(The audio version of this post can be found on the Interconnected YouTube channel):
Morris Chang, the founder and two-time CEO of the Taiwan Semiconductor Manufacturing Corporation (TSMC), gave an hour-long speech to a room of Taiwanese government and business leaders a couple of weeks ago.
The speech received some light reporting from Chinese media and hardly any coverage at all in Western media, even though the 90-year-old industry veteran shared valuable insights on the global semiconductor competitive landscape.
I did a summary Twitter thread of Chang’s assessment, but there is a lot more to learn from what he said, especially as we look ahead to the technology competition between the US and China, where semiconductor plays a major role.
Let’s dig into why, in Chang’s eyes, TSMC succeeded, who he thinks will be a fierce competitor (spoiler alert: Samsung of South Korea), and why both the US and China will have their work cut out for them, if they try to replicate TSMC in their own backyard. (Pictures are screenshots of Chang’s slides during his speech.)
Taiwan’s Non-Obvious Advantages
Chang laid out three major advantages that led to Taiwan’s success -- all of which are rather non-obvious to me and require some dissecting.
- Technical Talent: Taiwan boasts a large number of dedicated engineers, technicians and other technical workers, who want to devote themselves to the advanced manufacturing industry (emphasis reflects the slide’s red highlight).
This is perhaps the most obvious of the non-obvious advantages. Obvious because strong talent is a prerequisite to any successful venture. Non-obvious because in 1985 when Chang went to Taiwan to start TSMC, Taiwan was not known for being a technical powerhouse -- those distinctions belonged to the US, Japan and parts of Europe. Chang had to attract a group of senior level talent from the US, most of whom of Taiwanese descent, to return to Taiwan to get TSMC going.
Chang attributed the dedication of these engineers to the Taiwanese people’s hard working spirit, which he does not see in America. This observation is true to a large extent. The 24-hour R&D operation, dubbed the “Nightingale program”, that TSMC was running in the mid-2010s during Chang’s second stint as CEO, would be unthinkable in the US. But this assessment also obscures some structural differences. Being a small, not-so-diverse economy, working for TSMC carried both lucrative rewards and social prestige in Taiwan. Without nearly as much competition from other high-paying industries, like finance, consulting, software tech, etc., TSMC is the place to be! Thus, the company has an envious 3-4% employee turnover rate.
2. Local Professional Management: Taiwan developed a strong corp of local, competent managers, though these managers will not perform well in other countries.
This advantage has a lot of subtleties. While having good managers is key to any company’s success, it’s particularly important for the operations-heavy, efficiency-driven chip foundry business. The localization of quality managers is what Chang believes made TSMC successful in Taiwan, but by the same token, will make it less successful abroad in places like Arizona.
In other words, managerial talent does not transfer well across borders, certainly compared to technical talent. Chang, as a seasoned American executive, admitted feeling this challenge and requisite cultural shock firsthand when he was starting TSMC. Up until that point, he has spent his entire adult and professional life in the US.
3. Highspeed Railways and Highways: Convenient, high speed transportation enables quick deployment of thousands of technical and manufacturing workers between TSMC’s three hubs -- Hsinchu, Tainan, Taichung -- without these employees needing to relocate their families.
This is the most non-obvious of the three advantages in my opinion. The fact that you can have close to a one-day commute between these three locations gives TSMC maximal flexibility to optimize its human resources to achieve both manufacturing and R&D output. According to Chang, these workforce redeployments can often last up to a year. If workers are forced to move their families every year due to reassignments, turnover rate will inevitably go up.
This advantage underscores an important point: the quality and convenience of basic infrastructure, like highways and highspeed railways, are crucial to not just the supply chain side of advanced manufacturing, but also the human management side. It is also why Chang is bearish on any attempt to replicate what TSMC has in Taiwan in a place like Arizona, where the geography is a massive sprawl with no railways to speak of.
Morris’s Own Ambition
One other non-obvious and under-emphasized advantage is Morris Chang himself -- a highly intelligent, driven, and ambitious individual who happens to be of Taiwanese descent.
Chang had wanted to become CEO of Texas Instruments, but he was passed over for the job. So he promptly left for General Instrument as its No. 2 executive at the time, no doubt in preparation to become CEO. That is until Chang realized General Instrument was not the company he wanted to take over, so he promptly left for Taiwan. (General Instrument became defunct 12 years after Chang left; he made the right call.)
