Thread by Staś Butler | 貝德旭

twitter.com
1 Recommender
1 Mention
Recommend
Post
Save
Complete
Collect
Thread
I'm at @CWM_en's Taiwan semiconductor event featuring "Chip War" author @crmiller1 & TSMC founder Morris Chang, with opening speech by Taiwan VP (and potential future president) @ChingteLai. Will try and capture some of what passes here!
Lai: Semiconductors are not about one technology or one industry, we're talking about why a small chip can drive changes in geopolitics. Why they're the holy grail for so many countries around the world.
Lai: Morris Chang is a living dictionary and history book for the semiconductor industry.
Lai: Very proud of what TSMC has achieved. It is an asset to the world. But TSMC needs support from Taiwan and the rest of the world because of the semiconductor supply chain's distribution.
Lai: Need to continue to provide support for TSMC and protect to ensure it can continue to grow. Because TSMC determines the development of the world, the shape of our lives, climate change.
Lai: Net-zero by 2050 unachievable without TSMC.
Lai: TSMC shall continue to stay where it was raised, Taiwan, and continue to grow here. This is the government's commitment to TSMC, the country's economic development, and the world.
Lai: Gov't needs to improve environment in Taiwan for big enterprises to help drive Taiwan forward. Democracy is also essential, just like sunshine, air and water. Without it, TSMC cannot be sustained.
Lai: Taiwan will continue to strengthen its defense to maintain stability in the Indo-Pacific region. But that's a responsibility Taiwan shares with the world, and with China. A war with China would not hurt just Taiwan, but the whole world. War would have no winner.
Lai: Just like with Russia and Ukraine -- invaders also face many challenges.
Lai: If we can deter China, we will be able to do the world a favor. Peace is in everyone's best interests.
CW publisher points out that Chris Miller is the same age as TSMC (!)
Miller: What I can add is from the perspective of a historian. My argument is that you cannot understand the modern world without putting semiconductors at the center of your analysis, and TSMC at the center of that.
Miller: Chip industry emerged from Cold War arms race, with the U.S. miniaturizing computers to put in the nose cones of missiles.
Miller: Nowhere in the world economy has seen such extraordinary progress, doubling every two years for half a century. The same for aircraft travel would mean traveling near the speed of light.
Miller: The success of the companies that have defined the industry is as much down to technological capability as it is down to business capability.
Miller: If there was a Nobel Prize for business, several people in this room would be recipients.
Miller: The fact that we're having this conversation in Taiwan is a surprise. If you'd asked someone in 1960/1970 which country would have the most advanced technology, few would have guessed Taiwan.
Miller: Technological advances, business acumen but also visionary product design are all essential to industry's advances.
Miller: Every decade or so there is a new use case for semiconductors that produces a vast new source of demand: defense, then corporate computers, pocket calculators, PCs, smartphones, now AI.
Miller: 90% of the world's most advanced chips made in Taiwan. More than 1/3 of new computing power too.
Miller: In 1976, Texas Instruments was already considering the origins of the foundry business model that put Taiwan at the center of the map.
Miller: Foundry model is critical to why TSMC is the largest and most advanced maker of chips in the world.
Miller: So if it was Texas' idea, how did this end up in Taiwan? U.S. was a victim of its own success. Silicon Valley companies were profitable and set in their ways.
Miller: Visionary government figures partly behind Taiwan's success. If you look at the development of Taiwan's industry, you see very strategic government thinking for creating an environment for growth.
Miller: Almost impossible for anyone living anywhere in the world to go without touching a product produced partly by TSMC.
Miller: Semiconductor shortages of the previous years have illustrated that they're not about phones and PCs, they're about everything.
Now to the forum part, involving Morris Chang. Wu Yin-chuen as moderator.
Wu jokes that when she first interviewed Morris Chang decades ago, she ran with a headline that TSMC was a 10-billion-dollar gamble.
Wu describes Chang as "the most underestimated businessman of the past 100 years."
Chang: The best compliment that one can pay to a book is to say I had written it, and that is the compliment that I would like to give [Chris Miller's] book.
Chang: I was a combatant in the war (Miller's "Chip War").
Chang: When I joined the industry, it was 1955 and the industry was still just transistors. It was U.S.-dominated.
Chang: Chip war began among American companies. And then a few years later, we suddenly found the Japanese [were involved]. It became U.S. vs. Japan. Although the intra-American war still continued.
