Independent development path of semiconductor enterprises under the national system

Latest update time：2018-05-08

Reads：

Source: Content from Wang Cong/Zhang Tianwen of Guojun Electronics.

As we continue to visit major domestic semiconductor projects and deepen communication with the industry chain, we increasingly realize that due to the large number of stakeholders, the development of China's semiconductor industry is not simply an economic or information security issue, but more of a management issue, a government-enterprise relationship issue, and a problem of interest game (which is probably also a problem faced by Japan, South Korea, and Taiwan in the past). Therefore, as a mainstream semiconductor analysis team in mainland China, one of the main directions of our future research will revolve around the above topics and may not be limited to investment itself.

This article focuses on the independent development path of semiconductor enterprises under the national system. The so-called national system in this context refers to the full support of the government. On the one hand, the development of semiconductors cannot be separated from the strong support of the government, but in the end it still depends on the comprehensive strength of the enterprise. In an organization with government management as the core and a complex internal personnel structure, it is easy to have no one responsible, no one managed, and redundant personnel.

We have carefully analyzed the history of semiconductors (mainly Japan, South Korea, and Taiwan) over the past 50 years, using a large amount of data. We believe that mastering core technology research and development , improving cost depreciation awareness, and improving investment decisions are the two core elements, and more importantly, the wolf culture of semiconductors (manifested in a keen sense of the market, a strong desire to seize the market, the ultimate pursuit of speed and efficiency, and a strict accountability system).

1. The development of the semiconductor industry requires the spontaneous role of enterprises to be played with strong support from the government

The history of the rise of Japan and South Korea through "government will + enterprise investment + technology introduction" shows that the development of the semiconductor industry is a process in which enterprises give full play to their role as market entities and their subjective initiative with strong support from the government, and neither of these can be missing.

On the one hand, since the semiconductor industry is a capital and technology intensive industry that requires continuous huge investment, is high-risk and highly cyclical, it is difficult for a company that operates purely on market rules to survive on its own. Therefore, its development cannot be separated from the strong support of the government.

For example, Japan has implemented the "government, industry and academia" trinity. From 1976 to 1979, the Japanese government, together with five major companies including Fujitsu, Hitachi, Mitsubishi, NEC and Toshiba, as well as the Electronics Research Institute of the Japan Industrial Technology Research Institute, invested 72 billion yen to establish the national R&D project "VLSI (Very Large Scale Integrated Circuit) Program" . The Japanese government provided about 28.4 billion yen in funds for the project, which is equivalent to 2-3 times the annual R&D expenses of the five major companies participating in the plan. This government-industry-research project has been a huge success, with 1,210 patent applications, which is one of the most important reasons for the rapid rise of Japan's semiconductor industry from the early 1970s to the mid-1980s. In addition, Japan also protects its newly started memory companies in terms of trade strategy by raising tariffs, restricting quotas, and purchasing local products .

Another example is South Korea, which implemented the chaebol system of "industry and commerce". From 1986 to 1993, it implemented the eight-year "VLSI Technology Joint Development Plan". In 1986, the South Korean government invested heavily in the research and development of 4M DRAM as a national key project. Led by the Korea Electronics and Telecommunications Research Institute, Samsung, LG, Hyundai and six Korean universities jointly tackled technical problems. In three years, the project's research and development costs reached 110 million US dollars, and 57% of the investment was borne by the government. By 1989, the time difference between South Korean and Japanese companies in launching 4M DRAM on the market was successfully reduced to zero, marking the elimination of the technological gap between South Korea and Japan.

On the other hand, as the main body of market competition, semiconductor enterprises dominate R&D, operation, and management. The competitiveness of a country's semiconductor industry ultimately depends on the comprehensive strength of the enterprise. Especially considering that most semiconductor projects with full government support have a strong government background, and the government itself often does not have a full understanding of the semiconductor industry, nor does it have the experience and ability to manage enterprises. At the same time, for latecomer countries, the introduction of technology is often accompanied by the introduction of talents, but in an organization with government management as the core and a complex internal personnel structure, it is easy to have no one responsible and no one managing. Only when private enterprises fully participate in management can the government's resource support and the enterprise's management capabilities be effectively combined. Therefore, this article focuses on the independent development path of semiconductor enterprises under the national system.

2. Independent development path of semiconductor enterprises under government support

With the strong support of the government, enterprises can play an active role in mastering core technology research and development, improving cost depreciation awareness and improving investment decisions, and cultivating a wolf culture.