Chang went to Taiwan in 1985 to become an entrepreneur and call the shots as CEO, at the ripe age of 54. Even today, when entrepreneurship is much more socially accepted and capital to invest in startups is as plentiful as ever, if you start a company in your 40s, you’d be considered “old”. Even though he has already had a successful career as a corporate executive, “Old Man Chang” was still ambitious and had more he wanted to do and prove.
Chang was also not a “nerd” from the beginning per se, like Gordon Moore (of Moore’s Law), who was doing chemistry experiments and making explosives in his garage during high school. Chang had many non-technical interests. As he shared with then Stanford president, John Hennessy, in 2014 in a Stanford Engineering Hero Lecture (video below), when he enrolled at Harvard as a freshman in 1949, he was very interested in politics, economics, and other social sciences -- not engineering. But there were no Chinese American politicians or notable businessmen, just laundromat, restaurant, and store owners. The way Chang saw it, the only way to get a middle class job was engineering or academic research, and Harvard didn’t have an engineering program at the time. Chang transferred to MIT and studied mechanical engineering out of pragmatism and lack of Chinese American role models!
Even though the rest is, as they say, history, had there been an Asian American governor, like Gary Locke, or senator, like Daniel Inouye, at the time, Chang might’ve been running for public office, not running semiconductor companies. And had he gotten the No. 1 job at Texas Instruments or decided to take over General Instrument, he would have never returned to Taiwan. TSMC would’ve never existed, and we may be living in a totally different world.
Our times shape our heroes. Our heroes shape our times.
Global Competitive Landscape
With Taiwan’s advantages and his own lived experience in mind, let’s look at what Chang thinks of the global semiconductor competitive landscape, namely the US, China, and South Korea (screenshots of his slides below).
The US: Strong Competitor with Structural Weaknesses
The US is still a strong competitor given its vast resources in land, water, and affordable electricity. These are attributes Taiwan, a small island, will never have. However, America does not have the three ingredients that made Taiwan the right place to enable TSMC’s success.
In his speech, Chang was gentle but firm in criticizing Americans as simply not as dedicated and hardworking as the Taiwanese. Along the same vein, he does not think America has the needed management talent nor does he believe Taiwanese managers will do well to fill the void in America. (It’s an assessment TSMC certainly hopes to prove wrong with its $12 billion investment to build a 5nm foundry in Arizona.) He was also dismissive of Arizona as the right location to concentrate America’s semiconductor manufacturing development, for reasons I’ve already outlined.
Why is Arizona getting so much investment and attention? Its recent status as a hotly contested swing state may have at least a little to do with it.
Thus, what America has Taiwan does not. What Taiwan has America does not. And it's more than a little ironic that Intel is trying to replicate TSMC, after arrogantly turning down the opportunity to invest in TSMC 35 years ago when Chang was trying to raise money for it.
Because America’s weaknesses are more structural, federal and state level subsidies are temporary bandages at best. Even putting the timeline issue aside, TSMC’s raw investment commitment of $100 billion in the next three years dwarfs that of Intel’s $20 billion investment in the next four years and the Biden administration’s proposed subsidy of $50 billion combined.
China: Weak Competitor
China’s semiconductor ambition and eye-popping investment amount from the central government grabs a lot of headlines, but Chang dismissed it as a weak competitor and spent less than a minute talking about it.
In his eyes, China is one to two years behind both the US and Taiwan in the logical chip design sector, and more than five years behind TSMC in manufacturing. China is not a threat. This dismissal is also implicitly aimed at some of his former TSMC top lieutenants, like Liang Mong-Song and Chiang Shang-Yi. Chang recruited them as part of the founding team of TSMC. Liang is now the co-CEO of China’s national chip foundry, SMIC. Chiang re-joined SMIC as a Deputy Chairman after being CEO of another Chinese foundry, HSMC, recently exposed as a fraud (for more, see “China's ‘Semiconductor Theranos’: HSMC”).
South Korea: Very Strong Competitor
The only competition that seems to keep Chang up at night is South Korea’s Samsung. The reasoning is quite straightforward now that we know what Chang thinks are TSMC’s “secret sauce”. Both South Korea, the country, and Samsung, the company, display similar attributes as Taiwan and TSMC, respectively. Of course, the same lens would also suggest that Samsung’s ventures in the US are doomed to fail, because top-notch Korean managers don’t necessarily perform well in America or elsewhere. For what it’s worth, Samsung’s Texas foundry is manufacturing Tesla’s most important component, its custom-designed Full Self-Driving chips. Is Samsung proving Chang wrong?
