. Now he runs manufacturing and supply chain operations. Advanced equipment capabilities are becoming more expensive.

You got to get it right, otherwise it could be very, very costly. Since Andy Grove retired in 1998, Intel has seen a series of chief executives who have gone back and forth about how much the company should focus on the costly manufacturing end of the chip business. The most recent turnover happened in February when Bob Swan was replaced by Pat Gelsinger, who started at Intel in the seventies at age 18. 30 years at the company.

I mean, I, you know, I joke I went through puberty I started at Intel so young. At age 25, Gelsinger led the architecture of the 486 processor, then rose to chief technology officer by 2001. He left in 2009, and after leading VMware as CEO for nearly nine years, returned to run Intel this year. We needed a technology leader to help reestablish the technology company, this company that essentially put the Silicon in Silicon Valley. Gelsinger's made some major moves since he took the helm, most notably the decision to double down on manufacturing. For decades, the markets have rewarded

giants like Apple and Qualcomm for being fab-less. But the chip shortage has made manufacturing chips a more attractive business, allowing TSMC, for example, to raise chip prices as much as 20%. It takes time to build this infrastructure, but the good news is the world is rallying behind building additional capacity. Intel is adding capacity by building a huge new fab at its massive campus outside Portland, Oregon. D1X-Mod3 is about 250,000 square foot per level of the building. We got an exclusive first look inside the expansion called D1X-Mod3.

And what exactly are you manufacturing? We're manufacturing the latest generation of microprocessors for Intel and working to enable Intel's accelerated process for IDM 2.0. At the company's Intel Accelerated event in July, Gelsinger laid out an aggressive IDM 2.0 roadmap for how it plans to ramp capacity and catch up with big leaps in processing technologies. By 2025, Intel says it will surpass the chipmaking capabilities of both TSMC and Samsung. We are on a march for yearly innovation, setting a pace for ourselves and the industry to not only get back but to get ahead again.

Why should anyone trust Intel again? And so Intel will have to make a bunch of promises, both verbally and financially, in my opinion at least, to get anyone to listen. Intel has 15 fabs all over the world: China, Israel, Ireland and the U.S. in Oregon, Arizona, New Mexico and Massachusetts. It has assembly and test sites in Vietnam, Malaysia, Costa Rica and China and the U.S. It says it makes 8000 products, outputting 2 billion units a year for some 2000 customers.

Now it's expanding that production, specifically in the U.S. and Europe. It's got a major fab expansion underway in Ireland and is reportedly in talks for projects in Italy and Germany. Doubling down in its capacity requirement to support the growing needs of their customers around the world is absolutely a significant responsibility of what Intel has got to go drive. And in March, Intel announced it's spending $20 billion to build two huge new fabs in Chandler, Arizona. It broke ground in September this year with plans to output chips for PCs and data centers by 2024.

It's a very long time to build the concrete, the chemical delivery, the electrical systems. All of this needs to be perfect for a fab to run for something that's creating lines and dimensions that are 10,000 times smaller than your hair. When we toured the fab project in Oregon, semi-trucks were dropping off some of the 1200 massive tools used to make the chips.

All of our tools tend to be in the millions of dollars, tens of millions of dollars. They weigh anywhere from 10000 pounds to 100000 pounds. We also got a rare look inside the fabs bustling clean rooms, donning bunny suits that help keep dust and other particles away from the minuscule circuitry on the chips. We're talking about clean room that is 10,000 times cleaner than a heart surgery room. It's about the equivalent of about 20 American football fields is the amount of space we have here, which is clean room space. It's filled with yellow light to prevent exposing the chips to shorter wavelengths of light than the lithography machines use to print designs on the chips.

We have different chemistries and gases that we use to make our chips here at Intel, and we segregate those exhaust streams into these ducts you see here and are subsequently treated so that we're environmentally responsible in providing clean air coming out of our factories. Making chips also takes a massive amount of water, not a plentiful resource in the Arizona desert. We're currently out at the Ronler Acres water treatment facility, where today we've reclaimed over 2 billion gallons of water and reuse back into our manufacturing systems and process. We utilize approximately 9 million gallons a day and we can serve about 95% of that. The chips being made here are ten nanometer, used in PCs and data centers. Only TSMC and Samsung can currently make five nanometer chips, the most advanced node on the market.

