[Haitong Electronics] Exploring the opportunities facing China's semiconductor industry from the historical laws of memory development

Latest update time：2021-09-06 02:09

Reads：

Click on the blue font " Semiconductor Industry Observation " under the title to follow ICBank

recommend : Black Technology WeChat Platform kejihei (Welcome to follow)

Learning from history, exploring the opportunities China faces from the historical laws of memory development

(This report is from the electronics industry research team of Haitong Securities. If you want to communicate further, please contact Haitong Electronics Analyst Chen Ping 18502138751 )

Investment Points

From the perspective of output value, memory and microprocessors are two major components of the semiconductor industry, accounting for 22% and 19% of the output value of semiconductor products respectively . In the development of the semiconductor industry, memory and microprocessors are inevitable topics. This article analyzes the historical context of memory development in detail, analyzes the characteristics of memory industry development from a historical perspective, and obtains inspiration for the current development of China's memory industry.

History is a mirror for understanding the rise and fall of things. Memory has gone through half a century of development and has experienced many changes. The capacity has grown from K to T , a difference of 1 billion times. From a regional perspective, memory originated in the United States in the 1960s , moved to Japan in the 1980s, and now the industry leader has moved to South Korea. In terms of technology, the technology used in memory is becoming more and more cutting-edge in the entire IC industry. Today, NAND has always used the most advanced semiconductor technology and even leads the entire semiconductor trend in 3DIC packaging. Studying the history of memory has important implications for our current development of the memory industry. Looking back on the history of the memory industry, we hope to gain inspiration for industrial development from the historical trajectory and explore the development ideas of China's current memory industry. And from a historical perspective, we can correctly grasp the historical opportunities for China's current development of the memory industry.

The history of memory has given us at least three insights: the right time, the right place, and the right people. The so-called right time requires a good historical opportunity. The fundamental reason why the memory market shifted from the United States to Japan in the 1980s was the result of the rapid rise of the Japanese economy after World War II. The semiconductor industry shifted to South Korea in the 1990s , which was also the result of seizing historical opportunities - Japan was in the " lost decade " , while South Korea enjoyed the feast of the " Four Asian Tigers " ; the so-called geographical advantage requires the use of local advantages, whether it is the advantages of technology or capital. The so-called right people emphasizes working together, the government provides strong funding, enterprises are proactive, and domestic resources must be united and work together to avoid fighting alone or even vicious competition.

China faces a historic opportunity to develop memory, perhaps the last opportunity. First, the international semiconductor industry is shifting to China. Second, China's consumer market is huge, with sufficient demand to drive consumption. Third, new growth engines have emerged in the memory industry, such as the Internet of Things, big data centers, smart homes, wearable devices, etc. Fourth, with the emergence of new technologies, old technologies have reached a turning point. China is expected to find a shortcut to developing memory and narrow the gap. Finally, the strong support of national policies for the semiconductor industry is a good opportunity for the Chinese industry to develop memory.

The key for China now is to seize the historical opportunities, give full play to its own characteristics, focus on future opportunities in emerging fields, and truly innovate with people as the center, so as to take a different path with its own characteristics from large companies. With the mentality of "ten years of hard work to make a sword", we should make products seriously. Through Huawei HiSilicon and others, we see the hope of the rise of the semiconductor basic industry in China, and also the hope of the gradual growth of the memory industry in China.

For specific stocks, we recommend "Huatian Technology". The development of the memory industry will have a significant driving effect on the progress of the country's entire semiconductor industry. If the country makes a major investment in the memory industry, the upstream and downstream of the industry chain will benefit significantly. In addition to wafer manufacturing companies such as SMIC ( 00981.HK ), listed companies recommend paying attention to upstream and downstream supporting companies in the industry chain: downstream packaging companies Huatian Technology (has reached a cooperation agreement with Wuhan Xinxin), Changdian Technology, and Tongfu Microelectronics ; upstream companies such as Shanghai Xinyang will also benefit from the localization trend of semiconductor materials. Non-listed companies are recommended to pay attention to China Micro Semiconductor, Anjie Microelectronics, Longsys, GigaDevice ( IPO queue) , etc.

text:

This report draws inspiration from the historical laws of the development of the memory industry to China's development of the memory industry, and makes suggestions for the country's development of the memory industry. Here, we put the conclusions of inspiration in the front for your reference. Criticism and correction are welcome!

3.1 Revelation 1: Follow the will of heaven - the rise of an industry must have historical opportunities

Follow the will of heaven and rise straight up

For a large industry to transfer to another region, there must first be a historic opportunity. This opportunity may be the overall development direction of the industry or the economic prosperity or decline of a certain region. From the half-century history of memory, we can find that there have been two successful major regional transfers in the memory industry, namely, from the United States to Japan in the 1980s and from Japan to South Korea in the 1990s .

The rise of Japan's semiconductor industry in the 1980s took advantage of the rise of the Japanese economy in the world. Japan's economy rose rapidly after World War II, and its national industry also developed rapidly. Take automobiles as an example. In the 1960s , Japanese manufacturing has always been synonymous with "low prices and poor quality". But soon, Japanese Toyota Motor dominated the low- and mid-end small car market in the United States, even surpassing American brands such as General Motors and Ford and German brands. In the early 1990s , the land of Tokyo could buy the entire United States in terms of market value, and the Japanese economy reached its peak. Therefore, the fundamental reason why the memory market shifted from the United States to Japan in the 1980s was the result of the historical wave of the rapid rise of the Japanese economy after World War II.

The transfer of the semiconductor industry to South Korea in the 1990s was also the result of seizing historical opportunities - Japan was in the " lost decade " , while South Korea enjoyed the feast of the " Four Asian Tigers " . In 1985 , the US policy towards Japan underwent a fundamental change, from supporting pro-US Japan to maintain the balance of the situation in Northeast Asia to curbing the excessive development of Japan's economy. The Plaza Accord was signed, and Japan began its "lost decade". From the 1970s to the 1990s , coinciding with the rise of the "Four Asian Tigers", South Korea's national economic strength reached its peak on the eve of the economic crisis at the end of the last century. The memory industry has always required a lot of capital and manpower investment due to standardization and scale. Therefore, it is not surprising that South Korean semiconductors replaced Japan in the 1990s .

