Motley Fool Money - You Want Chips with That?

Episode Date: November 18, 2023

Chips – the modern world runs on ‘em. But who are the players that bring these tiny technological wonders into existence?  Chris Miller is a Professor at the Fletcher School of Law and Diplomacy at Tufts University and the author of Chip War: The Fight for the World’s Most Critical Technology. Ricky Mulvey caught up with Miller to discuss: The semiconductor supply chain Intel’s turnaround and new focus on foundries  How TSMC balances secrecy, transparency, and trust when it comes to intellectual property Tickers discussed: TSMC, SSU, NVDA, AAPL, AMD, INTC, ASML Host: Ricky Mulvey Guest: Chris Miller Producer: Mary Long Engineers: Dan Boyd, Tim Sparks Learn more about your ad choices. Visit megaphone.fm/adchoices

This episode is brought to you by Indeed. Stop waiting around for the perfect candidate. Instead, use Indeed sponsored jobs to find the right people with the right skills fast. It's a simple way to make sure your listing is the first candidate C. According to Indeed data, sponsor jobs have four times more applicants than non-sponsored jobs. So go build your dream team today with Indeed. Get a $75 sponsor job credit at Indeed.com slash podcast. Terms and conditions apply.

There's nothing more complex or expensive when it comes to manufacturing right now, which is why there's only a couple of companies that can afford it and even fewer that can actually undertake these capabilities in a high volume manufacturing process. And the key to making chips is not just that you're making one transistor that's coronavirus size. It's that you're making them by the billions, literally by the billions, for $50 or $100, which is the price of a typical chip in a smartphone. I'm Mary Long and that's Chris Miller, a professor at the Fletcher School of Law and Diplomacy at Tufts University. He's also the author of the best-selling book, Chip War, the fight for the world's most critical technology. Ricky Mulvey caught up with Miller to discuss the link between semiconductors and the printing press, what it takes to build a chip the size of the coronavirus and the geopolitical implications of the AI arms race. Chris, I'm really interested in the story of TSM-T-Taiwan semi.

It's one we haven't really told on the show, but I think it has to start the separation of fabs and chips. Maybe the starting point for that is what Lynn Conway and Carver Mead saw. Well, before then, almost all chips were designed and manufactured by the same companies. But they realized that as chips were getting more complex to design and more complex to manufacture, that there'd be benefits to splitting these two processes, sort of like how authors don't write and print books themselves, Gutenberg figured out how to mechanize printing and now they're done by different groups. The same thing happened in chips. And so today there's companies that focus solely on design and then they basically email the design file to companies that focus solely on manufacturing. That's enabled

extraordinary specialization and huge advances in efficiency that's resulted from that. And Taiwan, I think even before Taiwan semiconductor, figured out how to insert itself into this semiconductor supply chain. What was the national priority that led them to do this? Well, there were two things that drove Taiwan to focus on semiconductors. One was just the desire to find employment for the country's workforce. They were moving people off of farms into cities or looking for manufacturing jobs. And electronics seemed like a good industry to bet on. And that proved right. But the second factor was that Taiwan was in the 1960s and 70s looking to make itself indispensable for the United States because they were facing down communist China just

across the Taiwan Straits. And they were worried that they couldn't rely on the U.S. to help defend them. And so Taiwan focused on semiconductors in part as a way to deeply embed themselves into U.S. supply chains, the U.S. economy, and to make sure the U.S. had multiple reasons to stay interested in Taiwan. Yeah. And then there's kind of in comes this character, Morris Chang, who becomes sort of foundational to the modern world that we need. have. And he's able to predict the future of these fabulous chip companies. But before that, he's trying to convince the executives at Texas Instruments that this future is coming.

