TSMC founder Morris Chang on the evolution of the semiconductor industry - YouTube

Good afternoon, everyone. Thanks so much for coming. It's awesome to see 10 to 50 more than packed for such a wonderful reason. A tremendous honor to have Dr. Morris Chang back with us at MIT today. Dr. Chang, not only being an MIT alum. That's even before the formal introduction. Not only the MIT alum, but in my opinion, one of the most important innovators and industry leaders of the past 50 years, if not more.

大家下午好。非常感谢大家前来。能看到这么多人出席这一美好的场合真是太棒了。很荣幸今天能在麻省理工学院迎回张忠谋博士。张博士不仅是麻省理工学院的校友。在正式介绍之前，我要说，他在过去50年甚至更久的时间里，不仅是麻省理工学院的校友，而且在我看来，是最重要的创新者和行业领袖之一。

Dr. Chang, speech today is part of our manufacturing and MIT distinguished speaker series. This event hosted in partnership with the School of Engineering and Dean Anantha Chandra Kossen. And now to make the formal introduction, please welcome Professor Cindy Barnhart, Provost of MIT. Thank you, John.

张博士，今天的讲话是我们制造和麻省理工学院杰出演讲嘉宾系列的一部分。这个活动是与工程学院和安南莎·钱德拉·科森院长合作举办的。现在，请欢迎麻省理工学院教务长辛迪·巴恩哈特教授进行正式介绍。谢谢你，约翰。

I'd like to welcome you all to MIT's manufacturing at MIT distinguished speaker series. Today's lecture by Morris Chang promises to be illuminating and inspiring. Welcome back to MIT Morris. We're honored to have you here. This is the fifth event in manufacturing at MIT's speaker series, which has included U.S. undersecretary of defense for research and engineering, Haidishu, Intel CEO Pat Gelsinger, Pennsylvania governor, Tom Wolf, and Ford of CEO Jim Lico.

我想欢迎大家来到麻省理工学院的麻省制造业杰出演讲系列活动。今天的讲座由张忠谋先生主讲，定会启发人心。张先生，欢迎您回到麻省理工学院。我们很荣幸能够在这里与您交流。这是麻省理工学院制造业演讲系列活动的第五场，之前的演讲嘉宾包括美国国防部研究与工程副部长Haidishu、英特尔首席执行官Pat Gelsinger、宾夕法尼亚州州长Tom Wolf和福特汽车首席执行官Jim Lico。

I'd like to start by talking about Morris's deep ties to MIT. And then I will talk about his role as founder, informer chairman, and CEO of Taiwan semiconductor manufacturing company, also known as TSMC, the largest chip manufacturer in the world. Morris earned bachelor's and master's degrees from MIT, both in mechanical engineering, after transferring from Harvard. We're very proud of that move. After leaving MIT, a moment that he said was pivotal, he went on to pursue a career in industry and earned a PhD in electrical engineering from Stanford University.

我想从谈论莫里斯与麻省理工学院的密切联系开始。然后我将谈论他作为台湾半导体制造公司（TSMC）的创始人、董事长和首席执行官的角色，TSMC是世界上最大的芯片制造商。莫里斯在麻省理工学院获得了机械工程学士和硕士学位，毕业于哈佛后转学而来。我们为这一举动感到十分自豪。离开麻省理工学院后，他继续在工业界发展，并在斯坦福大学获得了电气工程博士学位。

Morris has called his time at MIT the most important part of his formal education and has said that it had a profound impact on his life. As with many success stories, it may have been failure that ultimately drove Morris's future triumphs. In a recent interview with the New York Times, Morris said, the greatest stroke of luck in my life was failing to be admitted to MIT's PhD program. It was then that he decided to try his luck in industry.

莫里斯称他在麻省理工学院的时光是他正式教育中最重要的部分，并表示这对他的生活产生了深远影响。和许多成功故事一样，或许正是失败最终推动了莫里斯未来的成功。在最近接受《纽约时报》采访时，莫里斯说，我生命中最大的幸运是未能被麻省理工学院的博士项目录取。从那时起，他决定试试在工业界的运气。

His first job out of MIT was at Sylvania, and later, Morris moved to Texas Instruments where he stayed for 25 years, eventually assuming the role of the chief executive in charge of the company's worldwide semiconductor business. While at Texas Instruments, he foresaw Asia's potential to excel at advanced manufacturing. So he founded TSMC in 1987, just as the industry was looking to outsource its manufacturing to Asia.

他毕业于麻省理工学院后的第一份工作是在斯威尼亚公司，后来，莫里斯转到德州仪器公司，在那里工作了25年，最终担任公司全球半导体业务的首席执行官。在德州仪器公司工作期间，他预见到了亚洲在先进制造业方面的潜力。因此，他于1987年创立了台积电公司，就在行业寻求将制造外包到亚洲之际。

TSMC quickly became one of the most profitable chip makers, and today is the largest manufacturer of the world's most advanced chips and serves some of the largest global companies. Today, TSMC is a $500 billion business that produces chips for cars and iPhones and supercomputers. Chances are right now, you have TSMC technology in your backpack or your pocket or your hands.

台积电迅速成为利润最丰厚的芯片制造商之一，如今是全球最大的先进芯片制造商，为一些全球最大的公司提供服务。如今，台积电是一个市值达5000亿美元的企业，生产汽车、iPhone和超级计算机的芯片。很有可能，你现在的背包、口袋或手中都有台积电的技术。

In founding TSMC, he transformed the semiconductor industry by pursuing a simple yet revolutionary concept to focus purely on manufacturing. Starting from the competitors and creating its foundry-like platform and its heavy investments in R&D, TSMC became a destination for the industry, and Morris became a titan of semiconductors. Morris, you've been a part of the semiconductor industry since the very beginning, and it's incredible to see you still at the center of it today.

