Odd Lots - How the U.S. Lost Chip Dominance and How It Can Be Regained

Episode Date: January 14, 2021

The U.S. was once a manufacturing leader in semiconductors. That's no longer the case, given the rise of contract manufacturing and outsourcing, the dominance of Taiwan Semiconductor, and Intel's own design stumbles. But how did it come to this? And can it be reversed by government policy? On this episode we speak with Willy Shih, a longtime tech industry veteran and a professor at the Harvard Business School, to answer these questions.See omnystudio.com/listener for privacy information.

Thanks for listening to Odd Lots. Follow the show on Amazon Music for more future episodes or just ask Alexa, play the Odd Lots podcast on Amazon Music. Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Wisenthal. And I'm Tracy Allaway. Tracy, remember when I said a few weeks ago that I thought we should do a semiconductor odd lot spinoff series? No, not a whole, not a series, like a whole new show. The idea is a whole new show on the subject. That's not what you said. You said you wanted to transform odd bots into a semiconductor podcast. So this is the latest step in your plan to do that, I assume. Yeah. So we're not going to do that. And realistically, I don't think we're going to have a new

spin-off, Oddlots branded semiconductor show. But I do think, but we are going to do some more episodes on the topic because it really does just intersect with a lot of different themes. there's sort of the pure tech question of their role in the economy, why different companies are rising and falling. It also intersects with things like national security and sort of economic strength and some of the questions we've seen during the COVID crisis about, say, domestic manufacturing and reliance on manufacturing in other countries. So it's a rich load to mine.

Yeah. Well, I mean, you could go even further. So it kind of gets to, I guess, the heart of a large body of economic thought, right? How do you balance domestic interests with market forces? So when the entire market is telling you to outsource manufacturing to a cheaper place and they can do it more efficiently, is that the way to go for something as important as a sensitive technology like a semiconductor? And I think we're seeing those tensions over and over. I think we see them reflected in a lot of U.S. politics and the way people feel at the moment.

But of course, we also see that reflected in the U.S.-China trade war. And technology is really ground zero of that. And within technology, semiconductors are the epicenter of the ground zero. I'm mixing my metaphors, but I think you get it. The epicenter of the ground zero. No, I get it. I think you're absolutely right. Like, it's kind of a, it's a business topic, but it intersects with a lot of

of our favorite macro themes that we discuss all the time. And I think another sort of macro theme is, you know, we talked about it a lot in 2020, but the sort of rethinking the role of the public sector and rethinking the role of government spending, rethinking the role of government investment. And I think that's also, I mean, that's a, that was a big topic for us. It's a big topic nationally. I mean, I think, you know, if you look at one of the successes that we've had in this crisis, Most people would say the speed with which we were able to develop a vaccine for COVID was in part due to the government investment in Operation Warp Speed and, of course, the longstanding investment in pharma. So there's all kinds of reasons to sort of use this moment as a chance to reconceptualize what is the role of the public sector. How can the public sector spur more innovation, manufacturing, production, production.

So it's a great topic because it hits a lot of odd lots of themes, I would say. Yeah, that's right. And I think we have seen a proposal in Washington about how to boost domestic semiconductor capabilities. So, yes, it hits all the sweet spots. It's something that's happening right now as well. And I think it's going to be an interesting conversation. Yeah, I'm really excited about this. Our guest today is Professor of Management Practice at Harvard Business School, Willie She.

He writes a lot about the topic, about semiconductors, about manufacturing, U.S. manufacturing, in particular, what it takes to make it work. So we're going to dive into all of these questions with him. Professor Sheath, thank you so much for joining us. Well, thank you for having me, Joe. So I'm trying to think, like, where the best place to start is. But, you know, obviously we were talking about Intel, and for a long time, Intel is not just the leader in semiconductors. conductors, but the sort of like the vanguard of U.S. tech and tech manufacturing capability. And of course, the story over the last couple of years has been the problems they've run into in

manufacturing, also business model questions. But to start, why is it important for a country like the U.S. or a country like China for that matter, but why is it important for a country like the U.S. to have this capability at scale to produce high-tech goods such as semiconductor. Well, I think the reason semiconductors have attracted so much focus in this regard is because they are a building block for so much of modern technology. In fact, if you look at, and it's been really highlighted by the pandemic, kind of the role of electronics technology, communications technology, whether it is conducting class, and meetings and business on Zoom or it's ordering online, underpinning all this,

