Home > News content

TSMC: Taiwan's foundry giant who masters the fate of Huawei's chips |

via:博客园     time:2019/6/1 8:32:36     readed:613


Introduction to Meteorite

As the trade war between China and the United States has become more and more serious, there is a Taiwanese company that has received unprecedented attention. This aspect stems from its absolute hegemonic status in the industry. On the other hand, it is also located in Taiwan, in the Sino-US dispute. In a more subtle position, this company is TSMC.

Polishing stones Business Review Author: Wen Dongmei high

Jin Mei Editor

The occurrence of the Sino-US trade war has made the chip industry, which people usually pay attention to, become the focus. Although Huawei started the spare tire program in key areas such as high-end chips and mobile operating systems many years ago, there are still two parts in the supply chain that are difficult to give alternatives in a short period of time. One is based on the X86 architecture. The CPU chip, the other is TSMC's foundry business.

Wafer foundry refers to the production of semiconductor chips. The chip design companies such as HiSilicon, Qualcomm and NVIDIA only design parts. The production of products requires a company like TSMC to do OEM work. Different from the OEM in the general industry, the technical content behind the foundry of the chip industry is very high, and the evaluation standard is mainly the level of the process technology.


Now the most advanced in the world is the 7nm process. Only TSMC and Samsung can do it, and TSMC has an absolute advantage in market share. The 7nm mobile phone chips of giants such as HiSilicon, Qualcomm and Apple are all manufactured by TSMC.

In this industry, there are still some head enterprises such as TSMC, Samsung, Intel, UMC, Grofund and SMIC. Among them, UMC and Grofund have already announced the abandonment of high-end technology below 12nm. Abandoning the 7nm process, but the 10nm process has always been low. SMIC is the only mainland Chinese company in this area. It just achieved mass production of 14nm in January this year, which is two generations worse than the 7nm of TSMC, which represents the world-class level.

Therefore, the attitude of TSMC is very important to Huawei now. Therefore, TSMC said on the evening of May 17 that “there has been a complete system in place. After initial evaluation, it should be in compliance with export control regulations and never change shipments to Huawei. It is planned that Huawei will continue to be shipped, but the follow-up will continue to be observed and evaluated.

Some people think that TSMC will fully support Huawei's over-optimism. TSMC is now making such a statement because the United States has not imposed extreme pressure on it. Once the United States imposes extreme pressure on it, it will be difficult to say whether TSMC can stand up. After all, 80% of its equity is controlled by foreign institutions. The shareholder is Citibank of the United States.

And once TSMC joins the ban on Huawei under the pressure of the United States, Huawei's high-end mobile phone chips will be cut off. This will be a great blow to Huawei and the worst situation that may occur in the US ban on Huawei. This correlation analysis has been a lot. What we want to write today is why TSMC can become the absolute overlord in the field of foundry.


TSMC is the world's largest foundry company. Founded in 1987 and headquartered in Hsinchu, Taiwan, it is the world's first integrated circuit manufacturing company focused on foundry and the founder of the foundry model.

After more than 30 years of development, its market value has reached nearly 200 billion US dollars, accounting for more than 50% of the global market share. In addition, as mentioned above, the company's existing state-of-the-art 7nm process technology is in the position of the industry leader, only Samsung can match.


TSMC has dominated the global market and has secured the absolute status of the industry. It has a great relationship with its founder and Morris Chang, who is known as the semiconductor godfather of Taiwan.

Zhang Zhongmou graduated from the Massachusetts Institute of Technology. After graduating, he entered the semiconductor industry, which was just started. He joined Texas Instruments at the age of 27, and he was able to grow into the company's “three-handle” and senior vice president. Later, he did not agree with the company's strategic differences in the transition to the consumer electronics field. He returned to Taiwan in 1985 and founded the Taiwan Integrated Circuit Manufacturing Company at the age of 55 in 1987, referred to as “TSMC”.

At that time, the semiconductor industry giants were all in the IDM model, that is, a company had to complete the entire industry chain from design, wafer fabrication to packaging and testing. The investment required was huge and the threshold was extremely high. Industry companies were afraid of designing patents and other trade secrets. Competitors have basically not shared the wafer production facilities, which leads to repeated investment and overall resource waste. The industry cannot form a scale and promote the reduction of chip costs.

Zhang Zhongmou saw this problem and seized the opportunity to subvert the traditional rules of the game and position TSMC to do wafer fabrication services for semiconductor companies. At the beginning of its creation, it was “world-class”. After this model emerged, it greatly reduced the entry barrier of the chip industry, making a number of independent fabless chip design companies including today's chip industry giant NVIDIA appear one after another.