在创建台积电时，他通过追求一个简单而革命性的概念——专注于制造——改变了半导体行业。从竞争对手开始，创建类似于晶圆厂的平台并大力投资研发，台积电成为了业界的热门选择，而莫里斯也成为了半导体巨头。莫里斯，自行业刚起步以来你就一直参与其中，看到你至今仍站在行业中心，真是令人难以置信。

And fortunately for MIT, you're also central to our community. Morris became a member of the MIT Corporation in 1999, and today serves as a life member emeritus. In 2016, MIT reopened the doors of Building E52 after a renovation funded by a generous gift from Morris and his wife Sophie. The building, which houses MIT's Department of Economics and the Sloan School, is now known as the Morris and Sophie Chang Building, where future leaders come together to learn and to shape the future.

幸运的是，对于麻省理工学院来说，您也是我们社区的核心存在。Morris于1999年成为麻省理工学院理事会的成员，如今担任终身名誉会员。2016年，麻省理工学院在得到Morris和他的妻子Sophie慷慨捐赠的资金支持下，重新开放了E52号楼的大门，该建筑现在被称为Morris和Sophie Chang大楼，其中设有麻省理工学院的经济系和斯隆商学院，未来的领导者们在这里聚集学习，并共同塑造未来。

MIT researchers have greatly benefited from access to advanced silicon fabrication technology through the TSMC University Shuttle Program, and TSMC has been active in collaborating with MIT and new devices memory and AI systems. For MIT, Morris is an extraordinary example of the lasting impact of our alumni, the lasting impact our alumni have on the legacy of innovation at the Institute.

麻省理工学院的研究员们通过台积电大学搭车计划获得了先进的硅制造技术，受益匪浅。台积电积极与麻省理工学院合作，共同研发新的记忆体和人工智能系统。对于麻省理工学院来说，莫里斯是我们校友对学院创新遗产的持久影响的杰出范例。

Today, Morris will discuss his path in building TSMC, how the company continues to produce chips on the cutting edge of technology, his assessment of the US chips industry and TSMC's investments in new fabrication facilities in Arizona. He's a pioneer who envisions the future and gets right to work making transformative, industrial change a reality. We're inspired by your example and eager to learn from your experiences. So without further delay, I'll turn this program over to you, Morris. Thank you again for being here with us today.

今天，莫里斯将会讨论他建立台积电的经历，以及公司如何继续在科技前沿生产芯片，他对美国芯片行业的评估，以及台积电在亚利桑那州投资新的制造设施。他是一个勇于设想未来并着手实现变革的先驱者。我们受到你的榜样启发，渴望从你的经验中学习。现在，我将这个节目交给你，莫里斯。再次感谢你今天和我们在一起。

Thank you, Cindy, for introducing me. And thank you, Suzanne and John, for inviting me to give a talk here. Ladies and gentlemen, this is a privilege and a pleasure to be back at MIT to give this talk. And I face the largest gathering of faculty that I ever faced at lunch today. And I was telling them that memories simply fell into the back at me this morning after I arrived here. I lived in several dormitories in the early 50s. I was here. I was at MIT between 1950 and 1955. And I lived in several dormitories, including the graduate house, Baker house, and a few other whose names I don't even recall now. And a lot of distinguished professors.

谢谢辛迪介绍我。还有谢谢苏珊和约翰邀请我在这里发表演讲。女士们先生们，我很荣幸也很愉快能回到麻省理工学院给这个演讲。今天午餐时我面对了我迄今为止在教职员工群体中最多的聚会。我告诉他们，今天早上到这里后，往事涌现在我脑海中。上世纪50年代初我住在几个宿舍里。我在MIT读书时间是在1950年至1955年之间。我在几个宿舍里住过，包括研究生宿舍、贝克楼等，还有几栋我现在甚至都记不得名字的宿舍。还有很多杰出的教授。

And at lunch, one of the professors asked me whether there was anything at TSMC, the company that I started, whether TSMC, I had anything that I wished to do over again. And I said no. Nobody asked me whether at MIT there was anything I wanted to do over again. If anybody did ask me that I would have said yes, I wanted to study harder. And Cindy, I think just pointed out that I failed the PhD horrifying exam price, which was the maximum that they allowed me to fail, of course. And I think that in retrospect, later in life, I said that was the luckiest thing that ever happened to me because I changed myself, changed the course of my life after I failed the exam. But at the time, I really didn't think it was lucky at all.

在午餐时，一位教授问我，在我创建的台积电公司，是否有什么事情我希望能够重新做一遍。我说没有。没人问我在麻省理工学院是否有什么事情我想重新做一遍。如果有人问我，我会说是的，我想更加努力学习。辛迪，我想她只是指出我曾经在博士考试中惨败，这是他们允许我的最大失败。回顾起来，以后的日子里，我曾经说这是我遇到的最幸运的事情，因为我在考试失败后改变了自己，改变了我生命的轨迹。但那时，我并没有觉得这是幸运的。

Anyway, let me start this talk by telling you the history of the topic. This is the final version. Nothing is alive in chips manufacturing from Texas to Taiwan. But actually, it has gone through several changes. My first version was alive in chips manufacturing from Texas to Taiwan. Then my second version was alive in chips manufacturing from Texas to Taiwan via Japan. I'll cover that story a little later. And then the final version is what you see here. But I really regret the addition of lessons. I mean, lessons, too many lessons. We can do with fewer lessons, frankly, at least I can do with fewer lessons. Well, anyway, but I had already sent the topic to Suzanne today to share. Let me start. I need to first be a lawyer. Frank, that I had hip joints surgery just two and a half years ago. And I never recovered completely from it. If anybody, if your doctor tells you that you can recover from a hip joint surgery, well, he may be right if you're young. But if you're a 90-year-old like I was for a head surgery, he was lying. I want to first talk a little bit about the pervasiveness of chips. And then some highlights in the history of chips. And then I want to talk about the foundry business model. And then Taiwan's advantage is in chips manufacturing and then some lessons.

