The Silicon Empire: TSMC’s Revolution and Morris Chang’s Legacy

Taiwan, a small island outside of China with a population of just 24 million, has in three decades managed to become the most important nation in the world’s most important industry. The Taiwanese government’s bold bet on semiconductors has had an unimaginable impact on the pace of human innovation. This is the remarkable story of TSMC, and its iconic founder, Morris Chang.

• Morris Chang’s origins and legacy: Born in China, trained in the U.S., and crowned The Godfather of Chips in Taiwan.

• The birth of the first foundry: How TSMC’s unconventional business model started an unbundling of the entire semiconductor value chain.

• Philips’ crucial role in shaping semiconductor history: How the Dutch electronics giant helped fund and develop two of the industry’s most important companies.

• Intel’s major strategic blunders: Three major forks in the road that, in hindsight, led the U.S. chip giant down the wrong path.

• Inside a state-of-the-art fab: A deep dive into the costs and complexity of operating a semiconductor fabrication plant.

• The competitive advantages of TSMC: What makes the Taiwanese silicon empire Chang founded so uniquely positioned?

As a young boy born in Ningbo, China, in 1931, Morris Chang dreamt of making money telling stories. Chang’s father, on the other hand, was not as impressed with his ambitions of becoming a novelist or a journalist, and successfully persuaded him into mechanical engineering. Ironically enough, Chang would end up writing one of the most impressive business and technology stories of all time – etching the names of himself and his company into the history books forever.

In 1949, at 18 years old, Chang moved to the United States where he initially enrolled at Harvard University, but eventually ended up at MIT with a bachelor’s and a master’s degree in mechanical engineering. Failing his PhD exam twice, Chang decided to leave MIT and start applying his knowledge in the real world – famously turning down a job at Ford Motor Company in 1955 and instead joining a now defunct American semiconductor manufacturing company called Sylvania Semiconductor.

This path would lead him to Texas Instruments (TI) in 1958, where Chang climbed the corporate ranks during a successful 25-year tenure. In fact, TI early on paid for the opportunity of getting another shot at the PhD title, which Chang received in electrical engineering at Stanford in 1961.

Morris Chang talking to a production line supervisor at Texas Instruments.

Morris Chang talking to a production line supervisor at Texas Instruments.

At TI, just like at Sylvania, Chang was tasked with improving manufacturing yields, which basically means increasing the percentage of accurately produced chips per silicon wafer. Yield is, in the end, what determines the success of a semiconductor manufacturer. Yield management is therefore a crucial task where atom-level defects can render entire wafers – back then, including a few dozens or hundreds of chips – useless. This was a major issue at TI before Chang’s arrival, and his success soon earned him a strong reputation in the industry globally.

In the early 1980s, Morris Chang, still at TI, found himself in an unsettling position. He observed something startling: the company's Japanese factory was producing significantly more chips per wafer than its counterpart in Texas, meaning yields were far superior. This discrepancy gnawed at him. The reasons were quite simple – better-qualified staff, lower staff turnover, and a culture of maniacal precision and hard work that seemed innate in Japan but elusive in America. Chang drew a simple, yet profound conclusion: the future of advanced manufacturing lies not in the West, but in Asia.

Despite his vast accomplishments at TI and his global reputation in the industry, Chang hit a glass ceiling at the American chip giant, as he was passed over for the CEO position. This pivotal moment hurt him deeply, and it perfectly set the stage for what was about to take place. In a few years, Chang left TI for good, and took a position at General Instrument – a company that since 2014 is a part of Lenovo Group, but before that played a crucial role as a partner and later subsidiary of Alphabet in developing the Android ecosystem.

Before going into detail about the next adventure for Morris Chang, it is worth explaining the complex relationship between China and Taiwan. The latter, an island on the southeastern coast of China, was separated from the mainland after the Chinese Civil War in 1949. The Nationalist Party retreated to Taiwan after losing to the Communists, who established the People's Republic of China (PRC) on the mainland. Since then, Taiwan has functioned with a separate government, although the PRC views it as a breakaway province. The relationship remains tense, with Taiwan operating as a democratic state with its own identity, while China insists on reunification, leading to complex international dynamics and ongoing geopolitical tensions. The story we are about to tell does not make the situation less tense.