sort of the core building block are semiconductors. When you start to look at the numbers on what percentage of the manufacturing happens in the U.S. and how that has declined over the years, it's only about 12% for the microelectronics that we're talking about for things like Intel microprocessors or the chips that go over. the smartphones, and you look at how much of the dependencies is in Asia, and combine that with the rhetoric around dependency on China or areas that China claims like Taiwan, I think that has really brought attention to a concern that I would say has been around in some circles for the past 20

years. So I think there is a recognition that semiconductors are strategically important. important for the reasons that you just laid out. But we've discussed this in some previous episodes. It's an incredibly competitive industry and it has huge startup costs. How difficult is it to actually set up and create a semiconductor manufacturing capability? And what are the business decisions that kind of go into that? Well, there are a couple of things. that I think have evolved, as you get to the more advanced technologies, the cost of setting up kind of a minimum efficient scale

manufacturing facility, it used to be, you know, back 30 years ago, you could do this for a few hundred million dollars, okay, or at the beginning it was a few million dollars. Now to set up the most advanced leading edge fab will probably cost about $20 billion, okay? And that's a substantial capital cost that really strikes fear in the hearts of many CFOs as they look at the level investment. And when you make that level investment, you have to make it successful in order to earn a return. That means you have to be able to learn all those process technologies.

You have to be able to manufacture at a high yield. Okay, because if you don't have a high yield, then you're throwing out half for your output, and then your cost just, you don't have anything to absorb your cost. So it's a technological challenge and it's a financial challenge as well. So, you know, we're talking about Intel. You're the co-author of a book, Producing Prosperity, Why America Needs a Manufacturing Renaissance. So when thinking about this question, it seems like there's kind of two parts to it. One is, okay, something is going on at Intel. They're having tech problems.

They're having business model problems as well. But there's also a sort of broader U.S. manufacturing malaise or, you know, a reason that there's less and less of it happening here. How do you sort of disentangle the two things? how much of it is sort of macro factors about something that's going on in the U.S. that's causing this decline in high-tech manufacturing capability and know-how versus specific micro things going on at a company that happens to be stumbling. Well, Joe, let me start with the macro thing because I think, in essence, the financial risk

and the technological challenge are linked because what we saw, We've seen this type of crisis before. It was actually in the 1980s. And at the time, it wasn't China. It was a Japanese who were manufacturing memory chips, so-called DRAMs, dynamic random access memories. And in the 1980s, the Japanese had invested in manufacturing. And they were actually much better at it than American companies, even like Intel. Intel was founded as a memory company.

Many people forget that, right? But Intel, Texas Instruments was a big player in memories, Mastek companies like that. And the Japanese out-executed American manufacturers because they were delivering higher yields. So if you think about, I'm going to sink all this money into a factory, if I have higher yields, that means my cost is lower. and Intel really was pushed to the wall on memories,

and they decided to get out of memory manufacturing and focused on microprocessor manufacturing. At the time, you know, IBM had picked their microprocessor for the IBM PC. There were early comments in Intel where they weren't sure if, you know, this PC thing was going to be big. Okay, but in some sense, they were smart and they were lucky in moving to microprocessors, okay? And then they were able to escape this problem

of being the best manufacturer. Okay, now at the same time, what we saw in American industry was this separation of design from manufacturing, which was really enabled by a lot of these computerated design tools, right, which, you know, separated out the manufacturing process. And then we saw the rise of some Asian, companies in particular in Taiwan, but also Korea and in Japan at the time, who said, well,

we're very happy to do the manufacturing because we're not the experts of the design. Right. So we saw that bifurcation. Okay. And now as that has developed, what we've seen is as you go to the more advanced manufacturing processes, there were very few companies who could kind of keep in that game. Intel has historically been one of them. Because after they left the memory business, they invested a lot in manufacturing to give their microprocessor business a lead.

And historically Intel led with manufacturing in that regard. And they were able to beat competitors like AMD because their manufacturing processes were a generation or two ahead. So it didn't matter if you had a design that might have not been quite as efficient. Okay. Now, AMD also had their own manufacturing, but then they spun it off a decade ago into Global Foundries. And once AMD separated their manufacturing from their design, and, you know, they left Global Foundries and they shifted over to Taiwan Semiconductor, okay, once AMD now had access to a more advanced process than Intel did, they have been able to take a lot of share from Intel, right? So, you know, this problem has been a long time in developing,

but the roots of it are the manufacturing part is capital intensive and it's hard. It takes a lot of discipline. Okay. And you can make more money on the design side. Right. So, hey, if somebody else wants to do the manufacturing, let them do it. I'll be more profitable. And a lot of people have said, well, the U.S. semiconductor industry had a renaissance in the 1990s because a lot of those companies let go of the manufacturing part. Not everybody did, but a lot of them did, and they've been able to be very successful, companies like Qualcomm and Invidia in particular, and now AMD, by letting go of manufacturing. So this is the, I think it's called IDM model, where you design and manufacture chips, like what Intel has been doing versus the fabulous foundry model,

like Nvidia and Qualcomm. We got into this a little bit with, well, in the episode with Stacey Razgan, the semiconductor analyst over at Bernstein. I'm curious, like, the consensus in the industry right now seems to be if you move to the Fadless Foundry model where you sort of outsource your manufacturing to a foundry like TSM, that that's the way to go. But you're arguing that there are some hidden costs or some synergies that are. a company might lose in that model. Is that right?