总之，让我通过讲述这个主题的历史来开始这次谈话。这是最终版本。从德克萨斯到台湾的芯片制造中没有什么是活生生的。但实际上，它经历过几次变化。我的第一个版本是从德克萨斯到台湾的芯片制造是活生生的。然后我的第二个版本是通过日本从德克萨斯到台湾的芯片制造是活生生的。我稍后会讲一下这个故事。然后最终版本就是您现在看到的。但我真的很后悔增加了许多教训。我是说，教训太多了。坦率地说，我们可以少些教训。嗯，无论如何，但我已经今天把这个话题发送给了苏珊分享。让我开始吧。首先，我需要先做一个律师。坦白地说，我在两年半前接受了髋关节手术。但我从未完全康复。如果有人告诉你，你可以从髋关节手术中康复，那么如果你还年轻，他可能是对的。但如果你像我一样是个90岁的人接受了手术，他是在说谎。我首先想谈一下芯片的普及程度，然后介绍一些芯片历史上的亮点。然后我想谈一下晶圆代工业务模式，接着谈一下台湾在芯片制造中的优势，最后谈一下一些教训。

Pervasiveness of chips, I think we all know that. I should say that when I first went to Taiwan to start TSMC 38 years ago, it was 1985. And chips weren't nearly so pervasive as they are now. So I was told to give a few speeches to the Taiwan business circles and to the public sometimes. So I started to talk about semiconductors. I said, well, I said if you, there was no cell phone at that time. There were computers, but usually the computers were owned by companies, big businesses, and not by people, not by individuals. So I said, semiconductors, if you have a watch, if you have a digital watch, then you are not carrying a semiconductor with you. Well, very few people even had digital watches. So anyway, but the situation has completely changed in the last 40 years. And the semiconductors are very pervasive now. And in the National Defense, you know, drones, missile guidance and so on, industry in commerce, computers, daily life, smartphones, and cars, even washing machines. In the developed world of approximately two and a half billion people, nearly every person uses chips, chips products in their day-to-day life.

芯片的普及，我想我们都知道。我应该说，当我38年前第一次去台湾创立台积电时，那是1985年。当时芯片还没有像现在这样普及。所以有时候我被要求在台湾商界和公众面前演讲。于是我开始谈论半导体。我说，我说那时候没有手机。有电脑，但通常电脑是公司、大企业所有的，而不是个人。所以我说，半导体，如果你有一只手表，如果你有一只数码手表，那么你随身带着的就是一块半导体。当时很少有人甚至拥有数字手表。总之，过去40年情况已经完全改变。现在半导体非常普及。在国防方面，无人机，导弹制导等等，工业和商业中，电脑，日常生活中，智能手机，汽车，甚至洗衣机。在大约25亿人口的发达世界中，几乎每个人每天都在使用芯片产品。

Highlights in the history of chips. Well, just if you point a common knowledge, I guess, the conductivity of semiconductor lies between conductors such as metals and insulators, such as wood, hence the term semiconductors. Conductors of certain properties, which remains largely unexplored until 1947. Well, actually, you know, the word semiconductors, which I think virtually everybody understands now, but only when I first arrived in the United States in 1949, I bet you that only the scientists, probably the physicists and some chemists knew what a semiconductor was. I mean, that word was largely unknown until 1947.

芯片历史上的亮点。如果你指的是一个常识的话，我猜半导体的导电性介于导体（如金属）和绝缘体（如木头）之间，因此称为半导体。半导体具有某种特性，直到1947年仍然未被很多人探索。实际上，你知道的，半导体这个词现在几乎每个人都理解，但是当我1949年第一次来到美国时，我敢打赌只有科学家，可能是物理学家和一些化学家知道什么是半导体。我意思是，直到1947年，这个词几乎是未知的。

What the seven, sharply, body, and pertain, invented the transistor, which was based on semiconductors. And they invented it in Bell Labs, and that was worth a Nobel Prize. AT&T then began the experimental fabrication of Genesis. Not anybody else, just AT&T for five years. In 1952, AT&T decided they were quite unselfish. And, you know, they decided that the transistor was too important to keep it to AT&T. So the license changes the technology to various companies. Those companies included the big ones, well-known ones, well-known at that time, like RCHE, IBM, and also unknown ones, like Texas Instruments. Texas Instruments was very small and unknown at that time in 1952. And various companies began to produce, and those companies that got the license began to produce transistors.