Lack of time, futile efforts

Without a great historical opportunity, it is difficult to build such a large industry by relying solely on the personal efforts of a few industry professionals or relying on national support. There are many examples of attempts to enter the memory industry but ultimately failed. Behind these seemingly individual company failures, in fact, are the failures of the country / region after all-out efforts.

Take Taiwan as an example. Although the downstream assembly, packaging, LED and other markets are gradually being eroded by the mainland, the semiconductor industry has always been the pillar of Taiwan's economy. Taiwan developed the memory industry a few years later than South Korea, but it lost the "right time". One step late, one step late, and finally led to a world of difference thirty years later. Samsung began to announce the development of the memory industry in 1983. Six years later, Taiwan began to develop memory in a joint venture with Texas Instruments of the United States . At this time, the gap between Samsung and the international market had narrowed to one year. Five years later, Taiwan found that it needed to focus on the development of memory with the efforts of the whole island. More than a dozen companies jointly established the world's advanced independent research and development of DRAM technology. At this time, Samsung was at the same level as international giants in terms of technology. The next year, Samsung surpassed American and Japanese companies and launched 256MDRAM to become the leader of the memory industry, which has continued to this day. Five years later ( 2000 ), Taiwan Advanced withdrew from the road of independent DRAM , and Taiwan's memory industry became a foundry. After the 2008 economic crisis, Taiwan's memory foundry industry suffered losses, while Samsung continued to be an industry oligopoly.

3.2 Revelation 2: Leverage geographical advantages - local advantages must be leveraged for industrial development

Behind a large-scale industrial regional transfer, in addition to great historical opportunities, there must also be national advantages and support from the national level. Japanese and Korean companies were the first to start from the low end and gradually grow through technology introduction and cooperation. Every time the industry is transferred, the up-and-coming companies use their own characteristics to defeat the strong.

The rise of Japan's memory industry is largely due to the nation's meticulous and continuous improvement characteristics. In terms of cost and yield, Japan's management and industrial engineering are very good. For example, the 5s concept used in management by most companies today originated in Japan. It can be said that Japan defeated American companies that are mainly based on innovation through meticulous and strict industrial engineering management. With this feature, Japan surpassed the United States in home appliances in the 1960s , surpassed the United States in the automobile industry in the late 1970s , and challenged the United States in semiconductors in the 1980s . In terms of storage, Japan's latecomer is mainly because Japanese companies have better control over cost and yield. The Japanese government also provided a lot of support to the early semiconductor industry.

One of the most important aspects of the rise of the Korean memory industry is the unique capital power of Korea. Japan fell into a long-term recession in the 1980s, and Samsung took advantage of the situation to surpass several Japanese memory companies. For example, consumer electronics have always been price sensitive. Japanese companies fell into serious export difficulties due to exchange rate issues in the 1980s. At one point, they even needed to be at least two generations ahead of Samsung to have a price advantage. On the other hand, Samsung's success relies on its strong financial capital. Supporting the Asian financial crisis in 1997, Samsung relied on strong financial support to survive the difficult period of 300% debt ratio, and then quickly rose to become the world's top semiconductor company. In 2007 , Microsoft launched Vista , which led to a large increase in DRAM production capacity in the industry. As a result, Vista sales were lower than expected, resulting in an oversupply of DRAM and a sharp drop in prices. Coupled with the 2008 financial crisis, DRAM prices fell from US$ 2.25 to a bottom of US$ 0.31 . At this time, Samsung once again played the " counter-cyclical law " : 118% of Samsung Electronics' total profit in 2007 was invested in DRAM expansion business, relying on huge capital to aggravate industry losses. As a result, it dragged down Qimonda, the German company that ranked third in the industry in 2006 , and Elpida, the Japanese company that ranked fifth. Samsung relied on its special domestic political and business relations to carry out dumping and counter-cyclical investment with huge domestic capital, which is well known in the industry .

China is already facing a historic opportunity for semiconductor transfer, but to develop its own memory business in such a fiercely competitive environment, it must also have its own local characteristics. In the context of the gradual disappearance of the demographic dividend, more scientific means must be used to find its own advantages. For example, national strategic policies and financial support, the huge domestic consumer market, grasping opportunities such as the upcoming Internet of Things, and supporting domestic companies' overseas mergers and acquisitions.

3.3 Revelation 3: Harmony among people - government and enterprises work together

For an industry like semiconductors that requires extremely high capital investment, extremely fast equipment depreciation, extremely high barriers to entry, and requires years of technological accumulation, it is not easy to break out of a market that has already been dominated by others! It is far from enough to rely solely on the government's wishful thinking or the enthusiasm of individual entrepreneurs. For a weak country to defeat the strong with the weak, it often requires the whole country's strength, and the government and enterprises to work together, that is, the government should provide policy and financial support, and domestic enterprises should work together, truly attach importance to the semiconductor industry and invest enough resources, make good use of policies and funds, and strive to develop.

Take Japan as an example. Today, a large number of Japanese companies originated from the government's military industry during World War II. After the war, they were privatized at a very low price, changing from government-run to state-owned, and giving full play to the advantages of private enterprises for marketization. Therefore, Japanese companies have official origins. In addition, Japan has given a lot of policy support to the semiconductor industry. In 1963 , NEC obtained process technology licenses from Fairchild in the United States , and the government required NEC to share technology with other domestic manufacturers. Since then, large Japanese companies such as NEC , Mitsubishi, and Kyoto Electric have begun to enter the semiconductor industry. The Japanese government has also given strong support to the introduction of patented technologies and talents. The concerted efforts between enterprises and the concerted efforts of the government and enterprises have played an important role in the rise of Japanese semiconductors.