Chang: Reading Miller's book was a nostalgic journey. I lived through the war, or wars, and I know everyone in his cast of characters. Almost half of them are dead now. Notes that this is not a precise calculation...
Chang: One minor correction -- Chris probably overemphasizes the government role in TSMC's formation. Gov't was just an investor, and not a very willing one. But one key guy, K. T. Lee, believed in us.
Chang: Gov't started selling their 48% stock in 1994 as soon as TSMC went public. "Couldn't sell it fast enough."
Chang: But gov't helped form "nuclear group" of TSMC that came from ITRI. About 120 people, including 40-50 operators.
Chang: K. T. Lee was the only one who believed in me. It was mainly his support that resulted in the government being willing to invest in 48%. It wasn't 10 billion, it was a much smaller number.
Chang: Philips was the other investor we were told to get on board to get others interested. 28%, they invested. Gov't said Philips can go and help stump up the rest of the money, about 25%. Even Philips couldn't sell TSMC stock fast enough. To help, TSMC went public in New York.
Chang: When gov't sold down to 6%, I asked gov't to stop selling. That's what they have now: 6%.
Miller: When you founded TSMC and looked 10 years into the future, what did you anticipate?
Chang: First 2-3 years. 1991 on, growth was fast. Over 50% compounded growth over 10 years. But at first, I just wanted the company to survive. Didn't want to let down the gov't.
Chang: Expectations rose afterwards. I certainly did not expect TSMC to be as big or important as it became in the '90s and '00s. But from 2010, I really expected TSMC to be what it is now.
Chang: Gordon Moore was asked the same question and said "nah." I've answered you in a slightly more full way.
Miller: I was struck during research by the extent to which people are unaware their lives, business, work depends fundamentally on semiconductors, and a couple of companies that make them. We all know Steve Jobs, but we don't appreciate the chips inside these devices.
Miller: Your life tracks onto the history of the chip industry. What did the world look like in the 1950s when you were looking forward?
Chang: Foundry business model was like a puzzle for children. You have to fit the pieces together. My 1976 presentation was the initial piece.
Chang: I happened to chance upon the insight that semiconductors would be "pervasive," a word used by Pat Haggerty. I thought that was a bad word: sounds too much like "pervert." I preferred "ubiquitous."
Chang: I realized that chips would be everywhere, and the producers may not all want to have fabs. But it was premature. Design was the big thing. We were waiting for the 2nd piece: Mead and Conway. '76 was talking about 4k DRAM. Moore's Law was still in youth if not infancy.
Chang: The almost-last piece of the puzzle was Gordon Temple. Nov. 1984 I met him. He said I want to start a company with US$50M from you. I asked "do you have a business plan?" He said no, it's all in my head. Get back to you in 2 weeks.
Chang: Time passed, no business plan. I called, he said "Morris, I don't need you anymore. I only need US$5M now. I'm not going to build a fab. I'm going to subcontract that to Japan and maybe IDM." That was almost the last piece. Still needed financing and a place -- Taiwan.
Miller: Impossible to be self-sufficient in the semiconductor industry. Any country that strives for self-sufficiency is likely to spend a lot of money and get less good technology. That's why it's rare for countries, and companies don't welcome it.
Miller: Setting up a foundry has real challenges. TSMC has vast scale that makes it difficult to compete with. Intel is trying to build that up, but I'm a historian, so I'll refrain from predicting the future. But certainly does seem like it's not an easy task.
Miller: Will Intel make U.S. self-sufficient? All firms we discuss have international footprints. Nationality of a firm's HQ is important, but not worth over-emphasizing because of complex, international supply chains.
Chang: Jensen Huang of Nvidia said it best -- TSMC has learned to dance with 400 partners (customers). Intel has always danced alone.
Chang: How do you see chip supply chain in 5 years? Basic assumption is no war in Taiwan Strait, or btwn U.S. and China.
Miller: Clear that there will be more bifurcation, esp. at leading edge, between China and others. You see Japanese and Dutch restrictions.
Miller: Also in many countries, more emphasis on investment in domestic capacity, for econ. security & development. You see it in major and minor players. Should expect to see more money going into chips worldwide.
Miller: But concentration of production in certain countries has been durable for years or decades. That suggests change will be slow. I think world will look different, but not radically different.