七个尖锐的专家团队发明了基于半导体的晶体管。他们在贝尔实验室发明了这个技术，为此获得了诺贝尔奖。随后，AT&T开始进行Genesis的实验制造，持续了五年之久。1952年，AT&T决定放弃私利，认为晶体管太重要了，不应该只由AT&T独自掌握。所以他们转让了该技术给各个公司，其中包括了一些大公司如RCHE和IBM，还有一些在当时不知名的公司，比如德州仪器。德州仪器在1952年时还很小，不太为人所知，但随后开始生产晶体管。

All right, now jumping ahead, in 1958, Jack Kilby, he joined Texas Instruments in 1958. With me, we joined together. Well, not one of the same day, but only the same man, in the same man. May of 1958. In 1958, well, he invented the Indiegler circuit a few months after he joined Texas Instruments. And Robert Neuss, at Fairchild, the two separately, in a man's Indiegler circuit, almost the same time. And that was another Nobel Prize, the two of them. But actually, only Kilby got in the Nobel Prize, because by the time Kilby got in the Nobel Prize, Bob Noisher died. I think it was 2000, and Bob Noisher died. Bob Nois died in around 1990, I believe. And Kilby was the only one that got the Nobel Prize because of Indiegler circuit. And Kilby did say in his acceptance speech that if Bob Nois were still alive, he would have shared the Nobel Prize with him.

好的，现在快进到1958年，杰克·基尔比在1958年加入了德州仪器公司。我和他一起加入。好吧，并不是同一天，但是只是同一个人，在同一个人。1958年5月。在1958年，嗯，他在加入德州仪器公司几个月后发明了集成电路。而罗伯特·诺伊斯，在飞查德公司，他们两个分别在同一个人的集成电路上，几乎在同一时间。那就是另一个诺贝尔奖，他们两个。但实际上，只有基尔比获得了诺贝尔奖，因为基尔比获得诺贝尔奖时，鲍勃·诺奇已经去世。我想是在2000年，鲍勃·诺奇去世。鲍勃·诺伊斯大约在1990年去世，我相信。基尔比是唯一因为集成电路获得诺贝尔奖的人。基尔比在领奖演讲中表示，如果鲍勃·诺伊斯还活着，他会和他一起分享诺贝尔奖。

Now, fast forward to 1965, got a more prediction that, got a more, who was a close colleague of Bob Nois. He predicted that the number of geniuses in Indiegler circuit doubles every 1.5 to 2 years. The prediction was later known as Moore's Law, and remained valid for 50 years. And this was almost a miracle that something like this could be valid for 50 years. And Nobel Prizes are not given to engineers. But it was worth an IEEE Medal of Honor, which is often considered the equivalent of Nobel Prize for electrical engineers. IEEE, of course, stands for Institute of Electrical and Electronic Engineers.

现在，快进到1965年，有更多的预测，获得了更多，他是鲍勃·诺伊斯的亲密同事。他预测半导体电路中的天才数量每1.5到2年翻倍一次。这个预测后来被称为摩尔定律，并持续了50年。几乎可以说这是一个奇迹，这样的事情竟然能持续50年。诺贝尔奖不会颁发给工程师。但这却值得IEEE荣誉奖章，这往往被认为是电气工程师的诺贝尔奖。IEEE当然代表着电气和电子工程师学会。

The exponential increase of transistor on chip, and the corresponding decrease of cost of transistor allowed an ever wider range of applications. Chips started to become pervasive. Well, I will just give you an example. I looked it up. An example of how Moore's Law would work for a baby's network. Let's say your son is born today, and you give him a dollar. One dollar. Well, if that dollar follows Moore's Law, it doubles every year and a half to 2 years. Your son will become a business near by the time he's 50 years old, a billionaire in 50 years. And, well, he's been 50 years old. He's still well able to enjoy his wealth, okay? But let's say he lives longer. And by the year he, by the time he reaches 70, he will have become a trillion there. That's how Moore's Law works. You can verify what I just said. It's Wikipedia. You can look up Wikipedia, or you can do the calculation yourself. It's a pretty simple calculation in logarithm.

晶体管在芯片上的指数增长，以及晶体管成本的相应降低，使得应用范围变得越来越广泛。芯片开始变得无处不在。好吧，我就给你举个例子。我查了一下。摩尔定律在一个婴儿网络中是如何运行的一个例子。假设你的儿子今天出生，你给了他一美元。一美元。如果这一美元遵循摩尔定律，它将在一年半到两年内翻倍。你的儿子在他50岁的时候将成为一个商业巨头，50年后会成为亿万富翁。嗯，他已经50岁了。他仍然可以很好地享受他的财富，明白吗？但让我们假设他活得更长。到他70岁的时候，他将成为一个万亿富翁。这就是摩尔定律的工作原理。你可以验证我刚才说的话。这是维基百科。你可以查阅维基百科，或者你可以自己做计算。这是一个相当简单的对数计算。

Up to the 60s, almost all chips were bipolar. That's an important development. Up to the late 60s, almost all chips were bipolar. And US companies dominated the manufacturing of bipolar chips. And European Japanese companies played less than those, but they also made bipolar chips. But the US companies dominated. Now, the important development was MOS. MOS was developed in the late 60s. And it enabled smaller transistors than bipolar technology did. Moore's Law would not live for very long if it weren't for MOS, because when Moore predicted his Moore's Law, MOS had not been developed yet.

直到60年代，几乎所有的芯片都是双极的。这是一个重要的发展。直到60年代末，几乎所有的芯片都是双极的。美国公司主导了双极芯片的制造。欧洲和日本公司参与度不及美国公司，但它们也生产双极芯片。但是美国公司占主导地位。现在，重要的发展是MOS。MOS是在60年代末发展起来的。它比双极技术能够实现更小的晶体管。如果没有MOS，摩尔定律将不会长久存在，因为当摩尔预言他的摩尔定律时，MOS还没有被发展出来。

It was just bipolar. And bipolar, the density of transistors in bipolar was limited. And it was MOS that allowed Moore's Law to be valid for 50 years. And MOS was developed in the late 60s, after Moore's Law. So, in a sense, Gordon was a little lucky, you know. And it was smaller transistors than bipolar. And thus, Moore transistors per chip. Roughly, the maximum number of transistors on bipolar chip, say one square centimeter.