The South Korean government and Samsung Electronics model can be described as a national unity. The government has given strong support to Samsung Electronics. The South Korean government has a tradition of supporting conglomerates, and the tax incentives given to corporate R&D and equipment investment are higher than those of other countries. When Samsung Electronics first entered the memory industry in the 1980s , the government provided policy and self-support in the form of "official-civilian integration" R&D. In the 1990s , Samsung's LCD panel industry suffered losses for 9 consecutive years. The government successively granted Samsung more than 6 billion US dollars in policy loans at an interest rate lower than the domestic average . Today, Samsung's panels are firmly in the leading position in the industry. In the Asian financial crisis in 1997 , a large number of companies went bankrupt, and Samsung's debt ratio was as high as 366% . In 1999 , it still had a debt ratio of 166% , but Samsung still had enough government funds to survive the difficulties and then became the leader in one fell swoop. In the first decade of this century, according to incomplete statistics, the South Korean government gave Samsung Electronics a total of 8.7 billion US dollars in tax breaks. It is this kind of stable, continuous, ongoing and massive financial support that has enabled Samsung to not only overcome difficulties time and time again, but also to make "counter-cyclical investments" when the industry is in decline, increasing its own losses and those of the industry until its competitors are dragged down. Until the industry recovers, it can then dominate the world while making a fortune.

4Historical Opportunity Period

As we discussed above, opportunities are very important and often fleeting, so we need to seize them in time. Historical opportunities are often gone and never return. Without the right time, all efforts will be in vain. Taiwan developed memory five years later than South Korea, and missed the right time, which resulted in a huge difference decades later.

The transfer of semiconductors to developing countries is a historical trend . China is currently at the node of the transfer. Properly grasping this historical opportunity, from back-end to front-end, from reverse design to innovation, from low-end to high-end, is one of the paths for the development of China's memory industry. In addition, China is facing some development advantages.

4.1 Semiconductor industry shifts to China

At present, China is facing the historical trend of semiconductor industry shifting to developing countries. In October 2014 , IBM paid $ 1.5 billion to sell its semiconductor business. In Europe and the United States , semiconductors are already a traditional industry with meager profits, so it is inevitable to shift to developing countries such as China. From 2008 to 2013 , the regional structure of the global semiconductor industry shows that Europe and Japan have shrunk from 35% to 22% , while the Asia-Pacific region including China has increased from 50% to 57% . China's semiconductor industry has started from packaging and gradually explored upstream foundry, design, and innovation to establish its own complete semiconductor industry chain. This is also the practice of Japan, South Korea, and Taiwan, and it has begun to show results.

Data source : Investment database, compiled by Haitong Securities Research Institute

4.2 China's consumer market is very important

The latest research from DRAMeXchange , the memory storage division of TrendForce , shows that based on 2Gb chips, China's market consumption of DRAM and NAND in 2014 reached 4.789 billion and 7.036 billion US dollars, totaling 10.2 billion US dollars, accounting for 19.2% and 20.6% of the global production capacity respectively .

From the perspective of the DRAM market, the consumption of PC-DRAM in the Chinese market has reached 15% . As for mobile memory, due to the hot sales of smartphones and tablets in the Chinese market, the mainstream products have gradually shifted from standard memory to mobile memory. In addition, China's own smartphone brand manufacturers have become an important source of demand. Except for ZTE and Huawei, which have a high proportion of exports, most Chinese brand manufacturers still focus on domestic sales. Based on the total consumption in 2014 , China accounted for about 28% of the mobile memory demand, and it is expected to exceed 40% in 2015 .

4.3 New Growth Engine Emerges in the Storage Industry

At present, smart phones, the Internet of Things, and servers are considered to be the three engines of memory development. The strong market demand brought about by this is expected to bring about a sustained memory boom and ease the fierce competition in memory. In particular, the Internet of Things is unanimously optimistic, and some people even call 2014/2015 the first year of the Internet of Things.

Some concepts related to the Internet of Things are also moving towards the mass market, such as automotive electronics, smart homes, wearable devices, etc. The gap between these fields at home and abroad is relatively small, and it is easy to achieve independence with the support of national policies. At the same time, once these fields explode, they will become a new blue ocean for growth in the memory industry.

Mobile devices emphasize the M2C (object-to-consumer) architecture of human-machine interaction, and basically adopt the concept of semi-open system design. The Internet of Things emphasizes the M2M (object-to-object) interconnection architecture, with embedded system design as the mainstream. But whether it is M2M or M2C , the core system still needs to use DRAM . In 2014 , the Internet of Things (IOT) began to grow strongly. Unlike general consumer electronics that pursue large capacity and high speed, the Internet of Things pursues stability and low power consumption and has lower capacity requirements. Therefore, low-capacity DDR2 with the most stable reliability below 1Gb is widely used . At the same time, in the field of automotive electronics, new markets including electric vehicles, smart navigation, unmanned driving, and body sensing have brought benefits to niche DRAM . DDR2 is still the mainstream in the embedded system market such as automotive electronics or industrial control. Taking automotive electronics applications as an example, customers' specifications for DRAM require that in addition to high stability and reliability, high and low temperature resistance, the life cycle of each product is as long as 7 years. In addition, smart grid, smart LED lighting, three-network integration, mobile payment, set-top box IPSTB , etc., further expand the application scope of niche DRAM .

Unlike smartphones, the memory currently required by the Internet of Things mostly requires low power consumption and high stability, but not high capacity, so it can be achieved with relatively backward technology. This is one of the breakthroughs in the development of China's storage industry. For example, China's GigaDevice has been deeply involved in the NOR Flash segment and has begun to scale.

4.4 When new technologies reach a turning point, cutting in can narrow the gap

Looking at the development history of storage companies, new technologies are constantly emerging. Directly adopting new technologies and new standards to narrow the gap with foreign countries may be a direction that some companies can try.

Take 3D NAND technology as an example. Currently, NAND has reached 20nm in mass production , and the process has reached 1Z . Domestic enterprises are almost blank, lagging behind foreign countries by decades. However, when transferring to 3D technology, the more mature 30nm process plus 3D technology will be adopted. In this way, the gap will be narrowed to less than 10 years. If Chinese enterprises directly introduce 3D technology, they will find opportunities in the industrial chain.

Another new direction is new material technology, such as phase change memory, memory resistor, ferroelectric memory, magnetic memory, etc. Take RRAM as an example . RRAM products are not as abundant as PCRAM , but they are developing rapidly and are favored by Samsung, SanDisk, etc. The strongest RRAM technology is Crossbar , which was founded in 2010. Its RRAM chip can be mass-produced. The first sample is expected to be released in early 2015 and put on the market at the end of 2015. In the early stage, it is mainly aimed at the embedded market and licensed to ASIC , FPGA , and SoC developers. It uses a 3 -layer chip stack. When it is commercialized, it is expected to use a 16- layer chip stack with a capacity of 128GB , which can be mass-produced. At the same time, Crossbar claims that the technology can be reduced to 5nm . Another example of using new technology to achieve overtaking is Invensense , which was founded in 2003. After 10 years of development , the company's MEMS products have successfully squeezed into Apple's supply chain. The iPhone6 uses a six-axis accelerometer and gyroscope from Invensense .