During his brief stint at General Instrument, Chang received a surprising yet intriguing offer. Sun Yun-suan, the Premier of the Republic of China, now governing from Taiwan, asking him to return to Asia. The invitation was to head the Industrial Technology Research Institute (ITRI) – located in Hsinchu, Taiwan, roughly 300 miles from Chang’s birthplace in mainland China – a government-backed non-profit tasked with jump-starting the nation’s technological aspirations.

Taking the helm at ITRI, despite Chang’s long and already successful career in the semiconductor industry, became his most important career decision yet. He was given a mandate that was as ambitious as it was daunting: to propel Taiwan onto the world stage of leading-edge semiconductor manufacturing by putting substantial resources to work. The government, much thanks to Chang’s know-how, was early to recognize the importance and potential in advanced chip production, and bet heavily on it with next-to perfect timing. This was just around the time when the West recognized Asia as the perfect low-cost outsourcing partner for the manufacturing of everything from consumer electronics to textiles and apparel.

So, in 1987, at age 56, Morris Chang founded TSMC – an acronym for Taiwan Semiconductor Manufacturing Company. A company that would rewrite both the rules and the world map of the semiconductor industry, and help companies like Apple and NVIDIA reshape how we interact with technology in our daily lives.

Morris Chang founded TSMC in 1987.

Morris Chang founded TSMC in 1987.

The capital for this expensive endeavor was supplied by the Taiwanese government, through National Development Fund (NDF), its sovereign wealth fund, and some of the crucial production technology and intellectual property licenses were secured through the Dutch electronics giant Philips. This meant that NDF took a 48% stake in the company, and Philips just under 28%, while the rest was backed by smaller private investors such as wealthy local families with ties to manufacturing.

Both Intel and Texas Instruments famously turned Chang down before he turned to Philips. Two very costly mistakes for America’s chip sector, in hindsight, and unfortunately not the last major one from Intel. Just three years before, Philips developed a joint venture with ASM International that would also turn out incredibly important for the industry. This was a company named ASML – today the most important semiconductor manufacturing equipment company in the world with a monopoly in so-called extreme ultraviolet (EUV) lithography. We will dive more into this technology later on.

This meant that by the mid 1990s, Philips owned 28% of TSMC at its 1994 IPO and closer to 70% ASML at its 1995 IPO – arguably two of the most important companies in the world today. Philips gradually sold out of both companies, and had in 2007 fully exited both. Since then, TSMC and ASML have returned over 9,500% and 78,300% respectively, and both companies combined are worth roughly 56x times that of Philips itself as of this writing.

Chang's vision for TSMC was revolutionary. He founded the company in 1987 with an idea of building the world’s first pure-play foundry. A business dedicated solely to manufacturing chips for other companies, instead of designing and selling its own branded products. He understood from his experience in TI’s fabrication plants – known in the industry as fabs – around the world and the rapidly rising demand for more and increasingly advanced chips that in-house manufacturing would become too complex for most chip designers. The days of unbundling were here.

This, at the time audacious, departure from the globally prevalent integrated device manufacturing (IDM) model – where you do everything from design, manufacturing to marketing in-house, like Intel and Samsung still to this day – would become TSMC's defining characteristic and the basis for its success. By focusing entirely on manufacturing and perfecting its yields, TSMC could meet the rising demand of chip companies all over the world, selling one of the industry’s most expensive and complex tasks on tap.

What Chang foresaw was how each step of the chip development process would become almost exponentially more expensive and complex as the nodes shrank. Keeping up with each step under one roof soon enough became very challenging, creating an explosion of fabless companies (meaning no in-house fab). Companies focused solely on chip design, leveraging advanced design tools from firms like Synopsys and Cadence, as well as instruction sets from Arm. The cost of just designing a leading-edge chip today is north of half a billion dollars. This model allows companies like NVIDIA and Tesla to innovate rapidly and scale efficiently without spending multiple billions of dollars in CapEx and R&D building its own fabs and perfecting its yields.