它只是双极性晶体管。双极性晶体管的密度是有限的。是MOS技术让摩尔定律保持了50年的有效性。MOS技术是在60年代末发展起来的，摩尔定律之后才出现。所以，从某种意义上说，戈登有点运气，你知道。MOS技术比双极晶体管更小。因此，每个芯片上有更多的晶体管。粗略地说，双极晶体管芯片上的晶体管数量最多是一个平方厘米。

That's a fairly typical chip size, one square centimeter. The maximum on bipolar was about a thousand, ten to the third. I know that, because I tried it. I tried to put 10,000 transistors on a chip by bipolar and it failed. And actually, Intel, at the same time, tried it. I'm not sure they failed too, you know. That's why they, when the MOS, we all, when the MOS dropped. But the maximum number of transistors on MOS chip started at ten to the third in 1970. The same time that we found out bipolar was limited to ten to the third. MOS started at ten to the third in 1970 and has now increased to ten to the tenth.

这是一个相当典型的芯片尺寸，一平方厘米。双极晶体管的最大值约为一千，十的三次方。我知道这个数字，因为我尝试过。我曾试图在双极晶体管上放置一万个晶体管，但失败了。实际上，英特尔在同一时间也进行了尝试。我不确定他们是否也失败了。这就是为什么当MOS技术崛起时，我们所有人都转向MOS技术。MOS芯片上的晶体管数量最初在1970年达到了一千，与我们发现的双极晶体管的限制相同。MOS在1970年起从一千开始，现在增长到十的十次方。

That's where the trillionaire and the billionaire stuff, you know, that I just told you came about, you know, in 2023. In this day, now, right now, the most dense chips that TSMB makes, TSMB makes, has almost 20 billion transistors in all one chip, 20 billion. That's, that's, that's two times ten to the tenth. MOS displays bipolar in the 70s and 80s. And as chips became more complex, technologies began to move from chip manufacturing to chip architecture and design. And that's an important development too, very important, you know.

这就是关于亿万富翁和亿万富豪的事情，你知道的，就是我刚告诉你的，发生在2023年。在今天，现在，TSMB制造的最密集的芯片，每个芯片里有近200亿个晶体管，20亿。那就是，那就是十的十次方的两倍。MOS在70年代和80年代显示出双极性。随着芯片变得越来越复杂，技术开始从芯片制造转向芯片架构和设计。这也是一个重要的发展，非常重要，你知道。

The first generation semiconductor people were mainly physicists and chemists, chemists and electrical engineers who worked on processing, how to make transistors or integrations. The later generation, the second generation, I mean, they don't, they don't work on how to make integrations anymore. They work on how to design, how to architecture, integrate that. So the later generation semiconductor people, I'm the first generation. I mean, I mean, when started in the semiconductor industry, you know, design's been simple. They only, it's thousand transistors on a chip, how complicated it can get, you know, it's thousand transistors. So, I, I didn't want to do simple things like that.

第一代半导体人主要是物理学家、化学家和电气工程师，他们致力于处理过程，研究如何制造晶体管或集成电路。第二代，也就是后来的一代，他们不再致力于如何制造集成电路了。他们致力于设计、架构和集成电路。所以后来的一代半导体人，我是第一代。我是说，我是说，在半导体行业开始时，设计很简单。只有一千个晶体管在一个芯片上，有多复杂呢，你知道，只有一千个晶体管。所以，我不想做那种简单的事情。

I worked on how to make them, how to make them more complex. I'll make them all all right. Then somebody had to design the architect, the complex circuits. And that's the second generation. That's, that's when, you know, the microprocessor, the Intel people that invented the microprocessor came along. That's when, you know, all the Qualcomm people and Qualcomm, you know, I mean, they are no longer a part of people. They are computer scientists. And they, they think, they think at home. They don't have to work in the factory. The first generation like me had to work in the factory. These people, you know, sure, they come to the office sometimes. But, but they, well, they can't. Anyway, they can mainly, you know, work at home.

我研究如何制造它们，如何使它们更加复杂。我会让它们都做得很好。然后有人必须设计那些复杂的电路。那就是第二代了。就是那时候，你知道，那些发明微处理器的英特尔人出现了。就是那时候，你知道，所有的高通人和高通，他们不再是普通人。他们是计算机科学家。他们在家思考。他们不必在工厂里工作。像我这样的第一代必须在工厂里工作。这些人，当然，有时候会来办公室。但是，他们，嗯，他们不能。无论如何，他们主要是在家工作。

I almost skipped over my contributions. As chips, well, I was on, on that, on the second pair with chips became more complex. Technologists began so on. I talked, talked about that already. In the 70s, 80s, Japanese companies mounted a charge for leadership in chip manufacturing. A fizzled in the early 90s. Why did fizzled? Various reasons were given. But one of the most important reasons, I think, was that, was the PASA accord. PASA accord. This was reached in 1985. The same year that I went to Taiwan. There's no connection between the two, but it was just happening at the same time. The other accord made the Japanese yen up raise. The exchange rate was almost doubled in two and a half years. 1985, you look up the record. 1985, the Japanese yen, $20. I think, $250 or something like that. $250 yen to $3,000. Two and a half years later, it was $100 something to $1,000. That's, the Japanese had to fold up with an exchange rate twice. There were other reasons. But anyway, the Japanese onslaught fizzled in the early 90s, actually in the late 80s and in the early 90s.