Chang: Agree that there will be bifurcation of supply chain. I think U.S. has started to practice industrial policy on chips. Part of that is slowing down China. "I have no quarrel with that. In fact, I might say, I support it."
Chang: But in manufacturing technology, China is at least 5 or 6 years behind Taiwan. They are making their most advanced chip with difficulty. TSMC was making that chip 5-6 years ago with ease.
Chang: I certainly support U.S. policy to slow down China's progress.
Chang: U.S. official says Taiwan is a dangerous place and U.S. cannot get chips from there. Echoed by Janet Yellen. That is Taiwan's dilemma.
Chang: US already has 39% of the whole IC supply chain, with many chokepoints. 11% of manufacturing. My question is, with the CHIPS Act, what is their goal? Do they want to get to 30-40%? Or maintain essential supply for national security? If so, that's not that high: maybe 2%.
Miller: Over past years, more of a sense of risk looking at geopolitics, intensified by Russia-Ukraine war. Plus more concentration in the chip industry, memory in Korea, advanced logic in Taiwan. Industry has a difficult balance to strike. Efficiencies of scale vs. risk.
Miller: Geopolitical angle is seen as more dangerous today than 5 years ago. Balance between efficiency and diversity can be struck.
Chang: Pervasiveness and ubiquity of semiconductors is primarily due to its ever-cheaper cost. Used to be $2-3 per transistor. Now probably 1 nanodollar. A billionth of a dollar. Just about the "most disinflationary thing now."
Chang: If you give up Taiwan's advantages and move to the US, the 1st thing that will happen is cost will go up. I estimated 50% higher, that was an underestimate. Even worse than how I was underestimated by others.
Miller: I'll defer to your figures on Arizona's costs. I've heard Japan is not much more expensive than Taiwan. To me, it's not a question of finding the places, it's about getting the scale in those places. That needs balancing supply chain, and among gov'ts. This is key driver.
Wu: Is globalization dead?
Chang: I've said my piece. "No question in my mind that in chips sector, globalization is dead. Free trade is dead." Just look at how China has been embargoed. I agree with that and support that. Free trade slightly less dead, but in danger.
Wu: Will that help development of chip industry?
Chang: Of course not. I've talked about that. Costs will go up, and pervasiveness of chips will either stop or slow down considerably. "We're in a different game."
Miller: We use "globalization," but we mean "internationalization." Chips sold globally, but production dominated by very few countries. We will see shifts in market share, but success of business models as important as industrial policies.
Chang: Concentration exists because of competitive advantages. US has a big one in design. Designer should be close to the market: 1 reason why US has advantage in that sector.
Chang: Taiwan, Japan, SK have competitive advantage in manufacturing. "It's mainly the people and the work culture." We run continuously. If sth breaks at 1am, in US, it'll be fixed 9am. In Taiwan, it'd be 2am.
Chang: Engineer in Taiwan will get up, "wife will ask 'What's going on?'" He'll say something broke, and she will go back to sleep with no questions.
Chang: Taiwan's technical schools give TSMC very hardworking, professional and skilled technicians.
Chang: Another piece of the puzzle: in '81, TI discovered that new fab in Japan had twice the yield of Houston fab. I was told to find out why. 3-4% turnover among operators in Japan, 15-20% in US. Repair/maintenance, in US couldn't get the staff. In Japan, we got graduates.
Chang: In Japan, equipment ran 90%, in US, 60%. Basically stops production line.
Chris Miller now speaking to journalists.
Miller: Chip war has multiple battles, battles for market share, battles to get tech accepted, and battles between countries. There will never be an end to competition. But we are seeing more political concern and regulatory issues over past 5 years. Not going away any time soon.
Miller: Hard to decide which country is best at AI. When I look at ChatGPT, I'm impressed by results and the computing power necessary.
Miller: Reasonable to conclude that as countries invest in new geographies, they will get more efficient. Price differentials will decline over time. Gov'ts are also subsidizing, helping companies relocate.
Miller: The lesson of what went wrong with Soviet chip industry and what went right in Japan and Taiwan is importance of integration into international supply chain. No longer possible for China to follow that strategy. That's a real dilemma for Chinese firms.
Miller: RE: Robert O'Brien's comments about destroying TSMC, I'll leave that to him.
From Taiwan's VP and presidential candidate-in-waiting:

Mentions
See All