While large IDMs such as Intel did have competitive manufacturing capabilities to offer to third parties during this time, the American chip giant faced another dilemma; one of misaligned incentives and trust. The fact that IDMs actively compete with its foundry customers did not sit right with fabless customers such as Apple and Qualcomm, fearing their designs and technological advancements could be co-opted or deprioritized. This problem did not exist with TSMC. Trust and incentives, together with concentrating its efforts on the incredibly capital-intensive manufacturing part of the value chain, were two of the key ingredients in Chang’s vision.

One by one, even the larger IDMs were dropping out of the manufacturing race as TSMC and Samsung grew even more reliable. For example, the former American semiconductor powerhouse IBM sold its fab operations to GlobalFoundries in 2015, who itself had been spun off from AMD a few years before as it went fabless. The same goes for LSI Corporation, who sold its fab operations to TSMC.

Intel, on the other hand, is still trying to combine the IDM and foundry business models. Unfortunately, this, together with an astounding strategic mistake mentioned in the next chapter of this story, has made the former PC processor giant lose ground on both fronts – to companies like NVIDIA and Arm in the design infrastructure race and TSMC and Samsung in the node-shrinking manufacturing race.

However, Intel has recently secured the rights to the first of ASML’s latest High NA EUV systems – weighing no less than 150 tonnes – which means the story is far from over.

A defining moment in the company’s history came when Apple sought a partner to produce its own in-house processors for the iPhone – seeking to get rid of its reliance on its competitor Samsung, who ironically enough supplied processors to its largest competitor in the early days. This was just after announcing that Intel would supply its x86 processors to Apple’s Mac computers. Initially, Apple approached Intel for this special task, but Intel's then-CEO, Paul Otellini, turned Steve Jobs down.

At the time, Intel viewed the mobile phone processor market to be too small to be worth the R&D effort needed to deliver on such a deal. Remember, designing chips is incredibly expensive, so they had to be able to make up for it on manufacturing volume. This decision, in hindsight of course, might be one of the largest business blunders of all time, and Otellini himself has later said that they misjudged the volume part of that equation by something like 100x. Today, the smartphone segment accounts for roughly a third of chips sold, according to Chris Miller’s Chip War.

Turned down by their only viable option besides Samsung, Apple decided to take matters into their own hands. Luckily, they were one of the very few companies who actually could afford to think of designing their own chips. This was the start of Apple Silicon, which ironically enough later would lead to Intel losing its Mac processor business too.

Looking for a manufacturing partner, Apple naturally chose the world’s most advanced and trusted foundry to bring the new heart of its flagship product to life. Leveraging TSMC’s advanced process technologies to produce the breakthrough A-series chips for iPhones and iPads, and eventually the M-series chips for Macs was pivotal for both companies, and likely one of the key factors that propelled Apple to soon becoming the largest company in the world.

Bolstered by its advancement in user-friendly software, enabled through its increasingly advanced in-house designed chips, Apple managed to do what basically no other consumer electronics company had ever done until this point – create embedded digital lock-in effects in its hardware products, ensuring customer loyalty and streams of recurring revenue.

As per the last twelve months ending in Q3 FY2024, Apple reported a staggering $199B in revenue from iPhone sales alone. Keep in mind though, that this does not even account for its high-margin service segment, including revenue from the App Store, iCloud, and other subscription services.

The timing of this alliance could not have been better, even if it did not feel like it at the time. The global finance crisis had gripped the financial system, and the chip industry – being as cyclical as it is – ground to a halt. Chang, Chairman at the time and a seasoned veteran in the industry, had seen this story play out before though. Unless the world was about to go under, the companies who invested against the capital cycle would come out stronger on the other side.

Chang, 77 years old in 2008, fired his successor and retook the helm of the company, having stepped down in 2005. In the darkest hour, judging by both TSMC’s stock price and the outlook for the real-world economy, he launched a massive multibillion dollar investment program. The Taiwanese powerhouse was not about to lose its market leadership on an unforced error of caution. Of course, this turned out to be exactly the right move. The world did not go under and the market came roaring back, with an even stronger TSMC in pole position to grab market share.

Morris Chang in the the 2008 Annual Report outlook:
“...the global recession will likely result in a long period of slower business recovery. [...] Historically, TSMC has leveraged periodic challenges to become stronger. The challenges of 2009 are clear, and all TSMC employees are committed to do our utmost to emerge from this global economic crisis stronger than ever.”

At the time of TSMC’s initial public offering, the company did $735 million in revenue and $328 million in EBIT. At the time of this writing, those metrics have grown into $75.1 billion and $31.6B respectively – a ~16.5% 30-year CAGR with software-like 40-plus margins – and the Taiwanese chip juggernaut is now the 9th largest company in the world.

That means, in terms of revenue and profits, TSMC is 100 times bigger today than 30 years ago, still growing 25-30%. The fact that this is an extremely complex and capital-intensive business to scale and operate is hard to grasp when looking at TSMC’s income statement.

The company’s cutting-edge fabs are producing millions of wafers with transistors measuring a mere fraction of a micron every year. Depending on the die size and yield, a 300mm wafer can produce anywhere from a few hundred to several thousand chips, which means TSMC is producing chips in the billions every year.

But volume is not everything when talking about the types of chips produced in Taiwan. Leading-edge chips, such as the iPhone and Mac processors, are obviously vastly more expensive than, for example, chips powering a microwave, like the analog power management integrated circuits made by the likes of Texas Instruments.

The relentless race for lower nanometer processes, where "lower nanometers" refers to the shrinking size of transistors on a chip, has driven innovations that push the boundaries of performance, power efficiency, and miniaturization in electronic devices. Lower nanometer technologies allow for more transistors to be packed into a given area, in essence. In 1961, Fairchild Semiconductor’s breakthrough chip, Micrologic, had four transistors. Today, a leading-edge chip like NVIDIA’s H100 Tensor Core GPU has roughly 80 billion transistors on it. As you might have guessed, it can only be made by TSMC.

One of the most hard-to-replicate competitive advantages of TSMC is the economies of scale established through decades of outpacing its competitors, meaning their established fab footprint and the sheer amount of capital they can reinvest in R&D and CapEx. For example, TSMC alone invested closer to $40B in CapEx and R&D in 2023 – a down-year for the industry – compared to the entire U.S. government $53B multi-year stimulus program, the Chips Act.

Building and operating just one single semiconductor fab is an extraordinarily capital-intensive endeavor. TSMC currently operates 20. Just the startup cost of a state-of-the-art fab such as its famous Fab 18 in Tainan, Taiwan, can exceed $15 billion.

Fab employees are forced to put their belongings in see-through plastic bags when entering the facilities and to leave their smartphones outside – this, to ensure the safety of their customers' intellectual property and trade secrets. These fabs, spread across Taiwan and extending into China and the United States, are marvels of modern engineering.

The majority of the cost is related to the installment of advanced manufacturing equipment from specialized suppliers around the world, most notably ASML’s EUV systems. These are sometimes referred to as the heart of the fab – essential for etching the most advanced designs with atomic precision on silicon wafers.

EUV is needed to produce what is known as leading-edge chips, meaning the most advanced, such as Apple’s iPhone and Mac processors, NVIDIA’s data center GPUs custom-designed for the training of LLMs, Google’s quantum computing chips, or Qualcomm’s Snapdragon 5G modem chips. Roughly 90% of these chips, as well as roughly 40% of the world’s new computing power each year, are made in Taiwan according to Chris Miller’s incredible book Chip War.

ASML’s EUV lithography machines – measuring the size of an American school bus and weighing 150 tons – are sometimes referred to as the most complex machine on earth, and without TSMC, Samsung, and Intel, the technology might not have become commercially available. Or, available might be a strong word, it takes immense R&D and CapEx investments, operating efficiency, and scale to be able to operate an EUV machine profitably. Not to mention the price tag of over $200 million a piece.