我几乎错过了我的贡献。关于芯片，嗯，当时我正在，那个，和芯片相关的第二波芯片变得更加复杂。技术人员开始了等等。我已经谈过了这个话题。在70年代，80年代，日本公司开始在芯片制造领域发起进攻。但在90年代初就不了了之了。为什么会失败呢？有各种各样的理由。但我认为最重要的一个原因是PASA协议。PASA协议。这个协议是在1985年达成的。恰巧我去台湾的时候也是在同一年。这两件事之间并没有联系，只是恰好发生在同一时间。另一个协议让日元升值。汇率在两年半内几乎翻了一番。1985年，你可以查一下记录。1985年，日元是20美元。我记得是250美元左右。两年半后，变成了100左右1美元。日本人必须应对汇率翻倍的现实。还有其他原因。总之，日本的进攻在90年代初实际上是在80年代末就已经失败了。

Now, in 1987, Morris Chan founded Taiwan Seminary Family Broadcasting Company based on an innovative business model. It dedicated a semiconductor boundary. Again, this was not worth a Nobel Prize, but it was worth an IEEE Medal of Honor. Thank you. Since the 90s, for three decades now, Taiwan and South Korea have taken on an increasing role in chip manufacturing, while the U.S. retains a dominant role in chip architecture and design. And in some other things like equipment, manufacturing equipment for chips, U.S. has a dominant role. Of course, that role is shared by the Japanese equipment manufacturers and lately by the Dutch Holland manufacturers. But still, the U.S. has a very important role in other aspects than just chip design and architecture. In chip design and architecture, the U.S. role is more than dominant. It's almost unquestionably dominant. All right.

现在，1987年，陈俊贤基于创新的商业模式创立了台湾神学家家庭广播公司。 它致力于半导体领域。 再次强调，这并不值得诺贝尔奖，但值得IEEE荣誉奖。 谢谢。 自上世纪90年代以来，至今已有三十年，台湾和韩国在芯片制造领域扮演着越来越重要的角色，而美国在芯片架构和设计方面保持着主导地位。 在一些其他方面，比如芯片制造设备，美国也起着主导作用。 当然，这一角色也被日本设备制造商所分享，最近还有荷兰制造商。 但是，美国在仅仅芯片设计和架构之外的其他方面仍扮演着非常重要的角色。 在芯片设计和架构方面，美国的地位几乎无可置疑地占据主导地位。 好的。

Now, I mean, I fool around with the stegwim for so long. I eventually tried of it. I decided to just already explain it to you. If you will, if you will, just delete the red line for a moment. Delete in your mind. Just don't look at the red line. Don't see the red line. So what you have left now is just the back line. That was the old business model of semiconductor manufacturing. The same company will do R&D on design. And most of them will do their own design tools. But of course, companies like Cadence and Synops, well, I'm looking at the red line. Don't forget about it. Okay. And these are so-called IDM integrated device manufacturer. They're integrated because they do everything. All the back boxes. After designing the IC, they do the way for fabrication. They do the packaging and testing. And then they sell them. The only thing that TSMC did was to impose the red line. And just look at the red line enclosure. Wave of fabrication, research on the wave of packaging, and advanced packaging. And TSMC took that. Of course, later on, people would start to say, well, that's the heart. That's the heart of our IDM business. Well, at the time, it wasn't really considered to be the heart. But after TSMC took it and succeeded with that business model, well, people began to be a little ambious about it. Anyway, that's what the Foundry business is just the red enclosure.

现在，我的意思是，我一直在研究stegwim好长时间了。我最终对此感到厌倦。我决定直接向你解释。如果你愿意的话，只要把红线删除一会儿。在你的头脑中删除。只要不看红线。不要看到红线。现在你所剩下的就只是背线了。这是半导体制造的旧商业模式。同一家公司会进行设计研发。而大多数公司将会使用他们自己的设计工具。但当然，像Cadence和Synops这样的公司，嗯，我又开始看红线了，别忘了。好的。这些公司被称为IDM集成器件制造商。他们之所以被称为集成是因为他们做所有事情。所有的背景。在设计芯片之后，他们会进行制造工艺。他们进行封装和测试，然后将其销售。TSMC唯一做的就是强调红线。并且只看红线围栏。制造工艺波，包装的研究波，以及先进封装。TSMC采用了这一做法。当然，后来人们会说，哦，这就是我们IDM业务的核心。而当时，这并不被认为是核心。但是在TSMC采用并成功实施了这种商业模式之后，人们开始对此感到有些嫉妒。无论如何，晶片代工业务就是这种红色围栏。

Now, we did it in Taiwan. What were so special about Taiwan? Taiwan's advantage in chip manufacturing. Terrence, substantial supply of high quality in dedicated. I underlined technicians under all of them because they were all very important. Not just engineers. Not just engineers, but technicians, operators. And actually, day after tomorrow, I'm going to have a dialogue on the subject of education in the Asia society. And I'm going to say that when we talk about education, we usually emphasize education of the elite. Like the kind of education the MIT gives. MIT students get. That's the education of the elite. But why is TSMC successful in Taiwan? Because TSMC also gets good, well-trained technicians and even well-trained operators from a lot of trade schools in Taiwan. Trade schools. There are 10,000 miles away from schools like MIT or Harvard. There are students only aspire to make a good living as technicians. And even operators. Well, we don't really. Taiwan doesn't really have schools for operators, but operators, but they're willing operators. I mean, they don't turn over. They don't leave their job as soon as something better, something that pays more. They're almost like Japan.