2.1. Increase R&D investment and master core technology autonomy

Having an independent core technology system is an indispensable foundation for the development of the semiconductor industry.

For example, Japan relied on the VLSI project of government-industry-academia integration in the 1970s to gradually narrow the technological gap with the United States and surpassed the United States in the 256K DRAM era. For example, in the 1990s, the three major chaebols in South Korea invested heavily in 58 R&D centers and doubled R&D expenses to gradually get rid of South Korea's technological dependence on the United States. All of these reflect the importance of core technology autonomy.

On the other hand, Taiwanese companies, whether they are foundry companies represented by TSMC or technology importers represented by Nanya Technology, prefer a "short, flat and fast" profit model. TSMC has been constantly acquiring design companies that have independent R&D capabilities but have fallen into losses due to industry cycles, such as Taiwan's first professional DRAM manufacturer, Acer Semiconductor, Taiwan's only company capable of DRAM industry technology R&D, World Advanced, Powerchip, and World Semiconductor, etc., and transformed them all into wafer foundries, which to a certain extent stifled the independent R&D capabilities of Taiwan's DRAM industry. In addition, insufficient investment in R&D expenses is another reason why Taiwan has failed to form an independent core technology system. Compared with the average R&D intensity of 15-20% of revenue for Japanese and Korean DRAM giants, the R&D expenses of Taiwanese companies account for only 6% of revenue. This is related to Taiwan's annual payment of more than NT$20 billion in technology licensing fees to Europe, the United States and Japan, as well as huge investments in imported equipment.

Independent mastery of core technologies is the source of corporate competitiveness. If Japan and South Korea do not take the initiative in technology through cooperation between the government and the industry and academia, and do not expand vertically into the materials and equipment industries along the industrial chain, they will likely be crushed in the recession cycle, just like Taiwan's semiconductor companies that were in deep trouble during the financial crisis.

2.2. Improve awareness of cost depreciation and price competition and improve investment decisions

On the one hand, as large-diameter silicon wafers achieve micro-processing, the price of semiconductor production equipment is getting higher and higher, and equipment depreciation has become the most important cost component in semiconductor production. Therefore, it is particularly important to raise awareness of cost depreciation.

A negative example is that Japanese technicians often consider technology and cost separately. They focus on controlling operating costs (variable costs), but have a weak awareness of capital costs (fixed depreciation costs), believing that it is a management issue rather than a technical issue. For example, after building a new semiconductor factory, Japanese semiconductor companies will first clean the factory thoroughly inside and out, then carefully and slowly move in equipment, and only start semiconductor production after the machines are running normally and everything is properly arranged. They don't care about the depreciation cost of the invested capital during this period. Producing high-performance, high-priced products is taken for granted.

Korean semiconductor companies take a completely different approach. They install machinery and equipment right after the factory is built and roughly cleaned. They keep the clean room running around the clock and use high-frequency filters to produce products against the clock so that the invested capital can be put into use as quickly as possible, reducing the unit product depreciation cost and thus the price.

In addition, Japanese semiconductor companies pay great attention to the performance details when purchasing production equipment, and have an almost persistent pursuit of improving the yield rate, while Korean and Taiwanese companies pay more attention to the throughput (processing volume per unit time) and utilization rate of equipment. In fact, if depreciation costs are taken into account, as long as the throughput of factory equipment is increased, even if the yield rate decreases, the total cost can be reduced and a price advantage can be obtained.

On the other hand, the strength of cost depreciation and price competition awareness will also affect investment decisions, and thus affect corporate profits.

Taking Japanese IDM companies as an example, Figure 5 reflects the annual changes in the total equipment investment and total integrated circuit sales of 12 semiconductor companies in Japan from 1975 to 1999. It can be clearly seen that there is a positive correlation between the two. When sales increase, the equipment investment amount of the year will also increase; when sales decrease, the equipment investment amount will also decrease accordingly.

Figure 6 shows the annual changes in the sum of depreciation expenses and the sum of integrated circuit sales of the same 12 semiconductor companies. The two show an inverse relationship from time to time. This is because depreciation expenses lag behind equipment investment. When sales fluctuate downward, the depreciation burden will increase. For example, in 1985-1986, sales fell sharply, but depreciation expenses increased significantly due to increased equipment investment in the previous years when the market was booming. As a result, the huge depreciation burden was superimposed on the subsequent sales downturn, making the situation even worse. In the boom cycle, equipment investment is made in direct proportion to sales, but in the non-boom cycle, depreciation expenses and sales change inversely. This situation has occurred repeatedly and has become an important factor affecting the stability of corporate earnings.