This means that only a tiny number of companies globally can rationalize investing in these machines. Three of those are TSMC, Samsung, and Intel, which all three helped fund ASML’s last big R&D push to develop EUV in 2012, in exchange for an equity stake without voting rights each. Despite bringing the largest amount to the table back in 2012, Intel could not keep up with the pace over in Asia and did not start using EUV until 2023. TSMC and Samsung were years ahead, already producing chips at scale using ASML’s breakthrough technology.

Apart from ASML, there are a lot of other companies with quasi-monopolies supplying crucial equipment for the fabs. KLA, for example, provides advanced inspection tools that ensure the integrity of the wafers at every stage of production. Applied Materials supplies deposition equipment crucial for layering materials with atomic precision. ASM International offers epitaxy and chemical vapor deposition tools, while BE Semiconductor designs advanced packaging systems that enhance the performance and reliability of the final chips.

Beyond the equipment, the infrastructure surrounding a fab is equally crucial. Fabs require ultra-clean environments, vast amounts of water, and stable power supplies. Specialized gas plants, often operated by companies such as Air Liquide, are constructed nearby to provide the various gasses needed for semiconductor processing. Not to mention the logistics infrastructure needed to ship silicon wafers and chips.

Apart from paying billions of dollars to build a fab, operating it requires constant precision and discipline, especially in the meticulously controlled cleanrooms. These environments are designed to maintain extreme cleanliness, crucial for preventing contamination that could ruin entire batches of wafers which is why workers are forced to wear full-body "bunny suits" seen in the image above. Ruining just one of the wrong wafers, a wafer with one of the most complex chip designs etched into it, can be a $250,000 mistake.

A cleanroom in a semiconductor fab is classified by the number of particles per cubic meter. For instance, a Class 1 cleanroom has no more than one particle larger than 0.5 microns per cubic meter of air, compared to millions of particles in ordinary air. Every aspect of the cleanroom, from air filtration to construction materials, is designed to minimize particle presence.

The stakes are incredibly high every single minute inside a fab. A single speck of dust in the cleanroom can damage a wafer, rendering it unusable. Every step, from photolithography to chemical vapor deposition, must be executed with extreme precision to avoid compromising an entire batch, which could represent millions of dollars in lost revenue. Managing a cleanroom also involves handling materials and chemicals, such as ultrapure water and high-purity gasses.

Morris Chang on TSMC’s culture of discipline:

TSMC’s ability to successfully operate its fabs with such precision and efficiency, meaning ensuring a high yield, a skill that Chang and TSMC now have perfected for decades, at scale makes us think of at least two of Hamilton Helmer’s Seven Powers. Scale economies and process power, which are both described in detail in this article. This has led to an unprecedented, and almost scary, market dominance. As per Q1 2024, TSMC’s market share of the global foundry market, measured in dollars of revenue, amounted to over 60%. Almost six times larger than its closest competitor, Samsung.

It is hard to write about Taiwan’s incredible arrival on the global semiconductor stage without mentioning geopolitics. As the United States and China vie for technological supremacy, both in terms of computing power and advanced weaponry, TSMC finds itself at the epicenter of this high-stakes rivalry.

China's growing assertiveness regarding Taiwan adds another layer of complexity. Taiwan, separated from China via the 130-150 mile-wide ocean strip known as the Taiwan Strait, operates as an independent state, but is viewed by China as a breakaway province that should be reunified with the mainland. This territorial ambition significantly influences global perceptions and policies regarding Taiwan’s semiconductor industry.

Beijing has not only engaged in military posturing and provocative maneuvers but has also leveraged its economy to influence global attitudes toward Taiwan. The prospect of conflict or instability in Taiwan could have truly profound implications for the global economy. Here is what Chris Miller wrote in Chip War regarding this matter:

The U.S. government’s restrictions on Huawei, one of the largest fabless chip designers globally with a long history of working with TSMC, and export regulations for advanced equipment such as ASML’s EUV machines, underscores how strategically important controlling key technology is. This, of course, puts TSMC in a peculiar position.