现在，我们在台湾做到了。台湾有什么特别之处？台湾在芯片制造方面的优势。特伦斯，有大量供应的高素质专业人才。我在所有这些人才下面划了下划线，因为他们都非常重要。不仅仅是工程师。不仅仅是工程师，还有技术人员，操作员。实际上，后天，我将在亚洲协会进行有关教育问题的对话。我要说的是，当我们谈论教育时，通常强调的是精英教育。就像麻省理工学院所提供的教育。麻省理工的学生接受的就是精英教育。但是，台积电在台湾成功的原因是什么？因为台积电也从台湾许多贸易学校中得到了受过良好训练的技术人员，甚至受过良好训练的操作员。贸易学校。这些学校距离麻省理工学院或哈佛大学有一万英里之遥。那些学生的目标只是想过上为技术人员甚至操作员的好生活。嗯，我们并没有真正的为操作员开设学校，但是有热忱的操作员。我的意思是，他们不会随时离开工作，去找一个更好、更高薪的工作。他们几乎就像日本人。

I'm not going to have time to talk about the life in chips manufacturing from Texas to Taiwan. I'm not going to have time to talk about the Vietnam part. But in Japan, you know, when I was running Texas Instruments, Japan plants plants, the operators, first of all, we had our major plants, we had our huge dollars. And when I walked into the factory, all the operators started looking at me, a new visitor. Curious. But when I walked into a Japanese plant, no operator ever looked at me. And then I asked, what is the hotel where I asked the manager? What is the hotel where they are? These are young women, the operators. They don't leave unless they get married. A turnover rate is like 2% a year, whereas in Texas, it was like, you know, 15% when there was a recession. A recession meant fewer jobs available. And 25% when the times were good, you know, I mean, they all leave when there's a better paying job available somewhere else. Well, anyway, and that's deadly because it takes about three months, at least, to train and operate. And we have a 25% turnover rate. You can't do it. You can't do any manufacturing.

我没有时间讨论从得克萨斯到台湾的芯片制造业生活。我也没有时间来谈论越南那部分。但在日本，当我经营德州仪器的时候，日本工厂的操作员们，首先，我们有我们的主要工厂，我们有大量资金。当我走进工厂时，所有的操作员都开始看着我，一个新来的访客。好奇。但当我走进一个日本工厂时，没有一个操作员会看我。然后我问，我问经理说那里的旅馆在哪里。那些年轻女性就是操作员。除非她们结婚，她们是不离开的。离职率大约是每年2%，而得克萨斯的时候，它就像，你知道的，在衰退的时候是15%。衰退意味着更少的工作机会。而在繁荣的时候则是25%，你知道，这意味着他们会在其他地方有更好的薪水时离开。无论如何，这是致命的，因为至少需要三个月的时间来培训和操作。而我们有25%的离职率。你无法做到。你无法进行任何的制造。

Can't do any manufacturing. And I still remember, you know, Jack Kilby, I was talking about, Jack Kilby and I joining Texas Instruments almost at the same time. I just, I can't remember, I mean, he, after he invented the, the Indigo Circuit, they told him to run an Indigo Circuit department. All right, so when he was doing that, and I was running a transistor department, and we were too walking towards the cafeteria to have our lunch. And then a bunch of ladies operators, you know, just Russia had because, you know, only 30 minutes was allowed for lunch at Texas Instruments. A bunch of lady operators, Russia had overtook us. And Jack Kilby said to me, well, I mean, these operators, do you realize Morris, the, some of them have never made a single good Indigo Circuit in her life? Well, because, you know, the turnover is so rapid and there's not enough training. Well, anyway.

我现在不能进行任何制造。我还记得，你知道的，杰克·基尔比，我说的是，杰克·基尔比和我几乎同时加入德州仪器公司。我只是，我记不起来了，我是说，他在发明靛蓝电路之后，他们让他负责一个靛蓝电路部门。好吧，当他在做那个的时候，我负责一个晶体管部门，我们两个正朝着餐厅走去吃午饭。然后一群女操作员，你知道的，因为在德州仪器公司只允许午饭30分钟，一群女操作员冲了过来超过了我们。杰克·基尔比对我说，嘿，这些操作员，你意识到了吗，莫里斯，她们中有些人一辈子都没有做过一个好的靛蓝电路？嗯，因为，你知道的，人员流动如此迅速，培训也不足。无论如何。

All right. I covered the low turnover rate and geographical concentration. You know, Taiwan is a relatively small island, but they are, we do have major facilities, factories in three cities in Taiwan. And however, they are all connected, all three cities are connected by fast trains, high speed trains. And at any time, including now, there will be a thousand or more engineers, not operators or technicians. A thousand or more engineers assigned to another city, another city other than his home city. They, we provide dormitories, TSMC provides dormitories. So they live in their dormitories. They go there from their home with the, on the rubber train. And they go there Monday morning and they live in the dormitories Monday night, Tuesday night, Wednesday night, Thursday night. And then Friday afternoon, they go home. At any time, they are over a thousand engineers doing that. Well, I mean, it's not fun life, really, but, but it's, it's effective for the company. And, and they are willing to do it. So, your record, your web ago country, and by the way, you know, learning curve theory. And I'm not going to explain you probably know it anyway, but it works. The learning curve experience curve theory. It works only when you have a common location and the common people. It doesn't work if you have a lot of branches or all the place. Learning is local. Learning curve works only locally. All right. All right.