At the beginning of the founding of TSMC, the semiconductor industry was in a downturn, and the growth rate of the semiconductor market in the world was rapidly declining. Without a market foundation and a new business model that has not yet been recognized, TSMC can only survive on a small number of orders in the first year or two. Business people often run around and have no new customers for a whole year.

In the face of adversity, Zhang Zhongmou's personal leadership played a vital role. He demanded that “everyone should set goals, achieve, set a higher goal, force self-reach”, and be good at motivating. Let employees “create new courage, new confidence, and re-sprint”.


As a big business in the semiconductor industry, Zhang Zhongmou has accumulated a lot of important resources in his 30-year career, which is one of the important reasons why TSMC can develop. In 1988, Zhang Zhongmou, who was then the chairman of the Taiwan Industrial Technology Research Institute, and Dyke, the newly appointed general manager, personally invited the CEO of the old friend Intel and the author of "Only paranoid to survive" Andy Grove. Taiwan came to investigate and accredit TSMC.

Grove is well known for leading Intel to survive many times of hardships, turning it into the most self-reliant company in the technology world, and bringing huge returns to investors. Americans have no tradition of opening up to friends because of personal relationships. Therefore, during the inspection, Intel checked the semiconductor's more than 200 processes one station at a time, and picked out more than 200 problems requiring TSMC to improve immediately. Moreover, once they pass the certification, they must first pass their permission to change the machine and change the process.

At that time, there was no uniform standard in the industry. Getting Intel certification means getting world-class certification, which is the best endorsement of TSMC's production capacity. With his consistently strong style, Zhang Zhongmou pursues perfection in the implementation of the strategy, adheres to high standards, strict requirements, and quickly improves and gives the results of Intel's approval, and finally won the certification and orders.

Moreover, after this campaign, Zhang Zhongmou also united the people within the company, establishing international standards, world-class management foundation and operating environment. After the development of the semiconductor industry began to pick up, Zhang Zhongmou seized the opportunity to attack the fire line. In 1994, he resigned as the chairman of the Industrial Research Institute and devoted himself to business operations. He led TSMC to the market on the Taiwan Stock Exchange and quickly expanded its scale. A company that makes money.

Looking back at the development process of TSMC at that time, now that it can develop, in addition to Zhang Zhongmou's personal experience and ability, he also benefited from two external conditions: First, before the founding of TSMC, there was already a sprout of the foundry model. At that time, many IDM companies came out of the door to design, lack of money to build Fab, and handed it to IDM to worry about the design solution leak, so there was a demand for OEM production. Second, the model conditions for pure foundry at that time were not yet mature, and the possibility of development was not obvious. Large semiconductor manufacturers did not notice that they left an opportunity for a startup like TSMC.

Like all disruptive innovation models, TSMC seized the opportunity to grow wildly. In 1995, revenues exceeded $1 billion. In 1997, Zhang Zhongmou put TSMC on the US New York Stock Exchange and achieved $1.3 billion in revenue that year. , a huge profit of 535 million US dollars. In the following years, because of the optimism of the outside world, both TSMC and Zhang Zhongmou have received numerous praises and various honorary titles.


From the quick opening of the market after obtaining Intel certification to becoming the most profitable company in Taiwan, the first ten years of TSMC's establishment have been relatively smooth, but the real thing that makes it realize the accumulation of technology is leading to the industry in 1990, which is based on IDM. ; Mountain plan & rdquo;. At that time, each IDM had its own process standard. In order to obtain these IDM orders, TSMC must adjust the process one by one according to the needs of different customers.

In this process, TSMC follows three steps: one is to use IDM's process technology; the other is to combine technologies developed independently; and the third is to jointly develop new technologies with TSMC and IDM. Through these stages, the TSMC has gradually completed the technical accumulation and has gathered a large number of outstanding talents.


After 2000, the rapid growth of design-only, Fabless expanded the foundry market. The number of Fabless in the top 20 semiconductor companies increased from 0 in 2000 to 4 in 2011. The increase in market demand has enabled the foundry business to enter a more rapid growth period.

Around 2000, Internet and technology companies disregarded earnings and blindly expanded. After the bubble burst, not only did the stock price collapse, but also the related entities. TSMC, which had a gross margin of 46% in 2000, was also affected. In 2001, revenue fell by 24% year-on-year, and net profit fell by 78% year-on-year. However, due to maintaining a sound financial position, it returned to 29% in 2002. Growth, with more than $5 billion in annual revenues, entered the top 10 semiconductor industry and became the world's largest professional foundry company.