On the contrary, Korean companies represented by Samsung are good at actively taking advantage of crises to carry out counter-cyclical investment operations. When the market falls into a depression cycle, they expand production on a large scale, use their production capacity advantages and fierce price wars to squeeze competitors out of the market, and thus seize the opportunity and make profits before the boom cycle arrives. "The more difficult it is, the more we need to increase investment," said Lee Kun-hee, former chairman of Samsung Group.

In the mid-1980s, the price of DRAM chips continued to fall. When Samsung launched 64K DRAM in 1984, its price had plummeted from $4 per chip to 30 cents per chip. Based on Samsung's production cost of $1.3 per chip at the time, the company would incur a loss of $1 for each chip sold. However, when Intel withdrew sadly and Japanese companies such as NEC drastically cut capital expenditures, Samsung went against the trend and not only continued to expand production capacity, but also invested in the development of larger capacity DRAM. By the end of 1986, Samsung Semiconductor had lost $300 million and its equity capital was zero. Subsequently, Japan and the United States signed the first semiconductor agreement in 1986, DRAM prices rebounded, and Samsung began to turn losses into profits.

During the 2008 financial crisis, DRAM prices plummeted by 90%. Samsung once again counter-cyclically used 118% of the previous year's profits to expand production capacity, launching a more intense price war against competitors. DRAM prices fell further and quickly fell below the cost of materials, which directly led to the bankruptcy of German manufacturer Qimonda and Japanese manufacturer Elpida in 2009 and 2012 respectively. Elpida was acquired by Micron at a low price. From 2011 to 2012, the demand for DRAM from PCs and other industries weakened, the market heat faded again, and revenue declined. However, Samsung once again increased its investment, with capital expenditures of no less than US$17 billion in two years, taking the lead before the industry recovery in 2013 and further squeezing out competitors. Taiwan ProMOS went bankrupt in 2012, and Inotera was acquired by Micron as a wholly-owned subsidiary in 2015. Since then, there are only three major players in the industry: Samsung, SK Hynix, and Micron, of which Samsung and SK Hynix occupy 75% of the market share.

Behind the differences in investment decisions, in addition to the different market competition attitudes of enterprises, there is actually a gap in the long-term grasp of economic cycles and market demand. It can be said that behind Samsung's several gamble-style counter-cyclical investments is its accurate grasp of market demand and competitive landscape. The foresight behind this wolf-like and bold investment philosophy is worth learning from.

2.3. Cultivate an enterprising wolf culture

Wolf nature emphasizes enterprising spirit and aggressiveness, and does not admit defeat easily. Reflected in industrial development and business operations, "wild" corresponds to the spirit of pioneering and enterprising, "cruel" corresponds to the courageous determination to achieve the goal at all costs without fear of competition, "greed" corresponds to the persistent pursuit of continuous progress and never stop, and "violent" corresponds to the high morale of ruthlessly overcoming all difficulties. In addition, wolves have an extremely keen sense of smell and the habit of fighting in groups, which correspond to market insight and team spirit in business management. In the development of the semiconductor industry, wolf culture is manifested in a keen sense of the market, a strong desire for market expansion, an extreme pursuit of speed and efficiency, and a strict accountability system.

First, we must have keen market insight, always pay attention to market trends, and take market demand as the source of technological innovation.

Taking the comparison between Japanese and Korean semiconductor companies as an example, Japanese semiconductor companies generally have the problem of over-emphasizing quality details and ignoring changes in market demand. Many Japanese company managers and technicians do not pay attention to consumer demand and refuse to invest in collecting market information. Samsung, a Korean company represented by Samsung, proposes that competitiveness must change with the times, environment, and market changes. In the memory business department of Japanese companies, there are no more than 10 dedicated marketing personnel, and their status within the company is very low, while in the corresponding business department of Samsung Electronics, there are 230 dedicated marketing personnel. The dull market sense has caused the market share of Japanese companies to be rapidly eroded by South Korea. The decline of DRAM is a typical example.

Japanese companies once had great success in the long-life DRAM market for mainframes. However, after IBM launched the personal computer in 1984, personal computers gradually surpassed mainframes in the market. Mainframes require a long service life for DRAM and are expensive, while personal computers require a lower service life for DRAM than mainframes and have stricter price requirements. Micron Technology of the United States was the first to respond to this, followed by Korean companies led by Samsung Electronics, and each company mass-produced DRAM for personal computers. However, Japanese DRAM companies did not take active countermeasures, and the long-life, high-priced DRAM products made in Japan were not well received in the new market, and were suppressed by Micron Technology and Samsung Electronics with their short-life but low-priced products for personal computers.