好的。我讨论了低员工流动率和地理集中度。你知道，台湾是一个相对较小的岛屿，但我们在台湾的三个城市都有主要设施和工厂。而且，这三个城市都通过快速列车、高速列车相连。任何时候，包括现在，都会有一千个以上的工程师，而不是操作员或技术员，被分配到另一个城市，而非他们的家乡城市。TSMC提供宿舍，所以他们住在宿舍里。他们从家里坐上火车去那里，周一早上过去，周一晚，周二晚，周三晚，周四晚在宿舍里住，然后周五下午回家。任何时候，都有一千多名工程师这样做。嗯，我是说，这并不是一种有趣的生活，但是对公司来说是有效的。而且，他们愿意这样做。所以，你的记录，你的网络合作，顺便说一下，你知道的学习曲线理论。我不会解释，你可能已经知道了，但它是有效的。学习曲线和经验曲线理论只在有共同地点和共同人员的情况下起效。如果你有很多分支机构或地方，它就不会起作用，学习是本地的。学习曲线只在本地起作用。好的。好的。

Okay. Large community experience. Yeah. All right. Oh, yeah. And the chip ecosystem has been built over the last three decades. Off-stream, misstream, and important global semiconductor equipment manufacturers such as ASML, or primary teams, and amateurs, or having service centers, training centers, and labs in Taiwan, and downstream packaging and testing companies. So, I think, yeah, we, yeah. Lessons. Well, if they, if they are in the lessons, the susceptibility of a country to chip manufacturing seems, this is my opinion, seems to be related to the status of economic development of that country. Frankly, the advantages that Taiwan enjoys today, those that I spend some time to discuss today, they were enjoyed by the U.S. in the 50s and 60s, because I saw them. I witnessed them. In the 50s and 60s, I mean, I was first in Pennsylvania in Boston. Look, look, one 28. Yeah, woven, woven. And then in 1958, I moved to Texas. And so, what we see in Taiwan, we saw in the U.S. back then. 50s and 60s. And I think for Taiwan, they enjoy the advantage now, but I think they will lose advantage. They will lose it to another country. I don't know who may be in there. Maybe Vietnam, maybe Indonesia, in another 20 or 30 years. The economic model we have been following, free market, free trade, globalization, all those are really in the past now for chips anyway. No globalization, no free trade, no free market. Well, I shouldn't say it absolutely. I do some left, no, some free market, free trade over there. But a lot of it, a lot of free market, free trade over there is already gone. And that's the most efficient and resilient model. Deviations from it, industrial policy, subsidization, self-shoring will make the world chip industry less efficient and less resilient. Less resilient? How? Why? Well, you look at in the Cold War period, Soviet Union, they follow the planned economy model. They always have shortage of things that people really need. Because it's planned economy and nobody can plan the economy wisely. And the government cannot plan it. So the resilient way is the free economy, free market. The pervasiveness of chips due to value due to supply and trade efficiency, chip costs in Greece, its pervasiveness, will lessen. However, after having said all this, I need to agree with everybody, I think, that national security consortium, of course, overripe everything. And you cannot get away from that. I mean, heck, without national security, we lose everything. Everything that we value. So that's by all means avoid even a Cold War, if we can. Anyway, this is my message. Thank you very much. Thank you. Thank you. Thank you. Thank you. Thank you. APPLAUSE Go.

好的。大型社群经验。对。好的。哦，对。芯片生态系统已经在过去三十年建立起来了。上游、中游和重要的全球半导体设备制造商，如ASML，或者主要团队和业余者，在台湾设有服务中心、培训中心和实验室，以及下游封装和测试公司。 所以，我认为，是的，我们是。经验教训。如果他们处于经验教训中，一个国家对芯片制造的脆弱性似乎与该国经济发展的状态有关。老实说，今天台湾享有的优势，我今天花了一些时间讨论，这些优势在50年代和60年代美国享受过，因为我看到了它们。我亲眼见证了。在50年代和60年代，我是首次在宾夕法尼亚州和波士顿。看看，看看，一个28。是的，编织，编织。然后在1958年，我搬到了德克萨斯州。所以，在台湾看到的东西，我们50年代和60年代在美国看到的。我认为对于台湾来说，他们现在享有优势，但我认为他们将失去这种优势。他们将失去优势，会失去给另一个国家。我不知道是谁可能在那里。也许越南，也许印度尼西亚，再过20或30年。我们一直在遵循的经济模式，自由市场，自由贸易，全球化，所有这些对于芯片来说现在都已经过去了。没有全球化，没有自由贸易，没有自由市场。嗯，我不应该绝对说。我确实有一些自由市场，自由贸易在那里。但很多自由市场，自由贸易在那里已经消失了。这是最有效和最具韧性的模式。该模式的偏离，工业政策、补贴、回流生产将使世界芯片行业变得不那么有效和不那么有韧性。不那么有韧性？怎么会？为什么？嗯，你看在冷战时期，苏联他们遵循计划经济模式。他们总是缺少人们真正需要的东西。因为是计划经济，没有人能明智地规划经济。政府也不能规划。所以最具韧性的方式是自由经济、自由市场。由于价值、供应和贸易效率，芯片成本在希腊中的普遍化，将会减少。然而，在说了这一切之后，我需要同意每个人，我认为，国家安全联盟当然是最重要的。你无法逃避这一点。是的，没有国家安全，我们失去一切。我们重视的一切都会丧失。所以无论如何，我们都要尽量避免冷战，如果可以的话。无论如何，这是我的信息。非常感谢。谢谢。谢谢。谢谢。谢谢。谢谢。掌声。走吧。