Well, yes, I think so. And here I would argue that more broadly speaking, there's a set of capabilities that comes with your manufacturing, which can feedback into design as well. Now, I think in the chip industry, what we have seen is manufacturers have continued to work very closely with the foundries. And, you know, a lot of that capability is transmitted through like software design tools and so on. Right. So the chip industry has been able to, the fabulous companies have been able to leverage those manufacturing processes by working closely with their foundry partners.

Okay. I think the question in semiconductors is really what would it? be like for the U.S. if it was unable to manufacture anything domestically. And I think here the COVID pandemic has really shined a light on that because on a lot simpler commodities, things like personal protective equipment, okay, or and 95 masks and things like that. You know, we didn't have the manufacturing capability. And the question would be, what is, If Asian manufacturing capability was suddenly turned off, I'll point out that an awful lot of the semiconductor manufacturing capability is in three science parks in Taiwan, which is a seismically active region. And there have been minor disruptions over the last few decades in terms of typhoons and earthquakes and so on.

But, you know, there's also the risk of geopolitical disruptions. Okay, so I think that has heightened the sensitivity about having manufacturing concentrated in such a small geographic area. It's a little different, I would say, than the know-how from manufacturing in a lot of other processes. But you do see it in semiconductors in terms of the tool manufacturers. Okay, and here American companies have a large presence in tool manufacturing, okay, and being close to the actual manufacturing facility is very important. And you see a lot of those toolmakers moving their development activities and manufacturing activities closer to the market, be it Singapore or Taiwan or Korea or even China. You can get the news whenever you want it with Bloomberg News

now. I'm Amy Morris. Moscow here to tell you about our new on-demand news report delivered right to your podcast feed. Bloomberg News Now is a short five-minute audio report on the day's top stories. Episodes are published throughout the day with the latest information and data to keep you informed. Yes, there are other products like this from a variety of news organizations, but they usually rerun their radio newscasts throughout the day. That's not what we do. We create customized episodes that can only be heard on Bloomberg News Now. And we don't wait an hour to publish breaking news. When news breaks, we'll have an

episode up in your podcast feed within minutes. So you're always getting the latest stories and developments. Get the reporting and the context from Bloomberg's 3,000 journalists and analysts we're all over the world. Listen to the latest from Bloomberg News Now on Apple, Spotify, or anywhere you listen. Can you explain in a little more detail the mechanism by which learning by doing actual know-how in the manufacturing feeds through back to innovation because, you know, it sounds good, this idea that like, okay, if you're actually in the process of sort of building chips, if you're not outsourcing it, that also helps the technological innovation and that this idea, this dream of, okay, you can just separate the design and the manufacturing and still be a world leader,

that it's problematic. How does that work in practice? Walk us through a little bit. why that link is important? So I think the key idea here is tacit knowledge. Okay. And tacit knowledge is that which you know, but you maybe haven't written down or haven't really explained to people, right? So there is a lot of know-how

on the shop floor that comes from kind of learning as I produce more product. Then the question is, how is that knowledge captured and how is it communicated? Oftentimes, you will see there's this concept called design for manufacture. In other words, I was in one factory once where the people assembling this piece of equipment. I said, you know, the guys who designed this, they haven't thought about, you know, assembly sequence, for example.

It's like, you know, you can't put that part there because I can't install it. But there's a lot of those kind of tricks of the trade that come from practice in manufacturing, which then feeds back into design that says, okay, when you design it, recognize that we have to do these things in manufacturing. And how might that change your design? So tacit knowledge is really the main concept. You know, that which I know, which I cannot have not or cannot explain necessarily. So we've sort of laid out the scene here and laid out the differences between the two business models and why semiconductors are important and also why the industry is different to a lot of other types of

manufacturing and design. What can actually be done about this? If we agree that Intel has followed, behind on semiconductors, that the U.S. should do more to boost its domestic capacity in order to have a more resilient supply chain and, you know, be able to withstand unexpected disruptions like the pandemic. What's the first step towards actually achieving that? Well, Tracy, I think this is a tough question. And a lot of people I've talked to agree on the problem, but, you know, I think there's really a searching for what kinds of steps would help. Now, what some people say is Intel needs to establish a foundry operation. Okay.