4.5 National Strategic Support

In some basic IC manufacturing industries, monopoly is the trend. Industrial development requires a lot of capital, a lot of technical accumulation, and years of huge losses. After making a profit, it still has to go through big market fluctuations. However, once profitable, it can become the peak of the industry and form a barrier to latecomers.

The government plays a huge role in supporting the rise of an industry, and the rise of Japan and South Korea has proved this. At present, China's 100 billion-scale semiconductor support fund, as well as the local funds and supporting policies linked to it, can show that the government has indeed regarded the semiconductor industry as a pillar industry for future development and has begun to provide real large-scale support. If the government's policies can be continuously implemented and truly implemented, this will play an important role in the rise of China's semiconductor industry and is also a historic opportunity for the development of China's semiconductor industry.

China has many policies to support semiconductors, such as Document No. 18 issued in 2000 and Document No. 51 issued in 2001. However, these policies have not been well implemented for various reasons. The major " core high-tech " projects during the 11th Five-Year Plan period have achieved results such as Loongson CPU , Kirin operating system, and Huarui DSP , but there is still a long way to go before semiconductor independence. China's semiconductor imports have exceeded oil, and the government has decided to invest hundreds of billions of yuan to support the semiconductor industry. Various local governments have also introduced support measures. How to prevent funds from flying around and how to reasonably plan to use funds in the right places may be areas that need careful study.

A huge industry needs good supporting services to be self-reliant under the fierce international competition, and the most important one is financial services. For the storage industry, which has high investment, high technical barriers, and often takes ten years to see results, strong financial support is indispensable. During the Asian financial crisis in 1998 , South Korea's financial operation model allowed Samsung to obtain a steady stream of financial support. The financial crises in 1998 and 2008 caused a large number of companies in Japan, Europe, and Taiwan to die, but Samsung's market share further increased after surviving the crisis with the support of the chaebol. At present, China's financial policies are moving towards standardization, and more and more domestic private equity firms are being born.

5. A crossroads of history: Should China develop the storage industry?

5.1 Importance and urgency of development

Storage involves national information security

China's semiconductor industry started late, has a weak foundation, and relies heavily on imports for a large number of equipment, which poses a serious threat to information security. In particular, the information industry faces more urgent security issues. The Snowden incident once again exposed the domestic information security issues, and the door to my country's information security is almost completely open to the outside world. Huawei, as the country's largest and most advanced telecommunications equipment manufacturer and mobile terminal equipment manufacturer, has had its high-level email servers monitored by the US government for many years without anyone noticing. This incident shows how urgent and important it is for China to develop national independent industries in key areas, which is related to national defense security, national industrial development, and the lives of ordinary people! This is an important reason for the country to develop national industries and replace imports. The semiconductor industry is the foundation of communications, and memory is the cornerstone of the semiconductor industry, accounting for half of the semiconductor industry. Almost wherever there are semiconductor equipment, there are memory. Developing domestic memory products is the cornerstone of national information security.

Memory accounts for a large proportion of China's imports

In 2012 , China imported more chips than oil, becoming the largest imported commodity, and the import growth rate is above 10% . Memory and microprocessors are the two major components of the semiconductor industry, accounting for 22% and 19% of the output value of semiconductor products respectively. From this perspective, the importance of developing China's semiconductors is even greater than that of energy. At the same time, the scale of China's semiconductor market has exceeded half of the global semiconductor market. Developing China's semiconductor industry is a major event for the country's economic lifeline. The storage industry accounts for 22% of the output value of the entire semiconductor industry. Therefore, it is a general trend for China to develop its own storage industry, especially in the current economic structural transformation. Of course, independent development does not mean challenging Samsung's leading position from the beginning, but starting from the back end step by step, gradually completing its own industrial chain, and finally realizing the autonomy of the entire storage industry.

5.2 Difficulty in development

Asset-heavy industry characteristics

In some basic IC manufacturing industries, monopoly is the trend. Industrial development requires a lot of money, equipment depreciates quickly, requires continuous investment, long-term stable policies, a lot of technical accumulation, and years of huge losses. After making a profit, it will also experience large market fluctuations. However, once profitable, it can become the peak of the industry and form a barrier for latecomers. This is especially true for the memory industry. From the previous analysis, it can be seen that the memory industry is a highly standardized industry with little product differentiation. It mainly relies on scale to win, which requires the most advanced production capacity (the latest technology), so it will cost a lot of money.

Memory volatility

The memory industry has experienced great ups and downs, and the market is extremely volatile. It is not uncommon for prices to drop from a few dollars to a few cents. Moreover, memory manufacturing companies are often very large in scale, because small companies cannot survive on production capacity, so the loss scenario is also very bleak. In addition, companies like Samsung will artificially create or exacerbate industry volatility, so volatility always exists cyclically, and good times are never long. This situation requires a strong ability to resist risks, which is not something that ordinary companies can bear. Even Elpida, which has the best memory resources in Japan, or Qimonda, which represents Europe, have declared bankruptcy during cyclical fluctuations, and the government can do nothing about it.

Highly monopolized industry

The memory industry is highly monopolized, with oligopolies accounting for more than 90% of the global market share . It is extremely difficult to get a piece of the pie from the world's top companies. If there is no good opportunity, and we continue to develop along the past path, we will never have a chance to catch up.

No sufficient conditions

For an industry like semiconductors that requires extremely high capital investment, rapid equipment depreciation, extremely high barriers to entry, and years of technological accumulation, there are many necessary conditions but none of them are sufficient conditions in order to develop one's own national industry amid fierce competition.