In response to these geopolitical pressures, TSMC has started diversifying its manufacturing footprint. The chip giant has for example announced plans to invest $40 billion in Arizona, and is about to begin construction on its first European chip plant in Dresden, Germany, in late 2024. The $11 billion facility, named the European Semiconductor Manufacturing Co (ESMC), is a joint venture with Infineon, NXP, and Robert Bosch, each holding a 10% stake.

The demand for more and more advanced semiconductors is insatiable, and industry experts often compare the pace of silicon innovation to a treadmill. Once you fall off, you are quickly so far behind your competitors – whether you provide crucial manufacturing equipment like the atomic layer depositioning systems from Lam Research or are designing breakthrough self-driving car chips like Tesla – that you risk never catching up unless they mess up too.

Practically every technological move forward in society is powered by semiconductors, from the smallest energy-efficiency improvement in next year’s vacuum cleaners to next generation’s breakthrough medicine. Below are six major technology trends that play directly into the hands of leading-edge foundries like TSMC, with its customers and partners battling over the most advanced designs and end-user applications.

Artificial Intelligence and Machine Learning: AI and ML are not just buzzwords anymore, but are now becoming integral to industries like healthcare, finance, and automotive – requiring incredible amounts of computing power to process incomprehensible amounts of data in real-time. A few key components that are crucial for this shift include GPUs for parallel processing, TPUs optimized for deep learning, ASICs designed for specific LLM tasks, and memory chips such as DRAM and NAND.

The transition to 5G: The global rollout of 5G networking is enabling faster data transmission and more connected devices, necessitating advanced chips to manage these high-speed connections and the increased data load.

The rise of cloud computing and data centers: Hyperscalers such as Amazon’s AWS and Microsoft’s Azure are driving demand for data processing and storage at an unprecedented scale. Apart from the usual chip suspects such as memory and advanced processors, the data center buildout is creating strong demand for ancillary infrastructure such as HVAC solutions and cooling, electricity management and backup systems, security solutions, server racks, and automation.

Electric vehicles and autonomous driving: The shift toward electric and autonomous vehicles relies on sophisticated chips for battery management, sensors, and real-time processing, enhancing vehicle performance and ensuring safety.

Internet of Things (IoT) and automation: As more devices become connected and tasks become automated, from smart home gadgets to industrial sensors powering heavy machinery, there is a growing need for energy-efficient chips that can handle complex tasks.

The advancement of consumer electronics: The push for even more demanding smartphones, personal computers, gaming consoles, and other personal devices requires cutting-edge chips to deliver better performance, enhanced graphics, and longer battery life. A nearby niche that is also benefiting from this trend is advanced display making, an industry powered by Samsung’s subsidiary SDC, and leading display manufacturing equipment makers such as Mycronic.

Beyond the impressive numbers and technological advancements, the story of TSMC is also one of human perseverance and bold risk-taking in the face of an uncertain future. Morris Chang’s leadership instilled a culture of excellence and focus that continues to this day.

The company is still maniacal about its so-called Trinity of Strengths; be a technology leader, competitive with the leading IDMs; be the manufacturing leader; be the most reputable, service-oriented and maximum-total-benefits silicon foundry – encapsulated in its motto “Innovative Technologies, Trusted Foundry”. A commitment to pushing the boundaries of what is considered possible while never compromising on the trust of its clients.

The legacy of the now 93-year old Morris Chang – still smoking his beloved pipe, just like he always did in his TSMC office, to the dismay of some of his colleagues – will be forever remembered, having played an indispensable role in the evolution of the most complex and important industry in the world. His unconventional foundry business model helped unbundle the industry, unlocking a remarkable wave of innovation in record time, propelling the likes of NVIDIA and Apple to become the most valuable companies in the world.

"Without strategy, execution is aimless. Without execution, strategy is useless." – Morris Chang.