While rapidly developing the company, Zhang Zhongmou is also constantly concerned about industry trends and eliminating possible risks. In 2000, Zhang Yujing, who was born in Texas Instruments, founded Shida Semiconductor. The development momentum was too fast. Zhang Zhongmou, who had a strong style of work, persuaded the largest shareholder of the World Bank to acquire it for $5 billion without knowing Zhang Yujing.

Zhang Yujing, who was on the north of the negative gas, set up SMIC in Shanghai and successively dug people from TSMC and the UN General Assembly. The digging staff revealed TSMC's trade secrets to SMIC, which allowed it to have four 8-inch and one 12-inch fabs from scratch in less than four years. In the face of SMIC's encroachment, TSMC filed a lawsuit against these employees who had been excavated. In February 2005, SMIC received a settlement fee of US$175 million. After November 2009, the agreement will receive SMIC. The annual payment of 200 million US dollars in cash and 10% of the shares, and forced Zhang Yujing to leave.

In the second decade after its establishment, TSMC's technological processes gradually took the lead. At the end of the 1990s, TSMC's process technology was still far behind Intel, but it was closer to the giants in the fast catch-up. In 1999, TSMC led the industry in the production of 0.18 micron copper process manufacturing services for commercial production. In 2001, TSMC introduced the industry's first reference design process to help customers developing 0.25 micron and 0.18 micron designs to reduce design barriers to achieve rapid mass production. In 2005, it was the leading manufacturer of 65nm chips.

As wafer tip technology evolved from 90nm to 65nm, costs began to rise, and many major companies including ST, Infineon, NXP, and Freescale stopped investing in advanced fabs. Texas Instruments also announced in 2007 that it would abandon the development of processes below 32nm. In 2009, when the wafer process was transferred from 40nm to 32nm, Fujitsu, Panasonic, Renesas, Toshiba and Sony all turned into light fabs, which created opportunities for TSMC.

By 2005, TSMC had achieved revenues of up to US$8.23 billion, net profit after tax of US$2.91 billion, and continued to dominate Taiwan with a gross margin of 43.6% in the face of increasingly fierce competition and lower overall profit margins. The largest industry company in the position of the most profitable company. TSMC, which has a good operating performance, continues to increase its R&D investment and is heading towards the industry.

However, during the 2008 financial tsunami, TSMC was hit, and revenue for the entire year of 2009 fell by 11.2% year-on-year. In response to the financial crisis, CEO Cai Lixing conducted cost control through layoffs and caused labor disputes. Zhang Zhongmou, who had retired for four years, moved the board of directors and dismissed Cai Lixing. On the occasion of the financial turmoil in June 2009, he returned to the front line as the company's CEO.


As the founder of the company, Zhang Zhongmou has a strong appeal to all employees and has the ability to stabilize the situation. As a heavy industry heavyweight, his influence on customers is also great. His comeback has enabled TSMC to recover quickly during the financial crisis. Revenue rose to 41.9% year-on-year. In 2012, it achieved the best performance in history, with revenues of US$17 billion, ranking first among Taiwan-listed companies.

Zhang Zhongmou, who started from the beginning of 54 years old, once again stood in the world semiconductor industry. In the era of TSMC, he was the person who created the industry. Later, during the development of TSMC, he created two industries of himself and his customers. He led the fabless professional design with the wafer foundry industry, which made the industry flourish.


After Zhang Zhongmou became the CEO, he drastically promoted the company's major decisions. First, he increased the capital expenditure in 2010 by a large stroke and increased it to 5.9 billion US dollars. With the gradual rise of mainland mobile phone brands, TSMC's production capacity has become in short supply, especially the 28nm process is often queued for purchase.

At that time, in order to keep the opponent far behind, TSMC launched four versions of the mainstream process for smartphone chip design, which was revised almost every year to make the opponents chasing after. The newly introduced process is not only faster than the previous generation, but also saves 30% of power, and the chip scale is still reduced by 4%. Through the "four-in-one" system, TSMC established a very high barrier in the 28nm battle, and 28nm has become its most powerful process.

After the establishment of Grofund, he tried to seize the market. The bottomless power combined with the advanced technology of the United States once made the outside world worry about the future of TSMC. However, I did not expect that the drums had just been knocked and the winners and losers had been divided. The 28nm process has a key technology choice —— Gate-first and Gate-last, Grofand and Samsung both choose the first gate, and TSMC uses the back gate. As a result, TSMC successfully mass-produced in one fell swoop, and the production yield of Samsung and Grofund was never improved.