I think I agree with that because in manufacturing, and especially in manufacturing semiconductors, you need to have volume, right? Because through manufacturing lots of volumes, that's where you get your process learning. That's where you get your process improvement. that's where you improve your yields, okay, and find out what works and what doesn't work. Okay. As Intel's microprocessor volumes come under competitive assault from other designs where people go to, say, TSM or a foundry to manufacture it, Intel needs to keep its volumes up.

One way it could do that is it could establish a foundry operation that would make both its own chips and chips for other customers. That has its challenges, though, because Intel has its own way of doing things, which have served its own needs historically. And so what it needs to do then, it needs to learn how to be a service business that will service the needs of other customers,

including its competitors. And that's kind of a tough transition, because that means it has to give priority to its customer slash competitor needs. Okay. And it's competitors who might use it as a foundry, then we'll ask the question of why do I want to help Intel? Because they've been a tough competitor over time.

And if I give business to this foundry, am I helping Intel? Okay. And the other question as a potential U.S. Foundry customer I'm going to ask is like, are you going to give me the same level of technology and process capability that I can reliably get in Taiwan today? So there are a lot of barriers to making a foundry business like that work. But I would argue that Intel needs to do something like that in order to keep its own. manufacturing volumes up. You know, why Intel? Well, because Intel is probably the last best chance for the U.S. to have a leader in this technology. Let me just actually back up here because I'm still sort of struggling with a key question, which is when we talk about the stumbles at Intel or the decline, I mean, here is a company that has been integrated. It's connected, you know, sort of It has this lots of tacit knowledge and having manufactured its own chips for so long, unlike some of its competitors. What is your core diagnosis for what's wrong with the company?

And also, you know, recently, I think it was the hedge fund third point, came out with some proposals, the activist firm, to try to revive the company. But, A, how would you sort of summarize your diagnosis of what went wrong there? and then what do you make of some of the suggestions by them? Well, I think Intel needs to invest heavily on process leadership. Ten years ago, they were two generations ahead of TSM. Now, 10 years later, they're behind TSMC, right? So they need to invest heavily on that. I think Intel historically has

tried to be very self-sufficient in terms of using a lot of their own tools and a lot of their own methods. I mean, I think one thing that would be helpful would be to work more with others in the industry in broadening the funnel of ideas that come to them. Okay, because in some sense, trying to do it all, you know, yourself in this very challenging technological era may not be the most competitive path. If you look at companies like TSM, they have heavily leveraged partnerships with ASML, who's the lithography leader for DPUV. They have heavily leveraged organizations like IMEC in Europe, which is kind of AASM. a shared research collaborative.

And TSM is very good about working with a broad range of parties. And I think that would help Intel, right? And that means they have to kind of set aside some of their established ways of doing things and open themselves up to maybe broader approaches to the competitiveness. Okay. Now, on the third point question, okay, I think I agree with the need for Intel to really establish a foundry business because, as I said earlier, I think volume is very important for them. But just the simplistic idea of separating their design business, which is designing microprocessors and manufacturing, which is, which is manufacturing semiconductors,

taking them down the path that AMD went down before and many other companies went down before, that will create value for shareholders. Okay, but then you have to ask, how likely is that foundry going to be absent some other moves to ensure that it has customers and ensure that it is competitive, right? So at a simple level, I could see how it would create a lot of shareholder value.

But is that the right answer for the United States? So one thing I'm wondering is, you know, we're having this conversation now. And there is a proposal in Washington, as I mentioned, to do some of this. And we can talk about that in some more detail in just a bit. but why hasn't this been done before or why hasn't there been more of a concerted effort to boost America's strategic capability when it comes to a technology that people agree is increasingly important? I think we have in some sense been looking at a rosy picture of the semiconductor industry

reflected by the success of fabulous companies like Nvidia, like Broadcom, like Qualcomm, like Apple, by the way. And we're seeing it as well in companies like Amazon and Google who say, I don't need the manufacturing. But you see the valuations of these companies and you see the financial success. As long as Asia represents a secure manufacturing supply base, one could argue if they want to subsidize that or they want to do a lower value work, then we should let them. And there are people in Silicon Valley I've talked to who said, you know, there's a strong argument for, hey, if Asian countries want to subsidize manufacturing, we should let them,

we should take advantage of that. And then the only question is what happens if you get cut off from that supply? Now, I think we tend to believe in the U.S. that people wouldn't cut us off from supply. Okay. In the last couple of years, the U.S. has really weaponized supply chains against other countries, notably against companies like Huawei or ZTE or, you know, a lot more recently. Okay, so then the question is, do we open ourselves to that kind of quid pro quo behavior? So let's talk a little bit more about national policy. And you talk about the case for why it should be a DC priority. Now, just to for my recollection or knowledge, in the latest, there was the Foundry Act, which sought to encourage more domestic investment. Is that been passed now? Well, there was the American Foundries Act. Okay. And then there was another act. Those got. merged into the National Defense Authorization Act, which was passed, but the funding still has to be worked out.

Okay. And if you read the language in the NDAA, I forgot exactly how many pages it was, but it was around 700 something. And, you know, so I went and found all the language on the semiconductor stuff. And I would describe it as aspirational, right, understanding that we're going to direct commerce department to do these things. We're going to direct all these people to report and all that. The challenge right now, I mean, I think congressional staffs have good intents here, right? But the challenge right now is this microelectronics leadership in this segment, you know, that Intel is in. There is a dearth of what I would describe as good ideas, good strategic moves on how to handle this. I've been racking my brain on this recently, but I think it's a very tough problem that has been decades of the making.

I hope people don't kind of get disillusion on these things. It's like my mother always told me problems that were a long time in the making or a long time in solving. This is not a problem that gets solved in three years or five years. If you look at TSM, which was founded in the 1980s, okay, they've been working very, very hard for decades to get to where they are. And they're not about to lose this either. So what is, I mean, you say there's this dearth of ideas, but obviously, and we talked about this on the previous episode, there's sort of a limit to how much you can achieve just by throwing money at the problem. But of course, money has to be part of it. So in your view, what would be the sort of basics of the beginning of rebuilding that capability domestically, both in terms of how much investment would need from DC? but also investment trained in the right way such that it really fostered a sort of self-sustaining and more robust capabilities.

So one of the ideas, Joe, that I have been floating is there are an awful lot of proposals that focus on the supply side. Like let's get people to build fabs in the U.S. You know, let's get people to do more R&D in the U.S. Okay. And that's all well and good. but if you build a new foundry in the U.S. and nobody shops there, it's not sustainable.

So what I've been saying is instead of only focusing on the supply side, we should also think about the demand side. And the reason I highlight that is if you go back to the 1960s, and I'm dating myself now, But if you go back to the 1960s, at the birth of the integrated circuit and the birth of the semiconductor industry in the United States, DoD and NASA bought 60% of all the integrated circuits that were made in the world. And that was something that helped companies like Texas Instruments, who was one of the pioneers. Okay. And Fairchild, which gave birth to Intel and others. Okay, but it was that demand that really helped foster the growth of the industry. It was a customer for kind of those early products

who worked with the suppliers to improve. Okay. So I've been saying it's like, let's focus on demand as well. Now, one of the things that we do have in the United States is we do have demand for a lot of high-end semiconductors. If you look at all these hyperscale data centers along the Columbia River or around the D.C. area or spaced across the country, let's take some of these big hyperscale data centers from, you know, Google or Microsoft or Amazon along the Columbia River. Well, where did those zion microprocessors come from? Well, probably Hillsborough. Oregon where Intel has a fab. That's one of their leading fabs in the U.S. And then what happened is, well, well, those chips then went to the finished wafers with the chips on them that were then sent to either Shandu, China or Vietnam or Malaysia,

where they were tested and cut apart into individual dye, and then they were put into packages. And then they would go to a distribution center somewhere in, Asia and then go to a server board manufacturer in China, probably. And then they'd ship them back to the Columbia River Valley, you know, 100 miles away from where the chips originated. And it's like, now, why do they take that round trip? Well, they take that round trip because in the early days of the chip industry, when you

had to cut these chips apart with a diamond saw and then put them into little packages, you'd have rows and rows of workers looking at. through microscopes, wire bonding those dye into packages. Okay. And it was very labor intensive. And semiconductor chips, you know, they have a very high value density. So it's no problem to air freight them over to Asia, have them packaged them and have them come back. Okay.

These days, all that stuff is automated. So there is no reason for it to go to Asia, except over the last four or five decades, Asia has developed the capability for chip packaging and automation and so on. So one of the things I would focus on is I would focus on let's bring kind of chip packaging back to the U.S. We still have companies in the U.S. like AMCOR, who are leaders in that technology, you know, and let's help them to do packaging in the U.S. and just kind of move a key part of that value chain where we have some of that demand.

You know, the other reason I think packaging, packaging is important is we're seeing the end of Moore's law, where it is becoming harder and harder to push the frontier on these advanced processes. Okay, but an awful lot of the horsepower you need in these chips come from lots of different processes and maybe some not so advanced. And you can get a lot of the benefit by going to what people call 3D packaging, where I'm going to put multiple chips into a single. package and then kind of bond them together with what are called Vias or, you know, basically I'm going to use multiple chips in one package because if I can cut the distance between them and pack them more densely, I'm going to get a lot of the benefits from Moore's law without having to go to the most advanced processes, right? So by focusing on parts of the value

chain that we can win over the shorter term, we're going to, get some of that manufacturing capability back, while you are working to restore process leadership, particularly looking at next generation, like what comes after all the technologies that we know will take us to three nanometer today with Moore's law, that has gotten very expensive. Investing in new process technologies, new technologies that, that will take U.S. and global semiconductor production to kind of the next era, that would be a good investment as well, right? So, I mean, I think there are things that we can do that will take investment, okay? But those also play to some of the strengths that this country has.

On April 4, 2023, around 2 in the morning, a man was found stabbed multiple times, on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. I want to go back to something you said earlier about the idea that the U.S. has been weaponizing supply chains, particularly in China and with its actions against Huawei.

We've seen Huawei not necessarily struggle to source semiconductors, but they've had to stockpile a bunch of them, and they're sort of starting to run through their supply because they can't get them in the way that they used to. There's a perception within China, or at least there's one strand of thinking within China, that this isn't necessarily the worst thing for the country, because it's going to end up accelerating their own domestic manufacturing ability.

So they're going to sort of rush through development, and the U.S. has in some ways done them a favor by sort of sparking a big fire under them to get this done. Is that a credible argument in your view? Or I guess in your view, what are the downsides and the benefits to targeting external manufacturing or targeting China in this way? Well, in some sense, what China is doing is they're on an import substitution drive now in response to this weaponization. A lot of economies have used import substitution very successfully to kind of bootstrap themselves

up. Taiwan, notably one of them, right? Korea, Japan, a lot of Asian nations have used that. Okay, will China be able to get there? Right. You will hear a lot of skeptics who say, well, China won't be able to get there fast because they still need technology from the West. Okay. And it may take them a while.

But, you know, I have friends in China who said, you know, you guys in the U.S., you're a young country. You're less than 300 years old. That's like, you know, one dynasty, right? And it's like they have a very different scale on time. Okay. And I would point out that the Europeans in the late 60s and early 70s were looking at the commercial hegemony and commercial aircraft in the U.S. And they were concerned about it.

They put together Airbus as a consortium, right? Now, Airbus had lots of political issues early on. Okay, but here 50 years later, Airbus arguably is the global leader. Right. Is 50 years a long time for China? Well, it is in the context of their five-year plans. Okay. But I wouldn't underestimate, you know, their ability to get there.

Now, I personally don't think that's a good idea, right? because I still believe that the world is better off if we trade with each other. But that also requires some notion of balance, right? I mean, what you're hearing out of China now is this so-called dual circulation strategy, which is they're going to be self-sufficient as much as possible, but they still want to be a source of manufacturers, for the world. That's something I would question, because if you don't want to buy from anybody else

and you want them to buy from you, I don't see how that's going to play very well over the long term. So I don't think you can have it both ways. But I wouldn't underestimate China's ability to get there. There are still some people, I think, who believe, well, China is just about copycats and so on. There are some very innovative people in China. They're some very innovative companies. So I would not underestimate their capabilities over the long term.

Might they spend much more money getting to parity in semiconductors than they would have if they took a more global approach, probably? Does that matter to them? Maybe not. I want to ask a slightly shift the conversation just slightly to another sort of timely topic, but it's related to this. So we're talking about these sort of global hubs of technology know-how. Obviously, Taiwan is a Q-on, also Singapore, Silicon Valley, has Silicon right in its name, the home of this industry. And lately, and this is sort of maybe more software outside of technology, but we've heard especially, like it really is accelerated in the last several months, people talking about moving to new tech hubs or creating tech hubs. So maybe people are annoyed with politics or the government in California.

And they're like, you know, we're all going to let's move to Austin, Texas, or let's move to Miami. And we'll just sort of set up a new, a new tech hub in this area and get away from, I don't know, I don't know, California regulation or whatever it is. What is history say about sort of what will make a tech hub succeed or not? And is this the kind of thing that can sort of be done if, you know, have a few venture capital leaders. They're all moving to Miami. We're going to set up the next Silicon Valley here. Does it work that way or does it sort of need to be more emergent and organic?

Well, I think one of the benefits of physical proximity, has always been the movement of people. And this goes back to what we were talking about earlier in terms of tacit knowledge. If you think about how tacit knowledge moves, it generally moves in the head of heads of people. I worked for IBM for 14 years. Okay.

And every disk drive startup in Silicon Valley was started by people who came out of the IBM research operations in San Jose, California, where they did disc drive development. every single one, right? So what happened is they had different ideas or they wanted to pursue different ideas, you know, Phineas Connor, Ellen Shugart, all those guys. And so that's one of the things that makes hubs productive, right, which is this spreading of ideas, this movement of ideas, movement of people. Now, the phenomenon you described, I mean, I lived in Austin,

in Texas for a while. And I like the no-income tax and the much lower cost of living there compared to California. I lived in California as well for a while. Those movements, I would argue, are really about being able to attract talent and lifestyle when you're unconstrained by having to be in a physical office. I don't know if Zoom and Microsoft Teams or virtual whatever, what the impact will be on that movement of tested knowledge between organizations. And if what we have is a new era of that kind of movement,

maybe we'll see some more dispersion. Okay. But I would argue one of the things that, really has helped Silicon Valley over time, has been the people movement between jobs, to the extent that they're willing to geographically relocate, which is something that has historically been very strong in the U.S., but has diminished over the last decade. It could work, right? But we'll see on that. Professor Shee, this is a great conversation. Before you go, I want to sort of, I was reading one

of your articles, and I thought this was super interesting and maybe super helpful for listeners, and that is the brutal math, the brutal mathematics of yields. And we've talked about it on a couple of previous episodes, but this idea you sort of enter it, you do this process. As you mentioned, a fab could cost $20 billion here, making this huge investment, extraordinarily complex, and then to hit your margins and to actually make a return, you need to gary. that if you produce that on the way to creating a finished product, that it actually works or that most of them actually work so that you're not throwing away a bunch or whatever, and that even if you have, say, like, 99% effectiveness on various steps, that that might

not be enough. Can you talk us through how yields are calculated and why the requirements for excellence are so demanding? Yes. Okay. So in semiconductors, the thing that makes this one of the most challenging manufacturing processes is that it has so many steps. Okay. And if you just do the math of saying, let's say I had 700 steps, okay, and let's say I have 99% yield for the first step. Okay, if I have 99% yield for the first step and then 99 for the second and 99 for the third, after each step, I'm getting fewer and fewer. Okay. By the time I get to 700 steps, I will get no good output. Okay. Therefore, what that means is I actually have to have my yields way up there.

That means I have to have 99.99 more than that, right? If I have 99.99% yield at each step. Most of us think, oh, 99.99% that's pretty good. Right. 99.99% times 700 steps will give me 50% yield out the other end. Okay, now let's think about the economics of a $20 billion fab that might produce, let's say, 20,000 wafers a month. Okay, and on those 20,000 wafers, if I have to throw out half of them, you can see the economic cost of that is unbelievable, right? So when you have an expensive fad, you need everything coming out of it to be good. And when you have a lot of steps, that means you've got to be almost flawless on every single step.

And that's why this process, this manufacturing process is so hard. I love that. That's a great place to stop. And I think a sort of perfectly clear explanation of how this industry is just so brutal. Professor She, thank you so much for coming on. Well, Joe and Tracy, thank you for having you. Thank you. I really like that conversation.

You know, obviously that last answer that he gave about yield math, to me, it's sort of like, what I really liked about his perspective overall. I mean, obviously that was super interesting is this sort of way business and technological demands sort of intersected. I feel like he just had like super clear and also like concrete examples of how that really works. Because we could talk and sort of like vague ideas about, oh, it's important to have technological know-how or it's important to, you know, be able to trade. with advanced partners, but Professor Shee really had a very concrete way of describing what's actually going on. Yeah, that was a good example. The other thing that really comes through in those discussions, I mean, it's coming

through in every episode that we're doing on semiconductors, but it's sort of the uniqueness of the industry, right? It's sort of like an exceptional industry in many ways. And again, that yields problem that Professor Sheed just described, as well as the startup up costs for creating a foundry, the idea that you used to be able to do it for a few million dollars, like back in the 70s or 80s maybe, and now it costs upwards of 20 billion. That kind of goes to show just how complex and advanced the technology has become, but it also, like, as it advances, of course, it becomes harder and harder to get in that game. And that's