Huge funds are very important. South Korea survived the Asian financial crisis in the late 1990s largely due to huge and stable financial support. Scientific and efficient management may be more important than funds - especially for some industries with thin profit margins and large scale - which is also what domestic companies lack. Talent and a good employment mechanism are the foundation, especially in the semiconductor industry. For example, Saudi Arabia, despite having enough money and sand, cannot produce a single IC . A good entry point is also very important. If you enter too early, the technology management, basic industry market, etc. will not keep up, and it will be futile; if you enter too late, the market will solidify, and then the situation will change and there will be no hope of sharing. If Taiwan entered the memory industry ten years earlier, or if China continued to move forward along the path of the first batch of domestic DRAM in 1975 , the situation might be very different. Sufficient size is also important. Although Singapore has more technology and funds, it has not become a heavyweight country in the memory industry. Taiwan's defeat in the memory competition with South Korea is also related to its small size. Policy support, strong execution, etc. are all indispensable.

5.3 China's Way Out

Opportunities exist

First, China has opportunities to develop semiconductors (memory). Although the current situation shows that the memory industry is highly monopolized and China's foundation in this area is very weak, it seems unrealistic to develop its own memory industry. However, from the laws of history, it can be found that industrial regional transfer is not impossible. If we go back to the situation when Samsung or Japan first developed memory, the powerful competitors they faced were no different from China's current situation.

China's opportunity is not to challenge the international oligopoly directly, but to find new features with a forward-looking vision and make itself bigger and stronger. For example, in the PC era, Intel had an absolute monopoly in processors, but Qualcomm, starting from the mobile Internet, posed a huge challenge to it. China is also actively working in some special memory fields, and this idea is desirable.

Get the timing right

The opportunity is gone, and Taiwan's failure is a profound lesson. As mentioned above, China is currently facing a good opportunity to develop the semiconductor industry, which is probably the last train to achieve independent innovation in the national industry. If China loses this opportunity, this trend will be seized by other emerging countries (such as the BRICS countries), which will be detrimental to China's long-term development.

Focusing on opportunities in emerging fields is an important way for China to realize the localization of the memory industry. The emergence of new materials and new processes, such as RRAM and 3D NAND , provides a shortcut for China to develop the memory industry, which is expected to greatly narrow the gap with the current giants.

Strategic Vision of Ten Years to Grow Trees

Regional transfer requires at least ten years of effort and will not happen overnight. Japan began to introduce semiconductor technology from the United States in 1963 and began to develop the storage industry in the 1970s . After 10 years of hard work, it finally surpassed Intel in 1980 and became the leader of the storage industry. Samsung Electronics began to develop the memory industry in 1983. After more than ten years of hard work, it finally surpassed Japan in 1995 and has since become the leader of the storage industry. Huawei HiSilicon, founded in 1991 , has finally achieved initial results after 24 years of development . The Kirin processor released in 2014 has far exceeded MediaTek's performance and can even compete with Qualcomm. If China wants to develop the storage industry, it must also have a strategic vision of ten years to grow trees.

The inherent attributes of manufacturing determine that it is different from communications and the Internet, and it takes years of accumulation before it can flourish. However, all roads lead to Rome. Whether it is Huawei's gradual growth through independent innovation, Lenovo's acquisition of overseas assets through capital operations and then digestion and absorption, or Tsinghua Unigroup's direct industrial integration through government power, it all gives people hope for the growth of national enterprises.

Therefore, the key for China now is to seize the historical opportunities, give full play to its own characteristics, focus on future opportunities in emerging fields, and truly innovate with people as the center, so as to take a different path with its own characteristics from large companies. With the mentality of sharpening a sword for ten years, we should make products seriously. Through Huawei HiSilicon and others, we see the hope of the rise of the semiconductor basic industry in China, and also the hope of the gradual growth of the memory industry in China.

6 Investment advice: Pay attention to the upstream and downstream of the memory industry chain

The current investment opportunities in the memory sector of China's A- shares mainly come from two areas: packaging and testing opportunities that may be brought about by IDM outsourcing, and related supporting opportunities brought about by independent innovation.

IDM outsourcing, domestic packaging and testing have opportunities

At present, most memory companies adopt the IDM model, which is to complete the design, manufacturing and packaging of memory chips by themselves. At present, with the dilution of profit margins, non-core businesses are gradually being divested by large companies. Micron has a large number of outsourcing orders, and Taiwan's Inotera, Chip Micro, and Powertech are its partners in wafer manufacturing and packaging and testing. The semiconductor manufacturing industry is shifting to developing countries, especially the packaging and testing industry, which is now gradually shifting from Taiwan to the mainland. For the memory industry, although it is currently dominated by IDM , non-core industry chains may be outsourced, and domestic packaging and testing manufacturers will face major opportunities.

Independent innovation, opportunities exist in the domestic supporting industry chain

The development of the memory industry has a significant driving effect on the progress of the country's entire semiconductor industry. If the country makes a major investment in the memory industry, the upstream and downstream of the industry chain will benefit significantly. Domestic memory chip design companies include GigaDevice ( IPO queue, NorFlash chip design), Shandong Huaxin, etc., and manufacturing companies include SMIC ( 0981.HK , memory chip wafer manufacturing), Wuhan Xinxin (unlisted, currently involved in Nor Flash wafer manufacturing), etc. In addition, if the memory industry develops vigorously in China in the future, there may be investment opportunities in domestic related equipment, materials, packaging and testing, modules, controller chips and other supporting facilities. We will keep a close eye on it.

For listed companies, we recommend paying attention to upstream and downstream supporting enterprises in the industry chain: Huatian Technology (which has reached a cooperation agreement with Wuhan Xinxin), Changdian Technology, and Tongfu Microelectronics ; upstream companies such as Shanghai Xinyang will also benefit from the localization trend of semiconductor materials.

The following section discusses the laws of the memory industry:

1Two regional shifts / three historical stages: each leading the way for decades

Reviewing the development history of the memory industry can provide some enlightenment for the current analysis of the memory industry. Looking at the development history of the memory industry over the past half century, it can be divided into three stages and two major regional transfers. After decades of leading the way, who will dominate the future?