As can be seen from Figure 8, Japan's DRAM share has been declining since the mid-to-late 19th century, from 80% at its peak to less than 20% in 2010, while South Korea's DRAM share has been rising, with a market share of nearly 60% in 2010, replacing Japan as the world's number one.

In fact, technological innovation is rooted in two factors: supply and demand. As the life cycle of new products in the semiconductor industry continues to shorten, corporate technological innovation is more affected by the ever-changing demand factors than long-term supply factors. In order to remain invincible in the fierce competition, the "wolf spirit" of keen sense of smell, good at catching and quick response is indispensable.

Second, we must actively expand the market.

During the golden period of semiconductor development in Japan, it had a vast market in the United States and its home country. South Korea first had the domestic market opened to it by the United States, and then entered the Chinese mainland market. Take SK Hynix as an example. During the Asian financial crisis in 1997, Hynix sold its TFT-LCD division to BOE for US$380 million. In 2004, Hynix and STMicroelectronics built a 12-inch wafer factory in Wuxi. The project with a total investment of US$2 billion received a 1 billion subsidy from the Wuxi Municipal Government. During the 2008 financial crisis, Hynix quickly turned losses into profits thanks to the strong demand in the Chinese market and the smooth production of the Wuxi factory, while Taiwan's DRAM industry has since fallen into decline.

Coincidentally, the counter-cyclical investment of Samsung mentioned above is also based on the huge markets of the United States and China. Actively expanding markets with strategic depth is crucial for enterprises to diversify their operating risks and improve their core competitiveness.

In contrast, after the Democratic Progressive Party came to power in Taiwan, it refused to move westward into the mainland. The lack of strategic depth in the mainland market was one of the reasons why it fell into deep trouble during the financial crisis.

Third, we must focus on strengthening personnel management and improving execution speed and operational efficiency.

For example, TSMC began to implement the "Nighthawk Plan" in order to develop the 10nm process, that is, the three-shift system in the R&D department. A R&D manager who has worked at TSMC for more than ten years mentioned, "TSMC's competitiveness is built on high-intensity manpower and time investment. The Nighthawk Plan is a plan proposed based on the reality that the overtime work of R&D employees is too serious. The purpose is to rationalize and institutionalize the working hours of the R&D department and complete the work through day and night shifts."

For example, Koreans are obsessed with speed. In 1983, Samsung began investing in the semiconductor industry and completed the construction of a VLSI (very large-scale integrated circuit) mass production plant that year, successfully achieving fully independent production, assembly and inspection of 64K DRAM, shortening the gap between Korean semiconductor technology and that of the United States and Japan from more than 10 years to 2-3 years.

"Samsung's 60-Year History" records: "Semiconductor factories need to produce ultra-precision products that meet the requirements of microns, and such factories are built for the first time, and they must ensure high production yields. The requirements of the project are very strict." At that time, Samsung employees carried out high-intensity construction 24 hours a day without vacations under extremely cold conditions. It took only 6 months to complete the entire line, which was about one-third of the industry average. When the imported equipment was transported to the newly built silicon wafer assembly line, the project staff found that the road from the highway to the assembly line was not paved, and the road surface was paved a few hours later. In 1989, only 6 years after South Korea officially entered the semiconductor industry, Samsung successfully mass-produced 4M DRAM and put it on the market almost at the same time as Japanese companies.

Samsung's speed is not only reflected in construction, but also in project technology research and development. Since Lee Kun-hee did not hesitate to double the research and development budget and hire the best software engineers, Samsung can often shorten the research and development time to half or even more of Japanese companies. For example, Samsung launched the Galaxy S3 in just 3 months, while the development of an ordinary smartphone usually takes a year.

In the semiconductor industry where technological changes are rapid, speed is the key to survival. Only by being faster and stronger can we seize the initiative in market competition.

Fourth, a strict assessment system with rewards and punishments should be established.

The semiconductor industry is a technology-intensive industry with high requirements for product quality. In addition, most semiconductor companies have state-owned backgrounds, so it is particularly important to establish a mechanism with clear rewards and punishments.