Maybe what's, what was the pitch? And why don't, why didn't the, the folks at Texas Instruments bite on it? Well, Morris Chang saw what Lynn Conway had realized that it would be very helpful to split design and manufacturing because of the efficiencies that would ensue. And he pitched that to his fellow executives at Texas Instruments. And they just couldn't see it. They already had a very profitable, very successful business. They were one of the leaders at the time in the industry. And this was a radically disruptive change to their business model. And so they pushed his idea out. And eventually they pushed him out. And so he left Texas Instruments in the middle of the 1980s and held on to this idea of having a foundry, a company that would only manufacture, do no design,

and do what Gutenberg did for books, but in the chip industry. So when we often think of angel investors, we think of Silicon Valley, maybe Sequoia Capital. Chang links up with the Taiwanese government to start TSM. And you write, quote, from day one, TSM wasn't really a private business. It was a project of the Taiwanese state, end quote. What's it like having the Taiwanese government is an angel investor? Well, I think the Taiwanese government deserves some credit for being very helpful in terms of the initial funding. in terms of providing a highly trained workforce in terms of making the environment, whether

it's tax or regulation, very conducive to chip making, while also not making any of the errors of being excessively intrusive on TSM's business. So TSM benefited from just the right amount of government support and just the right ways to launch itself in a very competitive industry. And with this new business model that Morris Chang had pioneered quickly win market share and by the 2000s, just 15 years after it was created, already was finding itself as a pretty irreplaceable part of not just the chip industry, but the entire tech sector. It's now at a point where there are companies that rely on it almost completely.

I think Nvidia can't make its most highly advanced chips without TSMC. So how does this company approach pricing? Because it's basically the only game in town, but it also wants to have a good relationship with the companies it works with. Well, that's right. It's almost a monopoly, but not exactly. There is a bit of competition from Samsung, for example, at the cutting edge. But you're right that for Nvidia, for Apple, for AMD, for many of America's biggest tech companies, TSM is either their only supplier for high-end ships or their primary supplier. But TSM also realizes that they've got to fit in

their customer's cost structure. And they've got an incentive to enable their customers to be highly successful. When InVIDIA was founded, for example, it started working with TSM very early on. And TSM has benefited as much as anyone from Nvidia's extraordinary growth. Now, it's one of TSM's largest customers. And so there's an alignment between customer success and TSM's success, which has made the company so successful. Yeah, at this point for Invidia, I understand why it wouldn't have foundries in its beginning, right? It only wants to focus on the chip designs. But at this point, there's a case. to be made for vertical integration when you're a trillion-dollar company, why doesn't

Nvidia, you know, try to dabble with the foundries, try cooking its own recipes? Well, the thing is you can't dabble in foundries because it's so expensive and so complex. If you're going to do it, you've got to go all-in. And all-in is a huge bat. For one thing, the capital expenditure required is enormous. A new cutting-edge TSM facility costs $20 billion. And so even for a company worth a trillion dollars, that's a huge sum of money for a facility that will be cutting edge for just a couple of years.

The second thing is, it's really, really, really hard. There's only three companies in the world, TSM, Samsung, and Intel that are anywhere close to being able to make cutting edge processor chips. And the complexity involved is such that despite the extraordinary skill at a company like Apple or Nvidia, they don't have the expertise that you need to manufacture chips. They've got it in design. They've got it in software. But chip manufacturing is an entirely different discipline.

And so it would take them year. I think, to ramp up the manufacturing know-how to actually make their own chips. You've written about one of the reasons it's so hard to catch up to TSM and a lot of these leading-edge chip makers is basically copying last year's design is a hopeless strategy. That's been mistakes for earlier when the Soviet Union was trying to build up their chip manufacturing industry, and now it's created this moat for TSM. So I guess, to color that, and something I have trouble understanding, like what's going on right now in chip manufacturing, where it's so far ahead of two years ago that one would be hopeless

if a company was following chip making from two years ago? Well, the key is Moore's Law, which has predicted accurately since the mid-1960s that the number of transistors per chip would double roughly every two years, which means that transistors have to shrink by a dramatic rate every single year, smaller and smaller. Today, the transistors, which are the tiny switches that flip on and off on the chips in your phone or on your computer, today each of those transistors is measured in nanometers, billions of a meter. They're smaller than the size of a coronavirus. And on your smartphone, there are 15 or 20 billion of them that are manufactured with basically perfect accuracy.