TSMC and Samsung have had a complicated and dramatic history. Back in 1989, Samsung tried (and failed) to recruit Chang himself, just two years after he started TSMC. The poaching has not stopped since and eventually people like Liang joined Samsung, which caused huge public uproars and high-profile lawsuits in Taiwan; the defection was akin to treason. When those top executives left for Samsung, Chang admitted it was the most challenging moment of his TSMC career.
South Korea and Samsung have always kept Chang up at night. That hasn’t stopped.
No Japan or Europe
Chang barely mentioned Japan and Europe during his speech, clearly not seeing either as a legitimate threat to TSMC.
Chang’s attitude towards Europe is not so clear, and I don’t know enough about him to make educated guesses. Looking at the industry as a whole, it does appear that the major European players have been playing a more complementary role to TSMC, not a competitive one. Leading chip companies like NXP are mostly fabless customers to TSMC, just like Qualcomm. Meanwhile, leading equipment makers like ASML treat TSMC as their most important customers. In 2020, TSMC accounted for 31% of ASML’s revenue -- the single largest source. That number will only grow given the global chip shortage and TSMC’s massive investment to fill it.
Chang’s attitude towards Japan is clear and has been expressed many years ago. In the same conversation with Hennessey I mentioned above, Chang shared his view that Japan’s refusal to embrace the fabless evolution hampered its ability to innovate and stay ahead. That evolution was in large part catalyzed by TSMC’s existence; self-interestedly, embracing fabless would also give TSMC more customers in Japan. In the end, Japan never did, but America did. Thus, much of the new chip design innovation accrued to American companies, but so began the hollowing out of America’s manufacturing capabilities.
Lessons to Learn?
Every runaway successful story is built on a mystical alignment of timing, resources, and an almost un-human level of hard work, patience, and dedication by extraordinary humans. TSMC is one of these stories. These stories are impossible to replicate, but are there lessons to be learned, especially for the US and China?
In this regard, there are more similarities than differences between the two superpowers. Both countries have a surprisingly short-term perspective when it comes to building their own semiconductor manufacturing, even though the TSMC experience clearly shows that it’s a multi-decade process. China has its “Made in China 2025” plan. The Biden administration’s proposed $50 billion program will surely not go beyond 2024, just given the American presidential electoral calendar. Despite governing via two different political systems, short-termism has infected both, because both leaderships are under pressure to “deliver”.
Multi-decade time horizon would be the first lesson I would draw.
There’s a second lesson I would draw: innovate to motivate.
As Chang poignantly pointed out, manufacturing is no longer a “cool place to work” that attracts the best and brightest in America; if you are young and smart, there are many industries to join with more money and prestige. You can say the same about China too, an increasingly diverse economy with plenty of ways to make money quicker than working in a chip foundry. And you can’t blame everyday citizens for making these choices; they are simply responding to the incentives available to them.
The nice thing about a smaller economy, like Taiwan and South Korea, is that while there are less resources, there are also less options, less distractions, and consequently a more singular focus when it comes to industrial policy. That context simply can’t be replicated in a large country.
The nice thing about a large economy, like the US and China, is their vast resources in terms of land, people, money, and the capacity to push boundaries of innovation that small countries would never dare to try. So to make “making stuff” cool again by only recreating the wheels of semiconductor manufacturing won’t work well, when Taiwan and South Korea already did that in the 80s and 90s, and especially when Moore’s Law is approaching its limit. Instead, a national call to action that motivates the best and brightest to make stuff to tackle climate change, quantum computing, space travel, or biotech -- all of which will need more semiconductors anyways -- is what big countries can uniquely deliver.
When Morris Chang started TSMC, I’m sure he never thought his creation would not only disrupt the most consequential sector of technology, but also give the world’s two superpowers plenty of heartburns and insecurities.
Heartburns and insecurities may just be what the doctor ordered for big countries to wake up to their senses and compete in a positive-sum way.