In fact, Intel relies on TSMC to make a good number of its chips. We are one of their key customers and that collaboration continues. To understand why Intel has chips made by one of its competitors it's trying to catch, let's talk about the different types of chips and the supply chain. Different sized chips are found in different types of electronics. Intel makes a lot of ten and 14 nanometer

server chips that function as the brains of computers, CPUs and powerful chips used in data centers, GPUs. Less advanced 28 to 40 nanometer chips are used most in the auto industry in components like anti-lock brakes and airbags. Bigger chips are also used in household devices like coffee makers or electric toothbrushes. Five nanometer chips, the most advanced chips currently made, are highly sought after for data handling and artificial intelligence processing, used in leading edge technologies like the latest iPhones, NASA rovers and F-35 fighter jets. Making five nanometer chips requires an extreme ultraviolet lithography machine that uses very small rays of light to etch the tiniest designs onto the chips. Only one company, ASML, makes these EUV machines and they cost upwards of $180 million.

Costs are going to go through the roof. But if you can't yield the process without it, like, you have no choice. Intel didn't buy EUV machines until a couple of years after TSMC, which explains why TSMC was able to reach five nanometer firs.

And now TSMC will be the first to make five nanometer chips in the U.S., building a $12 billion fab just down the road from Intel's new Arizona fabs. So where does all this leave Intel? It's currently in high volume production of ten nanometer chips after years of delays. Ten was supposed to be here in 2015. We still don't even know the reasons. Their rationale for why ten failed was that they just tried to do too much.

In July, Intel rebranded to avoid the nanometer based nomenclature used by other chip giants. Its seven nanometer chip, which it now calls Intel 4 or Meteor Lake, has been pushed back about a year. This recent delay, the first setback under Gelsinger, has seven nanometer in production for the second half of 2022, around the same time both TSMC and Samsung have committed to start production on their three nanometer nodes. They've been having process issues for ten years. 14 nanometers was delayed, 10 nanometers was delayed, seven nanometers is delayed, it's not like it's new. I still don't understand how you could let something slip as much as they have.

Like, it's shocking. We had some missteps. The strategy had become a little bit confused on the role that we're going to play in manufacturing for the long term. And now we're leaning back into that

with clarity, with clear urgency. The competition between chip giants and subsequent ramp in production is a positive for the chip shortage, which has impacted all types of chips. Apple is cutting its 2021 production targets for the iPhone 13. Carmakers like GM and Toyota have paused

production at some plants. When the personal computer market skyrocketed during the pandemic, it drew down supplies of CPUs and GPUs. Intel blamed this component shortage for its PC chip business shrinking 2% in Q3 2021, causing shares to fall more than 10% after earnings were announced in October. Intel stands alone as the only U.S.-based

company that designs and manufactures advanced chips at scale. Traditionally, it only manufactures its own designs. But now, in the face of the shortage, it's changing that. You're going to not only make our own wafers, you're also going to use those fabrication facilities to be producing wafers for customers that they want us to use their design. It's totally a right hand turn. We have always had much debate about it.

Intel is calling the new stand alone business Intel Foundry Services. We already have our first revenue with the Amazon packaging deal. Our next big customers like Qualcomm and the U.S. government. Foundry has been wildly successful for the other two at the leading edge, Samsung and TSMC, but analysts are not sure it'll work for Intel. Amazon presumably is going to be data center parts.

It's going to be that many units, right? Tiny, right? It's like five years away. And if it turns out that Intel is a viable founding partner, great. But if they're not, it's no skin off their nose. The only benefit I would see to using Intel is if you wanted something to be created onshore in the United States. And the government is greasing the rails with the Chips Act, a proposed $52 billion in subsidies for chip companies committed to making them in the U.S.

This is infrastructure. In 1990, 37% of the world's semiconductors were made in the U.S., but last year that was down to just 12%. A moonshot would be that the U.S. is at 30% of manufacturing in a decade or so in the future. And I think the Chips Act as it's structured today is a great step to start to turn that in a positive direction.

As I like to joke, God decided where the oil reserves are. We could decide where the fabs are. But analysts say much more is needed to help the U.S.

bounce back. If the goal of that money is to bring significantly more capacity onshore, it's not anywhere near enough. They need ten times that amount more. Because 92% of the world's five nanometer chips are currently made in Taiwan, the entire global chip supply is vulnerable to natural disasters common there, like earthquakes and its current drought and escalating geopolitical tensions between China and Taiwan, and subsequently the U.S.-China trade war. Every aspect of defense, intelligence and government operations is becoming more digital, and do we want to rely on foreign technology for those critical aspects of our defense and national security? I don't think so.

It's critically important for not just the global supply chain, but for the national security that we must maintain this journey. However, it is going to require Intel to put its playbook into work. The next steps in this playbook include a chip so efficient Intel's named it not with nanometers, but with an even smaller unit of measurement, the angstrom. Intel says the 18A, which is in development for 2025, will accelerate it past its competitors.

We will be the world's largest integrated design and manufacturer of silicon for the long term. It is a tall order and it is not my expectation that he will hit that. But if he could hit that timetable, it would put them back, in my opinion, on par with TSM head to head.

2022-09-01 13:25

Show Video

Other news