And so to manage to, A, accurately produce, precisely produce right now 15 billion transistors for a smartphone chip, all at the cost of $50 or $100, and then B, to shrink that dramatically further in one or two years, you know, understand why the complexity involved is so immense. And for companies that aren't already close to the cutting edge, even getting in the ballpark requires years of effort and many, many billions of dollars of investment. Yeah, for the layman, what's it take to manufacture something that's the size of a, I know there's basically the flattest mirrors ever created. You have ASML as a part of the process building these machines. But for the layman, what's it take to build a chip that's the size of a coronavirus? Well, it's the most complex

manufacturing process humans have ever undertaken. You have to acquire ultra-complex machines from companies like ASML, which you mentioned, or applied materials in California. And these tools are capable of manipulating materials at basically the atomic level. They can lay down thin films of materials, just a handful of atoms thick, or the ASML tool you mentioned, has the world's flattest mirrors that are capable of reflecting extreme ultraviolet light with almost perfect precision. They can carve patterns onto semiconductors. There's nothing more complex or expensive when it comes to manufacturing right now, which is why there's only a couple of companies that can afford it, and even fewer, that can actually undertake these capabilities in a high

high-volume manufacturing process. And the key to making chips is not just that you're making one transistor that's coronavirus size, is that you're making them by the billions, literally by the billions, for $50 or $100, which is the price of a typical chip and a smartphone. Seems like a good transition to start talking about the Chips Act, where we're trying to bring some of that cutting-edge technology to the United States. That seems really difficult to do, even if you're throwing more than $50 billion at semiconductor fab, fab plants. There's a lot of historical lessons from your book. I think there might be one parallel, which was sort of the end of the easy money era in Japan, where they were trying to really

build up memory chip manufacturing. Maybe it wasn't as successful as Sony or the Japanese government would have liked. Do you think there are any historical lessons from that capital spend by the government, or is there a better historical comparison for what's going on in the United States right now? I think the lesson from Japan is that you've got to not only focus on market share, which is what Japanese firms focused on and they succeeded in winning market share, but also profitability. And the problem the Japanese face was they poured money into building out their capacity. They did a great job in very efficiently manufacturing, but they ended up without effective business models, competing with each other, undermining each other's profitability.

And in the end, that just proved unsustainable once the government began removing the types of policies that made capital so cheap. I think in the U.S. right now, Now, the U.S. government's goal is not really about surging capital in the industry for the sake of just doing that. It's about closing the cost gap that does exist between producing the U.S. and producing in Taiwan or producing in Korea. And that's why the Chips Act is going to try to provide incentives that are 15 or 20 percent of the cost of a new facility to make it more competitive to produce chips in the U.S. So we're going to see over the next year a series of announcements from the U.S. Commerce Department of grants to different firms, both U.S. firms and foreign firms, we're going to be building new plants in the U.S.

Yeah. So TSM is one of the companies that's building plants in the United States, but something you've talked about is that they're not bringing the full cutting-edge stuff to the plants in the United States in part because it's a national security play for Taiwan to keep the most cutting-edge stuff in-house. Are there any horses in the race building these fabs in the United States that can compete with that real cutting edge stuff coming out of Taiwan? Well, there's, in addition to TSM, there's Samsung building a big new facility in Texas, and Samsung is just behind TSM when it comes to manufacturing technology.

It has much smaller market share, but in terms of the technology itself, it's not at all far behind TSM. And there's Intel, which used to be the world's most advanced manufacturer, but about seven years ago started falling behind Samsung and TSM. Now Intel, under their new CEO, Pat Gelsinger, has, over the last two years, undertaken a huge restructuring, trying to get its manufacturing back on track. And Intel says that by 2025, they're hoping to have caught up to TSM. And we'll have to see whether this pans out. But Intel certainly sees itself as competitive with TSM and with Samsung in the next couple of years. And so the best case is that in a couple of years,

we actually have three companies that are competing at the cutting edge, each of which will have facilities that are close to cutting edge in the United States. But it's not at all guaranteed that all through those companies will stay on track because they've got to roll out new processes every year or two. And it's certainly conceivable that one or several of them will slip up. Yeah, I want to focus on Pat Gelsinger's attempted turnaround at Intel right now. He has quite the challenge on his hands. I think one of the things that Intel did that really that you've talked about that is a real going to be difficult for them, is they were a designer and they were a foundry. And this ended up creating a lot of problems with their customers because, you know, if there's

a delay in getting a customer's chip made, there's always sort of that pressure of like, well, you're making your own chips at the foundry. Have you seen any movement in repairing a lot of those relationships at the foundry level for Intel? Is this something that's real, like, how necessary is this for Intel's turnaround? Or, you know, is it better for, let's say, investors, semiconductor observers to watch what they're doing and maybe advanced computing, artificial intelligence, design, that kind of thing. Well, I think if you listen to Pat Gelsinger, he's put this new foundry business at the center of Intel's new business model. And there's a really strong logic behind it because right now in chipmaking, there's a really direct relationship between the number of chips you produce, the cost per chip

and the technological capabilities that you're able to develop because you gather data from each chip you produce. And so you hone your technology over. over more and more waferes that you manufacture. And so right now, Intel produces far fewer chips in TSM. And so the foundry business is partly intended to ramp up the volumes that Intel are producing to benefit from the economies of scale and the learning that accrues from that. And so Intel is right now trying to convince other companies to manufacture with Intel. And this is going to be, I think, far from easy for Intel to do.

In the past, Intel's manufacturing was exclusively focused on serving Intel's in-house designers. They had just one customer. It was Intel, whereas at TSMC, the entire business model has been focused on being open, transparent, and flexible to a wide array of customers, since TSM has always had dozens and dozens of different companies that it works with. And so Intel not just has a business model shift, it also has a culture shift that has to undertake as it becomes more focused on serving different customers rather than simply optimizing its manufacturing to suit its in-house designers. So intellectual property and protecting that seems to be pretty important for these chip designers, chip manufacturers.

What does that balance look like for, let's start with TSM, because that's the model of a company that does it well, between secrecy and transparency for their intellectual property? That's right. It's a huge amount of trust that companies like NVIDIA or Apple place in TSM, trust to handle their intellectual property, but also, perhaps more importantly, trust to manufacture. their products on time and at the quality level. Apple for every single iPhone launch over almost the last decade, they've relied on TSM as a sole supplier of the critical chips inside of each iPhone. So if TSM is delayed, every single iPhone is delayed. And so that's the trust that really matters. Your core product relies on components that can only be produced by one company. You are betting very heavily on TSMC delivering. And TSMC's developed a reputation where they deliver and they've convince their customers that they can bet their entire businesses on the fact that TSMC will keep

delivering year after year. We've seen Apple try to move some of its production to India. Is any of this a response to that choke point with TSM? Or do you think this is more entirely a response to sort of the COVID restrictions in realizing that maybe we don't want to entirely rely on the People's Republic of China for iPhone manufacturing? So if you look at the iPhone manufacturing that happens in China and the iPhone manufacturing, manufacturing, you brought up in India, it's all really assembly rather than manufacturing.

All of the key chips, the processor that runs iOS, they're all made in Taiwan, whether the phones assembled in India or in China. And that's not going to change anytime soon. The operations in China and India are basically taking chips mostly imported from abroad. The main processor, the memory chips, the image sensors, those are all imported and putting them together inside of an iPhone. And so actually, the assembly process is relatively low value, much simpler than the manufacturing of the chips themselves. So it's easier to sort of transpose into a different area. I'm going to try to end this in a positive note, but I think there's first a maybe a more depressing question. You've linked sort of the AI weapons race with the Cold War. And in the Cold War,

there were pretty clear goalposts, right? Like, who's going to launch a satellite first? Who has the most hydrogen bombs in a barn. That seems a little bit different with like an AI weapons race, though. So for folks like me who are maybe a little scared by it, confused by it, don't know what to look for. What are the goalposts for these major countries in the AI weapons race? Well, I think it's very difficult to assess from the outside. And I think even militaries themselves are struggling to know how effective AI enabled systems will be on the battlefield. If you look at what AI enabled means, AI is being deployed in everything from predictive maintenance, so you know when to undertake maintenance on an airplane engine, to helping

drones fly more effectively to managing the communications in a context in which there's lots of jamming of communications, finding the part of the wireless spectrum that's free from jamming. So there's all these different use cases of AI. We're seeing in the Russia-Ukraine war right now, many of these being tested out in real time. But I think we shouldn't think of AI weapons as an entirely new category. We're going to have AI deployed to every aspect of how weapon systems are designed, how they're manufactured, how they are maintained, and then how they actually operate. And so rather than a new category of weapons, like nuclear weapons were new category of weapons, this is going to be an enabling capability

that will be visible in many different ways across weapon systems. And so people talk about AI arms control, for example, I think that's not really the right model to think of because there won't be an AI weapon. There will be AI across the battlefield and across the military's logistics and sustainment system. It'll be hard to say, this is AI and this isn't AI because everything will have lots of AI applied to it in different ways. Yeah, and then if I can end this, maybe this is not too hopeful, but you know, you've written about weaponized interdependence. And I know that we can play the game of like, will China invade Taiwan, Something that comes through in your book is that semiconductor manufacturing can create maybe not

peace, but at least stability, right? After the U.S. left Vietnam drew down its military presence in the region, it was able to sort of keep the dominoes of communism from falling by helping with these semiconductor plants manufacturing growing in that region. So are you seeing any of any semiconductor manufacturer? or these supply chains create a type of peace, even if it's uneasy between countries today? Well, I think the optimistic view is that because Taiwan is such a critical manufacturer of chips, no one will risk disrupting its ability to keep producing it. So there are some people who think,

look, China would suffer immense economic damage if it were to knock off Taiwan's chipmaking capabilities. Therefore, China won't try to do it. Now, the assumption there is that Chinese leaders are going to be primarily focusing on maximizing GDP or living standards or be afraid of economic costs rather than achieving their stated political goals of asserting control over Taiwan. And I think that's an assumption that might be true. But I guess my confidence in that assumption has declined in recent years, both because in 2022, we saw in Putin's war in Ukraine pretty clear example of a leader who was motivated by decision-making factors other than economic optimization. And because if you look at She's China, it's clear that the economy is no longer the

primary focus. She's presided over slowing growth, rising youth unemployment, and he's pursuing the policies that are making that slow down more and more inevitable. And so I think we can hope that economic rationality will prevail. But I guess when I look at the political dynamics in China, I'm less and less confident that Chinese leaders are going to make decisions solely based on what's better for GDP or what's better for the Chinese economy. I think you have to take them seriously when they say they believe asserting control over Taiwan is part of their political program and a central goal when they measure success or failure. Listen to what people say, not what you hope they'll say. Chris Miller, I appreciate your time and your insight. Love the book.

And thanks for coming on the show. Thank you for having me. As always, people on the program may have interests in the stocks they talk about. And the Motley Fool may have formal recommendations for or against. and buy or sell stocks based solely on what you hear. I'm Mary Long. Thanks for listening. We'll see you tomorrow.