张忠谋提出了促使台湾在半导体制造行业能成功的三大优势 -- 这些优势对我来说都不算很显而易见，直观上很易懂，但需要进一步剖析。
这也许是几个“非明显”优势中相对最“明显”的一个。说它“明显”，是因为大量的人才是任何成功企业的先决条件；说它“非明显”，是因为1985年张忠谋去台湾创办台积电时，台湾并不是以技术实力著称的 -- 当时的强国是美国、日本和欧洲部分国家。张忠谋不得不从美国挖了一批资深人才，大部分都是台湾后裔，回台湾做台积电。
3. 高铁及高速公路提供的交通方便：便利的高速交通使上千的技术和制造工人能够在台积电的三个中心 -- 新竹、台南和台中 -- 之间快速调动和部署，同时不需要这些员工为工作变化而搬家。
另一个“非明显”且未被充分强调的优势就是张忠谋本人 -- 一位高智商、有动力、有野心的人，恰好又是台裔。
张忠谋一直想做Texas Instruments的一把手，但公司没有给他这个机会。所以他立即跳槽到General Instrument担任当时公司的二把手，无疑是为有朝一日当CEO做准备。直到他意识到General Instrument不是家他想接管的公司，所以再次离开，去了台湾。（General Instrument在张忠谋离开后12年就倒闭了；看来他的判决是正确的。）
1985年，张忠谋去了台湾，第一次创业，坐上了CEO的宝座，成了一把手，当时他已经54岁了。即使在今天，创业的社会认可度已经很高，投资于初创企业的资金也空前的充足，如果你在40多岁时做创业，还是会被视为个“老人”。尽管他的职业企业高管生涯已经非常成功，但 "张老" 仍然雄心勃勃，有更多他想做的和证明的事情。
张忠谋也并不是一个传统的“工科男”，比如像戈登·摩尔（“摩尔定律”的摩尔）那样，在高中时期就在自家车库里做化学实验和制作炸药。张忠谋其实有许多与科技无关的兴趣。正如他在2014年与当时的斯坦福大学校长John Hennessy在一场“斯坦福工程英雄讲座”（以下有视频）中所分享的那样，当他在1949年作为新生就读哈佛时，对政治、经济和其他社会科学兴趣浓厚 -- 而不是工程。但当时在美国没有任何华裔政治家或知名企业家，只有开洗衣店、餐馆和商店的小生意人。在当时的张忠谋眼里，要想在美国爬到中产阶级的唯一途径就是做工程师或做学术研究，而哈佛当时都没有工程学科。出于务实角度和美国华裔榜样的缺乏，张忠谋才转学到MIT去学习机械工程!
尽管接下来的事情已经众所周知，如果当时在美国有一位像骆家辉那样的亚裔州长，或Daniel Inouye那样的亚裔参议员，张忠谋很可能会参政，而不是参与经营半导体公司。如果他拿到了Texas Instruments CEO的位子，或者决定接管General Instrument，也就不会回台湾。台积电就不会存在，而我们可能会生活在一个完全不同的世界里。
唯一让张忠谋夜不能寐的竞争对手就是韩国的三星。我们已经了解了张忠谋认为台积电成功的 "秘方" 是什么，所以他对三星的态度也并不让人惊讶。韩国这个国家和三星这家公司，都分别拥有与台湾和台积电类似的优势。当然，用同样的框架来看，三星在美国的投资也注定要失败，因为一流的韩国经理人不一定在美国或其他地方是一流的。值得一提的是，三星在德州的芯片制造工厂正在代工特斯拉最重要的部件，即自己设计的自动驾驶芯片。三星是否企图在证明张忠谋的看法是错的？
在这方面，两大国之间的相似之处多于差异。在建立自己的半导体制造能力方面，两国的态度看似都很短期，尽管台积电的经历清晰的证明，这是一个数十年的过程。中国有自己的《中国制造2025》计划。而拜登政府提出的500亿美元补贴计划期限肯定不会超过2024年，因为美国总统选举日程。尽管是两种不同的政治，但“短期主义”似乎是两者的共同点，也许是因为两边的最高领导层都面临着 "交付" 和 “业绩” 的压力。
像美国和中国这样的大型经济体的好处是，在土地、人口、资金方面拥有大量资源，并且有能力推动小国不敢尝试的创新。因此，仅仅重新造半导体制造这个“轮子”来把 "造东西" 变得更酷，效果不会很好，因为台湾和韩国在80年代和90年代时已经做到了这一点了，而且现在摩尔定律正在接近其极限。相反，号召全民投入制造与解决气候变化、量子计算、太空旅行或生物技术有关的东西 -- 所有这些都需要更多的半导体 -- 是只有大国才能“交付”的。