The second battle opened after Zhang Zhongmou's return was a confrontation with Samsung. After the rise of the foundry model pioneered by TSMC, there was a foundry fever around the world, and many semiconductor manufacturers participated. Currently, among the top eight foundries in the world, Samsung is the only IDM manufacturer and the largest IDM foundry. Foundry is one of Samsung's core business. Over the years, it has worked hard with TSMC to compete for the right to speak in the advanced process technology market. It has also been a topic of concern to the industry.

As the IDM of the Big Mac, Samsung has always felt that TSMC's foundry business is not good enough. Through strong investment, independent foundry business, and digging people, it has continuously challenged TSMC. In 2015-2016, Samsung won many large customer orders from TSMC, and the foundry business has seen significant growth. In 2016-2017, with the further maturity of TSMC's advanced processes, some of Samsung's OEM orders were robbed by TSMC.

At the beginning of 2018, at the Samsung Foundry Forum held in Seoul, South Korea, the relevant person in charge of Samsung said that the goal of 2018 is to increase the market share of foundry from the fourth place to the second place, surpassing UMC and Core, the future is going to surpass TSMC to become the industry's number one. However, in 2018, Samsung's foundry business achieved only a 4% micro growth.


As a high-tech heavy asset business, the amount of money that foundry needs to invest is huge. The capital expenditures of TSMC and Samsung account for 70% of the global foundries, while the TSMC family accounts for half of the total. The research and development cost of Samsung's 90nm process node is 280 million US dollars, and the research and development cost to 20nm has soared to 1.4 billion US dollars, which does not include the new production line production costs and construction costs. The investment of TSMC is even greater. Therefore, advanced process development has gradually become a giant game.

As a result, there were 22 manufacturers with 130nm process capability in the world, and the number of manufacturers capable of foundry with 16/14nm process technology has dropped to five. With 10nm, 7nm and more advanced process technology capabilities, only TSMC, Samsung and Intel are left.

In terms of advanced processes, TSMC is in an absolute leading position, with Samsung ranking second. As early as 2011, TSMC conquered the 28nm process, and the advanced process developed rapidly. Its 7nm process was mass-produced in the second half of 2018. It is expected that the revenue in 2019 will exceed 20%. 5nm has undergone risk trial production in April this year and is expected to achieve mass production by 2020. Samsung has announced that it will start production in 7nm in the second half of 2018, and advanced processes of 5nm and below are also planned.

As the two giants in the monopoly of the industry, TSMC and Samsung are the only ones in the world to have the mass production capability of 10nm process technology. As an all-round IDM manufacturer, Samsung has a certain competitive relationship with its OEM customers, owning the mobile phone market and Researched Exynos series SoC chips as a bargaining chip, have certain control ability for cooperative customers. Mobile SoC chips are the primary application area for top-of-the-line processes. Overall, TSMC and Samsung monopolized the SoC chip foundry using advanced process technology.

From a technical perspective, TSMC and Samsung are on a par with each other. It is expected that TSMC's 7nm EUV version (N7+), which is mass-produced in 2019, will be similar to Samsung's 7nm parameters. At present, TSMC's original 7nm (not using EUV) has been mass-produced, and it is the most advanced process in the market. It has a first-mover advantage in time.

In terms of EUV use, TSMC and Samsung are the top two subscribers to ASML, the equipment manufacturer. At present, TSMC has the most EUV equipment, with 10 units, followed by Samsung and 6 units. EUV is a necessary process equipment for processes below 7nm, which is of vital importance to the latest process production of the manufacturer. Due to the high precision requirements, TSMC also has in-depth cooperation with ASML in research and development.

TSMC's iterative speed is fast. After the semiconductor manufacturing technology enters the 14/16nm node, the development difficulty and capital investment of the FinFET process technology are greatly increased, so it is also regarded as the access standard for advanced process technology. TSMC's 28nm process has been in volume production since 2011, leading competitors for 3-5 years. Since 2014, it has started mass production of 16/20nm chips, and then entered a period of rapid growth. By 2015, the proportion of the two processes has reached 49. %.

In order to give full play to its technological advantages, TSMC attaches great importance to the rapid expansion of advanced manufacturing processes after mass production. For example, after TSMC's 130nm process was put into mass production in 2003, its revenue share rose sharply from 0 to 28% in just one year; the 28nm process revenue ratio was only used after mass production in 2011. It has climbed from 2% to 22% in one year.