For example, the annual salary of CEOs of Samsung Group companies is only 25% of the basic salary, and the other 75% is determined by the company's stock price increase, profit indicator economic added value, and the completion rate of established goals. Long-term competitiveness indicators such as R&D strength and marketing are also important criteria for judgment. The same is true for ordinary employees. 60% of the salary is basic salary, and the rest depends on performance. The principle of "grading according to ability and rewarding according to effort" is an important reason why Samsung Electronics has world-class competitiveness. In contrast to rewards, once problems arise, thorough accountability must be carried out. For example, once it is discovered that bribes have been accepted or there are major omissions in the product, no matter how high the position is, they will be dismissed without exception; once the performance is sluggish for more than three consecutive times, they will be permanently removed from the promotion list.

3. Implications for the domestic semiconductor industry

After the discussion in the second part, it is not difficult to see that in the development process of the semiconductor industry, the wolf model has more advantages than the Buddhist model. Based on this, the following suggestions are put forward for the development of the domestic semiconductor industry.

On the one hand, the government should always play a role as a pusher, continue to provide funding and supporting policy support, and guide the integration and allocation of official, industrial, and academic resources. Historically, the global semiconductor market has never been a fully competitive market (from the original text of "Ensuring American Semiconductor Leadership" presented by PCAST to the President) , and the development of semiconductors must rely on the power of the "visible hand". Specifically, the following measures can be taken:

(1) Continue to implement major national science and technology projects such as the 02 project, guide the industry and academia to cooperate in the division of labor, and carry out short-term, medium-term and long-term strategic layouts for the semiconductor industry. Externally, attract international leading companies and research institutes to set up R&D centers in China through financial subsidies and tax incentives, and consider their cooperation with domestic companies and the conditions for entering the mainland in different links such as design, manufacturing, and packaging;

(2) Provide universal individual tax benefits to employees of semiconductor companies, and reward existing senior executives and core technical personnel based on the progress of technology research and development and the completion of indicators such as company performance; provide preferential measures such as government subsidies and share exit income tax exemptions for high-end talents introduced from overseas, and provide their families with high-quality education, medical care, and pension resources;

(3) Improve the construction of basic disciplines related to semiconductors, expand the number of students enrolled in microelectronics majors, improve discipline settings and training programs, give priority to scientific research funding and faculty salaries in key colleges and universities, add practical courses that meet industry needs, encourage students to intern in the industry, and encourage leading companies in the industry to jointly train talents with scientific research institutions and universities.

On the other hand, as market players, semiconductor companies should give full play to their initiative in marketing, technology research and development, fixed asset investment, internal management, and external market development . Specifically, the following measures can be taken:

(1) Adhere to independent control of the core technology system, increase the proportion of R&D expenses to sales revenue internally, ensure R&D intensity, and actively introduce high-end talents from home and abroad in technology R&D and other related positions, and provide them with internationally competitive salaries;

(2) Equip the technical department with a dedicated product planning and marketing department, which will continue to pay attention to the market structure (product categories and forms, enterprise product lines and projects, factors affecting market size), the evolution of the market and products (product life cycle and corresponding profit margins and sales volumes), industry forces (existing competitors, new entrants, indirect competitors, suppliers, customers), and environmental factors (political, economic, social and cultural, technological, legal and regulatory, and physical environment), pay close attention to market trends, and make timely technological innovations based on market demand;

(3) Strengthen cost control awareness while ensuring reasonable quality, especially fixed asset depreciation control, improve equipment production efficiency, and seek a reasonable balance between yield rate and utilization rate;

(4) Improve investment concepts. In addition to considering the short-term input-output ratio when making investment decisions, we should also consider the industry operation cycle, market competition pattern, potential undeveloped markets, and the company's ability to bear the investment results over a relatively long timeline;

(5) Improve operational efficiency and execution speed. Motivate employees through salary incentives, improve work efficiency, improve the ERP system (a system that manages and uses the data collected by each department in a unified manner), and integrate all departments such as ordering, production, sales, inventory, logistics, accounting and personnel through uninterrupted and low-loss business processes. Reduce the time required for decision-making and implementation between the headquarters and branches, and move towards a low-cost, high-efficiency business model;

(6) Actively expand domestic and overseas markets through product exports, overseas direct investment, mergers and acquisitions, etc., enter the supply chain of important customers, increase market share, and diversify business risks;

(7) Reward market-oriented incentive mechanisms and strict accountability systems, use corporate stock prices, annual performance, R&D results, etc. as performance evaluation indicators, establish smooth promotion channels for outstanding employees, and at the same time establish a strict responsibility accountability mechanism to thoroughly hold individuals accountable for product quality accidents and other issues.

Today is the 1580th issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.