where the question of public support comes in. Yeah, because it's interesting, like, so obviously, obviously, there are companies in the U.S., and you mentioned some of them, you know, like AMD or NVIDIA, they've just done like phenomenally well that don't have their own fab. So you don't have to have that integration to work. And from a pure shareholder perspective, those companies have like shot the lights out, which then raises this question like for the government. It's like, okay, well, maybe you could just have this sort of total separation. and some people in a design shop somewhere in the U.S. or California or Austin, Texas,

and then you ship the design to somewhere in Taiwan, and then they produce it. And maybe that model is really great for shareholders, but it does raise the question of, A, how does that serve sort of strategic public needs? Right. But then B, is that sustainable, the sort of the long-term separation of designing and building? Because, you know, a big part of this whole core thesis is that the two. really do go together. Right. It's the short-term goals of shareholders who are interested in returns versus the long-term value of having a strategic industry that's competitive globally,

I guess. Yeah, another big macro theme to add to this conversation. You know, another thing that, again, is sort of thinking about how this fits in our other themes. His whole, the whole discussion of like demand side policy for tech. I thought it's super interesting because, you know, I think the clear, you know, it's like, all right, what do we need to do to create a more robust manufacturing sector here for high tech? And I think a lot of people would say, well, invest in R&D, create tax credits for building foundries here and so forth. But this idea that there is also an important demand side component, and of course, go back to the history of Silicon Valley, the idea that the Department of Defense and NASA as being,

NASA is being customers of last resort or first resort for a lot of these things was a big part of it. It's also interesting, you know, bringing into like the vaccine discussion, a lot of what has allowed the fast pace of vaccine development has been this sort of guaranteed buyer at the end, that the government pre-committed to buy all these shots from Pfizer, Moderna, Johnson, and Johnson, and if they were successful, and the ability of the government, on the government alone, to de-risk the process by promising to buy the end output. That's a super interesting point. Right.

And I mean, we've talked about the history of Silicon Valley before, and I think there's still a tendency to think about tech as this very entrepreneurial industry that, you know, people started in their garage and they're smart and they sort of get lucky and they're able to do it. it. And one of the things people don't really realize is the degree to which having that government backstop has actually helped the tech industry. I do wonder, based off the experience of the vaccine development, whether that starts to change. And again, as we've been discussing for most of 2020, whether there is going to be a larger recognized role for the government in sparking both demand and supply. Yeah. One of our previous guests, actually, I think we've had them on a couple of times, although maybe you weren't there for one of them, but Bill Janeway, the venture capitalist. Yeah.

He talks a lot about exactly this point and the idea of like government underwriting, guaranteed government demand and how important that is for the commercialization of innovation. And also like, you know, with an eye towards like puncturing a little bit of Silicon Valley's own myth making, that if the government just left tech alone, that they would do all these great things and, you know, so forth. than how much of that, all the industries that really take off, I'd say green technology or clean technology is another area in which the government has had a big role or could play a big role, both in terms of the supply side investing in R&D, but also the demand side in being a customer

for ventures that take a lot of risk, take a lot of money and therefore high risk. Yeah. Well, it seems like we might have a new example in the form of a U.S. semiconductor. is not too far from now, maybe, to add to that pile. Yeah, we'll see. Okay, shall we leave it there? Yeah, let's leave it there. This has been another episode of the Oddbots podcast. I'm Tracy Alloway.

You can follow me on Twitter at Tracy Alloway. And I'm Jill Wisenthal. You can follow me on Twitter at the stalwart. Follow our guest, Professor Willie Shee. He's on Twitter at Willie Shee underscore at HBS. Follow our producer Laura Carlson. She's at Laura M. Carlson. Follow the Bloomberg head of podcast, Francesca Levy, at Francesca today.

And check out all of our podcasts under the handle at podcasts. Thanks for listening. I'm June Grasso, inviting you to join me for the Bloomberg Law podcast. Every weekday, we help you make sense of the legal stories that shape the nation and the world. Listen for complete analysis of the biggest court cases, the latest actions from Congress and regulators. and the legal moves driving the markets. From corporate law to constitutional law and from state courts to the Supreme Court.

At Bloomberg Law, we go beyond the day's headlines. We speak with top attorneys, judges, scholars, and policy experts to break down what the rulings really mean. We do this every weekday, then bring you the best conversations in our daily podcast. Search for Bloomberg Law on YouTube, Apple, Spotify, or anywhere else you listen. On the East Coast, listen as you start your day. And on the West Coast, catch up in the evening.

That's the Bloomberg Law Podcast with me, June Grosso. Subscribe today wherever you get your podcasts.