Rapid expansion of advanced production capacity will help TSMC quickly seize customer resources and expand first-mover advantage at every advanced process node, and its production capacity structure will be significantly better than its competitors. Thus, higher product added value will bring higher Gross profit margin. Under the IDM model, Samsung's foundry division is located in the semiconductor segment and belongs to the third-tier business. The resources available are relatively limited. Even if the intellectual property rights and patents are well protected, the supply chain cannot be guaranteed. And avoid the impact of competition from the upper level.


For example, Apple's A4-A7 series processors are all OEMs in Samsung. After a series of patent litigations broke out in 2011, Apple transferred A8 to Taiwan Semiconductor Manufacturing Co., and A9 was handed over to TSMC and Samsung. A10 was all owned by TSMC. work. For the above reasons, although Samsung's foundry department is indistinguishable from TSMC in the progress of process technology, its background has determined that it is difficult to become a giant in the field of foundry.

In the competition with rivals, TSMC finally grew into an industry giant. In 2018, TSMC's consolidated revenue for the year was US$34.2 billion, up 6.5% year-on-year; net profit after tax was US$11.64 billion, up 3.3% year-on-year, and its market share was close to 60%. This year, the company's revenue, net profit and earnings per share set a record for seven consecutive years, and successfully mass production of 7nm process, leading other peers for at least a year.


TSMC can become the absolute hegemon of the foundry industry. The following points are worth learning:

The first is the positive and progressive culture of TSMC. In 2014, in order to effectively improve the speed of research and development of 10nm advanced processes, TSMC internally launched the “Nighthawk Project”, which arranged the R&D personnel to implement the 24-hour three-shift rotation without any suspension, which greatly improved the R&D efficiency. At the end of 2016, 10nm began mass production. In 2017, the climb began to climb, and by the fourth quarter, the proportion of revenue had reached 25%.

Secondly, it is willing to invest, continue to expand the scale of research and development, and ensure the status of its industry leader. In 1992-2017, TSMC's annual R&D expenses increased rapidly from US$5 million to US$2.7 billion, both in terms of quantity and growth rate far exceeding that of other competitors. Due to the increasingly complex and difficult technical challenges faced by Moore's Law, TSMC's ever-expanding R&D scale ensures that it can continue to provide customers with leading process technologies and solutions. The ever-expanding capital expenditures have helped the company build up barriers to competition and further stabilize and expand its dominant position.

TSMC currently has three 12-inch large fabs with a total capacity of more than 7 million wafers, offering full-generation and half-generation designs from 0.13 micron to 7nm, while retaining some of its capacity for R&D applications, supporting 5nm and Technological development of more advanced processes.


The third is excellent management. The law of economics is that too large a scale will restrict the profit level. The semiconductor foundry industry is highly intensive in technology and capital, and as the industry develops, it will show a more concentrated trend. As a leading industry leader, TSMC can maintain a gross profit margin of about 50%. Its stable R&D expenditure level, control of excellent operating cost levels, and good management and management levels are definitely beyond the reach of most listed companies.

The fourth is to take advantage of the technological advancement in advanced processes, take the lead in obtaining high interest rates and expand, and then wait for the latter companies to catch up, then rely on the scale advantage to fight the price war, so cycle, far away from competitors, Lay your own industry dominance.

In addition to the advantages of daily operations management and technology research and development, there are two other points to help TSMC develop and climb the industry. First, TSMC received support from the Taiwan authorities at the start. In 1987, under the impetus of the Taiwan authorities “Government”, Taiwan “National Science Association” invested US$100 million to create TSMC with Philips and some private capital. When TSMC was founded, it needed to build a 6-inch wafer fab. At that time, it needed to invest 200 million U.S. dollars, and there was a funding gap in addition to the Taiwanese development fund. TSMC got a share in the Dutch company Philips and provided technical assistance.

Second, one of the fundamentals that TSMC can lead the world is that they set a principle of never competing with customers. Looking back at Samsung's loss of Apple A10 orders, in addition to the performance on the A9 is not as good as TSMC, there are rumors that Apple is worried about Samsung stealing its related design, so terminated the cooperation with Samsung.

In addition, advanced packaging technology is used in conjunction with new technology research and development. The Apple A10 processor released in 2016 uses TSMC's FOWLP technology, which brings a new incentive to semiconductor practitioners and will bring good benefits to TSMC.

With the outbreak and deepening of the Sino-US trade war, semiconductors, especially the chip industry, have received unprecedented attention. In the future, China will need to take measures to accelerate the development of this field. As the absolute overlord of the industry, TSMC's establishment and development process has many places worth learning and learning from our company.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments