Morris Chang, founder of TSMC, expressed his support for US efforts to slow China's progress in chip production at a recent semiconductor forum hosted by CommonWealth magazine. However, Chang also raised concerns that US onshoring policies could lead to the price of chips increasing and warned that "globalization, at least in the chips sector, is dead."
Chang and historian Chris Miller discussed the past, present, and future of the semiconductor industry, which is at the heart of the global economy. Miller argued that we have overlooked the role of semiconductor companies in creating the modern economy and called Chang the most underestimated businessperson of the last hundred years.
Taiwan plays a critical role in the semiconductor manufacturing supply chain, with 90% of the world's most advanced processor chips produced in the country. Chang's innovative foundry business model, which enables the design and fabrication stages of semiconductor production to belong to separate companies, transformed the entire semiconductor industry and put Taiwan at the center of the map.
TSMC's success is attributed to its collaboration with the world's leading designers, tool producers, and customers, making it challenging for other companies and countries to compete with TSMC and Taiwan. However, both Chang and Miller agreed that there will be increased bifurcation in the supply chain between China and other countries, especially at the leading edge of the industry.
Chang supported the US industrial policy aimed at slowing down China's advances in chip production but warned that the US move towards onshoring and friendshoring has placed free trade in the sector in danger and could lead to the price of chips increasing significantly. Chang also highlighted the tension between risk management and manufacturing concentration, as too much concentration could have significant implications for the world economy.
The forum emphasized the importance of protecting TSMC and the Taiwanese semiconductor industry, with Vice President William Lai Ching-te calling on global leaders to work together to maintain peace and security, which is in the common interest of Taiwan, TSMC, and the world. In the face of increasing global geopolitical tension, the world cannot afford to underestimate TSMC and the Taiwanese semiconductor industry as chips will determine the future of the world.
The following are the key points discussed during the conversation between Morris Chang and Chris Miller on March 16th:
Morris Chang: About the book, I think the best compliment that one can pay to a book is to say, I wish I had written it. And that is the compliment. That is what I want to say about the book. I wish I had written it. Now, of course, I think Chris wrote it better than I could have written it. He has the historian's impartial position, whereas I was a combatant in the war.
The chip war is really a series of wars over the last 60-some years.
The semiconductor industry actually started with transforming the people, very quickly. And it started in 1952. Now, Chris's book basically started with the invention of an integrated circuit, which was in 1958. But when I joined the industry in 1955, it was still only transistors.
From the very beginning, it was an American-dominated industry. The transistor was invented by Americans in the Bell Lab. And the integrated circuit was invented by Americans subsequently in 1958. Now, the chip war started with wars amongst American companies. And I was a combatant right then. Texas Instruments versus Intel wasn't even there then, versus companies like Fairchild, Motorola, and so on. And then a few years later, we suddenly found that there were Japanese. So the war changed to Americans versus Japanese, although the war between American companies still continued. And well, and then later, of course, there were South Koreans.
Reading the book was a nostalgic journey. I lived through the wars. And I know everyone in his cast of characters. My impression is that I am almost the last Mohican.
There might be a few minor corrections. For instance, Chris has probably over-emphasized the government’s role in TSMC's formation. The Taiwan government actually was just an investor. And frankly, it was not a very willing investor. But there was one guy, a key guy, KT Lee. If it weren't for KT Lee, the government wouldn't have invested.
After the government invested, as soon as we went public in the stock market in 1994, it was seven years after we formed, the government started selling their stock immediately. The government had 48 percent. And they really couldn't sell it fast enough. There was another form of support that TSMC got from the government. And that was the original group of TSMC. It came from the IC project in the Industrial Technology Research Institute.
Chris Miller: When you founded TSMC and looked 10 or 20 years into the future, what did you anticipate the result would look like at that point?
Chang: We founded TSMC in 1987. The first two or three years were tough, but from 1991 on, we were growing very fast. More than 50% compounded annual growth rate. All through the 90s, the compounded annual growth rate was 90%.
When we first founded TSMC, I just wanted the company to survive. I just didn't want to let down the government who had basically funded half of it. But when we went into the 90s, it was a case of continually rising expectations. I certainly did not expect TSMC to be as big, as important as it was all the way in the 90s, or even into 2010. But from 2010 on, I really expected TSMC to be what it is now.
Gordon Moore was once asked, did you expect Intel to be so successful? And his answer was, nah.
Miller: Most people are totally unaware that their lives, their businesses, their work depends fundamentally on semiconductors. And they're even more unaware of the extent to which just a couple of companies produce the chips that they rely on. We'd all heard of Steve Jobs because we associated him with the iPhone.
But we don’t fully appreciate the extent to which the chips inside of all of those phones come from just a couple of companies whose capabilities have been in many ways equally as impressive as the devices that they empower. We've underestimated the role of semiconductor companies in creating the modern tech sector, the modern economy. And there are a small number of individuals like Morris who have played a critical role in that regard.
Morris, you founded the most important company in the industry. And so I'd be interested to hear your perspective of what the world looked like in the 1950s. Could you have imagined it would arrive at a point like this, or what did you expect from that position?
Chang: In 1976, I gave my presentation about the foundry business model to Texas Instruments. It was just like a puzzle. You have to fit the pieces together. The presentation I made in the 1776 Texas Instruments Strategic Planning Conference was, in my mind, perhaps the initial piece of the puzzle.
I believed that semiconductors will be pervasive.
Actually, the word pervasive was used by Pat Haggerty at that time. Texas Instruments or anybody couldn't possibly design all the products in semiconductors. Since it's going to be pervasive, then washing machines will use them, automobiles will use them, let alone computers from PC companies and so on. And they may not want to have fabs manufacturing, and therefore we can sell the chips according to their own designs, we can make them.
But that idea was premature. Remember, design at that time was a big thing. And in Chris's book, he described how Federico Faggin labored almost for a whole year designing the first microprocessor. We were still waiting for the arrival of Carver Mead and Lynn Conway. They were the second piece of the puzzle.
Of course, it was one transistor. If we talk about logic, we didn't even have 4,000 transistors on the same chip. So Moore's Law was still in its youth. The chips were not complex enough yet. And fabs were still relatively inexpensive. Back then it was probably $20 million. Right now, we're talking about $10 billion.
Yinchuen Wu: Do you think Intel will succeed in building a foundry business that can compete with TSMC and make the US self-sufficient in chip manufacturing?
Miller: If there's one thing that I've learned about the semiconductor industry, it's that it's impossible to be self-sufficient. If you look at the map up behind us, you'll see multiple critical countries that are involved in the industry. I think that any country that strives for self-sufficiency is likely to find itself spending a lot of money and getting less good technology as a result, which is why I don't think there are very many countries in the world that are actually seeking self-sufficiency.
In terms of Intel, I think one of the conclusions from the comments we've just heard is the challenges in setting up a foundry business model. We've seen a number of companies try to do so and compete with TSMC with varying degrees of success, but certainly TSMC has vast scale and advanced technologies that make it difficult to compete with.
We're going to see Intel try to build out their foundry business over the next several years. It certainly does seem like building a foundry business from scratch is not easy when there are already multiple successful foundry businesses in the world.
What's critical to remember about all of the firms we're discussing is that they're all firms with international footprints, multiple manufacturing facilities in different countries, with customers in different countries. So, in some ways, I think the nationality of a firm's headquarters is important, but it's also important not to overemphasize that reality because the supply chains are so complex and international.
Morris, do you want to respond?
Chang: As far as Intel being a foundry is concerned, I think Jensen Huang of Nvidia said it best. He said that TSMC has learned to dance with 400 partners. Intel has always danced alone. I really like your comments that TSMC's success is not by doing things on their own, but by embedding themselves very deeply in the international supply chain.
Buying machine tools from the US and Netherlands, and materials and chemicals from Japan. Taiwan's success story is not about Taiwan doing it alone, but a story of integration of the best in the world.
I think there's a basic assumption we need to make that there's no war in the Taiwan Strait. There's no war between the United States and China. After making the assumption, how do you see what the supply chain is going to be like in five years?
Miller: From my perspective, it's clear that there will be more bifurcation in the supply chain, especially at the leading edge between China and other countries. You see U.S. regulations, and now Japanese and Dutch regulations. Now bifurcation will probably be much less significant because there are no regulatory barriers right now for cooperation between firms or sales of equipment at the lagging edge. I think that's one key shift we're already seeing and will continue to see.
The second is that in many different countries, there's much more emphasis from their political leadership on investing in domestic capacity for a variety of reasons, for economic security concerns, for economic development concerns. You see this in all of the major companies and countries, and even from some countries that don't have major chip industries right now. Whether it's the United States, Europe, Japan, or India, they are all trying to become bigger players in the chip industry, this is a global trend. We're already seeing governments put really substantial sums of money behind this relative to what they've previously done.
I would hypothesize that if we were all to reconvene in five years' time and look at the global landscape, it would look different from today, but it wouldn't look radically different.
Chang: I agree that there will be bifurcation of the supply chain. I'm of course more interested in the American side of the supply chain. The United States I think has started to practice industrial policy on chips. And part of the industrial policy is to slow down China's progress in chips. I might say I support it.
Although in my opinion, in manufacturing technology, China is at least five or six years behind Taiwan. I look at the most advanced chip they are making, and they are making it with difficulty. But that chip, TSMC was making five or six years ago with ease. But still, I certainly support that part of American industrial policy, to slow down China's progress.
Now another issue is the so-called reshoring or friendshoring, which does not include Taiwan. In fact, Commerce Secretary Gina Raimondo has said repeatedly that Taiwan is a very dangerous place. America cannot rely on Taiwan for chips.
That is Taiwan's dilemma. The United States already has 39 percent of the world’s semiconductor equipment, manufacturing, design, and intellectual property capacity, with 11% of the world’s manufacturing. So what is their goal with the CHIPS act? To go back to 30, 40 percent? Or is their goal just to maintain the supply essential for national security? If it's just for national security, it doesn't have to be very high.
Miller: There's been a greater sense of risk when looking at geopolitics, intensified by the Russia-Ukraine war. There's been greater thinking about business continuity planning, about worst-case scenarios, and that's coincided with more concentration in the chip industry. And risk planning and concentration are two things that point in very different directions.
Right now the industry is at a difficult balance that it needs to strike, because there are extraordinary efficiencies that come from concentration. The economies of scale in chip making is dramatic, and we've seen the evidence of that. But there's also concern that if there's too much concentration and something goes wrong, the implications for the world economy could be quite large.
There are many ways things could go wrong. If you look at where the world has decided to put its advanced chip-making capabilities, it's largely in seismically active zones like Silicon Valley and in Japan and on this island. So you can worry about natural disasters, but obviously the geopolitical backdrop is seen in most countries to be more dangerous today than it was five years ago. I think that's what's driving some of these concerns about having more diversification of the manufacturing footprint.
Chang: One has to realize that the pervasiveness, the ubiquity of semiconductors is primarily due to its ever cheaper cost. Back in the 50s, we were talking about transistors back then. One transistor was $2 or $3. Now, a cell phone carries 10 or 20 billion transistors, which means a transistor is now one nanodollar, or one billion times cheaper.
Now, if you give up the competitive advantages of Taiwan and move it to the United States, move the manufacturing to the United States, the first thing that's going to happen, and in fact it's already happening, is that the cost is going to go up. Our estimates of the U.S. cost would be 50% higher than Taiwan. That was an underestimate. Maybe it's double. That's going to do something. It has some impact on the further expanding ubiquity of the chips. That's one thing that I see.
Is globalization dead?
Wu: Morris, you said in Phoenix last year, “globalization and free trade are almost dead and unlikely to come back.” Do you care to elaborate on that and how would that impact the future of the chip industry?
Chang: There's no question in my mind that at least in the chip sector, globalization is dead. Just look at the way China has been embattled. Free trade is not quite that dead, but it's in danger.
Wu: Do you think that will help the development of the chip industry in the future?
Chang: Of course not. The first thing that will happen is that the cost will go up. And when the cost goes up, the pervasiveness of chips will either stop or slow down considerably. So we're in a different game.
The reason that Chris started to write the book was because he realized that chips were so pervasive. We needed chips everywhere. Now he has finished writing the book and the game has changed.
Miller: I would say a couple of things. We use the word “globalization” to define the chip industry, but the correct word is actually “internationalization” because the chip industry sells its products globally, but in fact is dominated by just a couple of countries, Taiwan, Japan, US, Netherlands, Korea, and most countries play no role in the production of advanced semiconductors, which is interesting. They buy them, but they don't play a role in producing them.
So in some ways, the chip industry is very globalized. In other ways, it's actually extraordinarily concentrated in a couple of countries. And so I think what we are certainly going to see is some degree of shifts in market share between those countries, though probably not always in the way that governments intend or would like because the success of business models is limited.
Chang: It's indeed concentrating in a few countries because those few countries have competitive advantages in what they are doing. Actually, I think the US has an enormous competitive advantage in design. On the other hand, Taiwan, Japan, and South Korea have a competitive advantage in manufacturing. This is another subject that will require more time than we have today.
It's mainly the people and the work culture in the country. In chips, we run 24 hours because the equipment is so expensive we can't afford to stop running even for a few hours. When a piece of equipment broke down at one o'clock in the morning in the US, it will be fixed by next morning at nine. In Taiwan, however, it will be fixed at 2 a.m. That's a work culture competitive advantage. Taiwan has very, very professional, hardworking and skilled technicians.
Technicians, particularly repair and maintenance technicians, are very important. I think the year was 1981, all of a sudden TI discovered that the new fab that had just started in Japan had twice the yield of the old fab in Houston.
It is all because of the people.
In Japan, we have graduates from trade schools that are very willing to be trained to be precision equipment manufacturers. As a result, the equipment stays up 90 percent or higher. In the US, it's only 60 percent.