Looking at the development of the memory industry over the past half century, it roughly corresponds to three stages: the United States in the starting stage in the 1970s, Japan, South Korea and Taiwan in the growth stage in the 1980s and 1990s , and the oligopoly in the mature stage of this century. It has experienced two major regional transfers. The United States invented memory and dominated the market for about 10 years. In the 1980s , Japan's semiconductor industry followed the rise of the automobile industry and surpassed the United States in one fell swoop. The memory leader shifted from the United States to Japan, and the latter dominated the memory market for about ten years. In the 1990s , Samsung Electronics of South Korea rose, and the memory hegemony shifted from Japan to South Korea. The latter has dominated the memory market for exactly 20 years (from the first launch of 256MDRAM in 1995 to the present). In addition, the memory market has also experienced the decline of Japan, the failure of Taiwan, and the short-lived success of Europe. Now it is highly monopolized, and the oligopolies have divided more than 90 % of the market. The NAND industry is monopolized by Samsung, Micron, Toshiba / SanDisk, and Hynix . The DRAM industry is currently divided into three parts by Samsung, Hynix of South Korea and Micron of the United States.

1.1 The start-up period in the 1970s: the United States, mainly focusing on research and development

In 1966 , IBM invented the first DRAM chip. In 1970 , Intel launched the first 1K commercial DRAM chip , Intel 1103. The early DRAM had a small capacity of only 64 to 256KB and was directly fixed on the motherboard. At the same time, there were about a dozen DRAM companies in the United States. For example, Mostek MK4096 4 - kilobit DRAM was once very popular and once reached 3/4 of the market share. Intel was the first to use three-inch wafers to replace two-inch wafers to generate DRAM , which kept it ahead from the beginning. During this period, technology research and development (R&D) was the key to development. The rise and fall of every company was accompanied by the update or elimination of its technology. The good scientific research environment and strong scientific research foundation of the United States were the key to its domination of the entire DRAM market. At that time, there were about 40 to 50 DRAM companies in the world .

1.2 Growth period from 1980s to 1990s: Japan, South Korea and Taiwan, government support + talent and technology introduction + local innovation

After DRAM products entered the mature stage, Asian countries such as Japan and South Korea successively achieved catch-up through government promotion, technology introduction and cost reduction. In 1963 , NEC obtained process technology license from Fairchild in the United States , and the government required NEC to share technology with other domestic manufacturers. Since then, large Japanese companies such as NEC , Mitsubishi, and Kyoto Electric have begun to enter the semiconductor industry. Japan has spent a lot of money to purchase a large number of integrated circuit patents, and at the same time pays great attention to quality improvement and its own innovation. For example, Japan was the first to use CMOS technology to develop 1M DRAM . In the early 1980s , the reliability of Japanese DRAM products was higher than that of the United States. Competing on cost is not the specialty of American companies. In the 1980s , Japan replaced the United States as the main supplier of DRAM . For a long time, the top three semiconductor companies in the world were NEC , Toshiba and Hitachi, and they once had 90% of the market share in the DRAM industry . At this time, Intel in the United States withdrew from the market due to serious losses in DRAM and turned to microprocessors with higher technical requirements.

South Korea's semiconductor industry started later than Japan, and also started from the end of the industrial chain. Its model is similar to that of Japan, namely, government promotion, technology and talent introduction, and cost reduction. In terms of national policies, South Korea has given strong support to DRAM . For example, with the support of South Korean state-owned banks, South Korean companies have built a large number of factories with low capital costs and the latest technology, squeezing out competitors through scale reduction of costs. This method was particularly devastating to small businesses during the DRAM storm ( the cyclical rise and fall of the DRAM market). At the same time, Samsung can rely on strong financial support to carry out "counter-cyclical investment", that is, when peers reduce investment and reduce losses during the industry downturn, Samsung increases investment and increases losses, thereby narrowing the gap with its competitors when it is weak, or further widening the gap with its competitors when it is strong. In terms of technology introduction, Samsung's early DRAM technology came from Micron's authorization. At the same time, South Korea has attracted a large number of overseas technicians to return home, and also attracted a large number of scientific and technological workers from other countries (such as Japan). Samsung's talent strategy can be described as a real sword and a gun. Samsung's conditions for poaching people from Japanese companies were once to double their salaries, solve housing problems and solve the work, study and life of engineers and their families in South Korea, which even led to lawsuits. Samsung Electronics hired Koreans who had worked in American semiconductor companies at high salaries, which were 4 to 5 times higher than the president's salary. It was this strategy of truly valuing talent and technology research and development that laid the foundation for the magnificent building of Samsung Electronics. In the 1990s , Japan's " lost decade " coincided with the appreciation of the yen and the economic downturn. South Korea's DRAM took the opportunity to rise and replace Japan. In February 1983 , Samsung founder Lee Byung-chul said that he would build Samsung into a global memory chip company. In the same year, South Korea launched 64K DRAM , which was 4 years behind the international leading level . When 1M DRAM was launched in 1986 , the gap had narrowed to 1 year. In 1992 , it launched 64M DRAM in the same year as American and Japanese companies . In 1995 , it had surpassed American and Japanese companies to launch 256M DRAM , and has since occupied the largest market share of DRAM . In the 1990s , South Korea's DRAM market share once reached 65% , and this proportion has been maintained to this day.

Taiwan's DRAM was slightly later than South Korea's, and its model was similar to that of Japan and South Korea, but it failed to follow the path of success of Japan and South Korea. In the early days, Taiwan also obtained authorization through cooperation, such as in 1989 , PC manufacturer Acer and Texas Instruments jointly established Taiwan's first DRAM factory, Texas Instruments . Many overseas Chinese returned to Taiwan to invest in the DRAM industry. In cooperation with the Electronics Institute of the Industrial Technology Research Institute, Taiwan has independent DRAM technology. In 1994 , TSMC took the lead in establishing World Advanced with more than a dozen companies to develop Taiwan's independent DRAM technology. Taiwanese invested about US$ 20 billion in six years. However, Taiwan's DRAM not only failed to surpass Japan and South Korea, but also failed in its road to technological independence.

Generally speaking, the global DRAM industry flourished during the growth period of DRAM . Before 1999 , there were 14 DRAM manufacturers, including Samsung and Hyundai from South Korea, Micron and IBM from the United States, NEC , Hitachi, Mitsubishi, and Toshiba from Japan, Siemens from Europe, and Powerchip, ProMOS, Nanya, Winbond, and World Advanced from Taiwan.

1.3 Mature stage of this century: market mergers and reorganizations, oligopoly

After entering the new century, DRAM entered a mature stage. Like many other industries, DRAM entered an oligopoly stage through mergers and reorganizations.

First, Taiwan withdrew from the mainstream stage of memory. In 2000 , the withdrawal of the world's advanced DRAM business from independent research and development marked the end of Taiwan's road to independence. Then the foundry model suffered a heavy blow in the 2008 financial crisis. In the past, Taiwan's DRAM industry was dominated by small and medium-sized enterprises, such as Powerchip, Prochip, Mosel, and Winbond. The only large group was Nanya Technology and Inotera, which were supported by Formosa Plastics Group. Inotera was originally established as a joint venture with German memory giant Qimonda , and was later taken over by Micron. In 1998 , DRAM was in oversupply and the industry suffered huge losses. Taiwan's DRAM gradually turned to foundry, and after 2000 , the world's advanced also withdrew from the independent research and development of DRAM business. Taiwan completely adopted the technology of other companies, such as Prochip's technology from Hynix, Powerchip from Elpida, and Nanya Technology and Inotera's technology from Micron. Foundry is very developed in Taiwan, and it also has good profits in good years. But in the DRAM storm in 2009, Taiwan's DRAM industry disappeared . As a result, NUST and Winbond transformed into niche DRAM manufacturers, Powerchip became a wafer foundry, and ProMOS transformed into an IC design company.

There are many reasons for the failure of Taiwan's DRAM industry. For example, the timing of entry was too late, the market technology was already relatively mature, and it entered the stage of competing on cost. Especially in the first decade of this century, it had entered the oligopoly stage, and the opponents were all world giants, so there was not much room for latecomers to survive; there was no advantage of its own, and there was no good opportunity. Japan's DRAM came from behind by competing on cost with the United States (the same is true for Japanese cars), and South Korea came from behind when it encountered the Japanese economic depression; without independent technology, it was controlled by others in critical moments and could not integrate resources. In the DRAM storm in 2009, Taiwan's DRAM belonged to different camps and each fought on its own, which led to the failure of the government's integration plan; affected by the Asian financial crisis in 1997 , Taiwan's economy was in recession, and it failed to stick to the path of independent innovation at a critical moment, and there was no chance later; there was a lack of sufficient financial backing in the crisis. In 1997 , Samsung's debt ratio was 366% during the Asian financial crisis, and it still had a debt ratio of 166% in 1999 , but it still had continuous support from domestic financial capital, and finally survived the financial winter and avoided bankruptcy. Since then, it has been advancing all the way and has become the overlord of DRAM and NAND Flash .

In addition to Taiwan, Japan and Europe also withdrew from the DRAM business in the first decade of this century . In 1999 , the DRAM businesses of Hitachi, NEC , and Mitsubishi merged to form Elpida in an effort to avoid losses. Between 2001 and 2002 , IBM and Toshiba withdrew from the market, and Hynix was established as an independent semiconductor business unit of Hyundai in South Korea. By 2003 and 2004 , there were only five major camps left in the global DRAM industry , including Samsung and Hynix; ProMOS and Elpida; Powerchip and Qimonda; Nanya, Inotera, and Winbond; and finally Micron; After the 2008 financial crisis, DRAM prices fell from US $ 6 per piece in Q4 2006 to US $ 0.6 per piece in Q1 2009 , and the global DRAM industry suffered huge losses. This directly led to the bankruptcy of Qimonda in Europe in 2009 , the acquisition of Elpida in Japan by Micron in 2012 , and the transformation of Taiwan's DRAM companies. Since then, the global DRAM market has been monopolized by three oligopolies, occupying more than 90% of the market share. Among them, Samsung Electronics, SK Hynix and Micron accounted for 40% , 29% and 24% of the market share respectively .

2. Volatility of the Memory Industry: Determined by Genes

Looking at the development history of the memory industry, in addition to international regional transfers, there is another characteristic, which is cyclical fluctuations and monopoly in the fluctuations, which is reflected in both the DRAM and Flash industries. We believe that this volatility is rooted in the industry's genes and is innate and unavoidable.

The volatility of the memory industry is determined by the special properties of memory products. The standardization and scale of products determine that manufacturers can always more easily expand production by reducing size, and then increase production capacity by expanding factories. However, the demand for memory market is volatile, affected by macroeconomic conditions, flagship products, seasons and other factors. This greedy expansion gene leads to drastic fluctuations in market supply and demand, which leads to periodic storms. During the storm, small companies with weak resistance are eliminated, resulting in more and more monopoly.

2.1 Products: Standardization, scale, low user stickiness, and large market fluctuations

The biggest feature of the memory industry is its high degree of standardization. For example, a DDR3 memory stick, no matter who the product is, what the capacity and internal packaging method are, must follow the standard interface, so almost all devices with DDR3 interfaces can be used universally. For both users and manufacturers, replacing a memory is much more convenient than replacing a CPU . As a result, it is difficult for memory manufacturers to form effective stickiness with users.

Under the circumstance of standardization and homogenization, the more reasonable way out for memory is to win by quantity. At present, the output value of memory accounts for 22% of the entire semiconductor industry . Almost all devices need memory, from servers to PCs to mobile devices to the current Internet of Things. Therefore, storage is a basic industry. In such a universal and large-volume situation, upgrading capacity seems to be more attractive than upgrading interfaces, especially in the field of consumer electronics, whether for memory manufacturers or users. Compared with the upgrade of CPUs , which requires a lot of redesign, the capacity upgrade of memory is much easier. As a result, compared with accessories such as CPUs , motherboards, and graphics cards, the upgrade of storage devices is much smoother, and the product life cycle is much longer, while the capacity upgrade is relatively easier and easier to hit the user's " pain points " . For example, the iPhone 3Gs produced in 2009 had only 256MB of memory and 32G of storage space . Now eMMC has been raised to 128G by iPhone 6 , and the memory has been raised to 4G by Android .

Another characteristic of storage products is that demand fluctuates greatly. This fluctuation is reflected in both price and quantity. First, electronic products are extremely susceptible to macroeconomic influences. Due to the large demand, once the overall economic situation is sluggish, the demand for memory as the " basic raw material " of the electronics industry fluctuates greatly. The Asian financial crisis in 1997 , the Internet bubble in 2002, and the financial crisis in 2008 all directly affected the demand and price of the DRAM industry. The Nand industry also encountered twists and turns in 2003 , 2007 , and 2012 , and survived hard. The price fluctuation of the DRAM industry far exceeds the fluctuation of the macro economy, and the price can even fall by 90% . On the other hand, the demand for memory is subject to seasonal fluctuations because the demand for PCs , mobile devices, etc. is seasonally volatile, such as the peak season at the end of the year. In this way, DRAM manufacturers can only stock more in the off-season, and the risk of inventory is very high. Once the market fluctuates, the inventory may suffer serious losses. In addition, single products also have a great stimulus to the storage industry. For example, the iPhone's launch conference in the fall of 2014, the surge in shipments brought benefits to the memory industry, and the increase in mobile phone storage to 128G further stimulated the storage industry.

2.2 Periodic Memory Storm

As a commodity in the electronics industry, the cyclical volatility of the memory market is well known. Its volatility is mainly determined by the characteristics of the memory itself: memory products are in high demand, are key components in the semiconductor industry, and have a high degree of standardization, fast replacement, weak user and product stickiness, and are easy to replace. Market capacity can be pushed up quickly, such as smartphones from the previous 512M to 2G or even higher. This makes large companies always have the urge to expand their scale, and this approach seems to always be right for large companies. When the market is good, expansion can squeeze the market with small profits and quick turnover. When the market is bad, large companies have strong tolerance and can drag down their opponents.

For example, due to the impact of the Asian financial crisis, DRAM supply exceeded demand in 1998 , and the industry suffered heavy losses. Around 2000 , due to the impact of the bursting of the Internet bubble, DRAM product prices fell by more than 30% in the third quarter of 2002. When the financial crisis broke out in 2008, DRAM prices fell from US$6 per chip in Q4 2006 to US $ 0.6 per chip in Q1 2009 , and did not start to pick up until the end of 2011 .

2.3 Fierce competition and oligopoly

Under the influence of periodic storms, the monopoly of the DRAM industry has become increasingly serious. The competition in the DRAM industry is fierce. In the 1970s, there were as many as 40-50 companies . At that time, the industry was in its infancy, and companies were mainly concentrated in the United States, which had strong R&D capabilities. In the 1980s , products entered the growth stage, and the quality and price of products determined the size of the market. In the 1980s, Japan replaced the United States by reducing costs and once occupied 90% of the global DRAM market . After the 1990s , South Korea took advantage of the economic decline in Japan and rose to become the main supplier of the DRAM market, with a market share of 65% , which has been maintained to this day. At the end of the last century, there were still 14 DRAM manufacturers in the world. In the first decade of the new century, the DRAM industry was surging. After a decade of drastic mergers and reorganizations, there were only 5 left in 2004 , and now there are only three oligopolies left, occupying 90% of the market share.

The NAND industry is also experiencing a similar monopoly trend. Currently, the international NAND market is dominated by Samsung, Toshiba / SanDisk, Micron, SK Hynix, etc., ranking the top four, accounting for a total of 99% of the market share. Among them, SanDisk is Toshiba's NAND Flash business partner. Its market share in China is basically the same as that in the international market. The specific process will not be repeated.

2.4 The storm continues, and the expansion genes of major manufacturers continue to emerge

In the past two years, after DRAM recovered from the pain of the 2008 financial crisis, a three-oligopoly situation has been formed. Major manufacturers have been restrained in expanding production capacity and shipments, resulting in continued profitability in the DRAM industry in the past two years. This positive factor will continue in 2015 , making the short-term market more optimistic.

However, by studying the development history of the memory industry and major related companies, it can also be found that the long-term volatility of the memory industry has not changed. The reason is that major memory manufacturers all have the gene to expand production capacity.

Take Samsung as an example. There are countless precedents in history of wantonly expanding production capacity, increasing shipments, and frantically lowering prices. It is also notorious in the industry. In 2006 , three senior officials of Samsung Electronics were sentenced to seven months in prison by the US Department of Justice for allegedly manipulating memory prices . This move is particularly prominent when the industry is in recession. The consequences range from hitting competitors to directly bankrupting competitors. For example, in early 2010 , DRAM prices began to stabilize and rise after a round of sharp declines, and Samsung quickly expanded its production capacity, causing a wailing in the industry. This common practice runs through the past development history of Samsung Electronics and is even one of Samsung's " success secrets . "

We can look at the latest expansion moves of the major giants:

Samsung's latest expansion: Samsung has 4 factories for DRAM products, all with 12- inch production lines, all built in Hwaseong Park, South Korea, with a monthly output of 395,000 pieces. Samsung has already adopted 20nm mass production in terms of process technology, and the gap in process technology with Hynix and Micron has widened to two years. Samsung's new expanded plant Line 17 will have a monthly production capacity of 50,000 pieces, all of which will be converted to DRAM production , and it is expected to have begun gradual production. On October 6 , 2014 , Samsung announced a new plant expansion plan in Pyeongtaek, Gyeonggi Province, South Korea, which is expected to start construction in early 2015 and may be officially put into production as early as the first half of 2017 , and it is very likely to focus on DRAM or NandFlash products.

Driven by Samsung, other manufacturers are also actively expanding their production capacity, and the new capacity will begin to take effect in 2016 .

SK Hynix's latest expansion: Hynix currently has one 8 -inch wafer production line and two 12- inch production lines in South Korea, one 8- inch production line in Oregon, USA , one 12- inch production line in Wuxi, China , and production lines in Taiwan. It has a long-term partnership with Taiwan ProMOS and has a research and development center in Europe. SK Hynix's new expansion plant M14 is expected to start production in Q4 2015 and mass production in 2016 .

Micron's latest expansion: Micron's semiconductor manufacturing plants are located in the United States, China, Japan, Malaysia, Puerto Rico, Singapore and Taiwan. In 2014 , DRAM products mainly used 30nm production lines. In 2015 , Micron will put two 20nm factories into production. Micron 's 20nm factory in Hiroshima, Japan is progressing rapidly, and at the same time , it is conducting research and development of tens of nanometers and several nanometers nodes in Boise , Idaho, USA .

In view of the above analysis, we believe that the biggest characteristic of the memory industry is the high volatility of the business cycle. In the short term, the three-way competition will continue for a while, but there is still a hidden worry of monopoly in the future. At least from Samsung's perspective, it is a good choice to form a situation similar to the CPU dominated by Intel . In the future, we will analyze the industry status, characteristics and development trends in detail from the aspects of DRAM , NandFlash , Nor Flash , etc.