In the next 30 years, there will be endless work in China's semiconductor industry!

-Meng Jie, Chairman of Qualcomm Inc China

This article is a speech from King's Landing at an offline event, which has been sorted out and posted online.

The problems solved in this speech

1. Opportunity: is it possible for China to rise in the semiconductor industry?

   If so, which link and which companies will have the most opportunities in the industrial chain?

2. Risk: an industry with high returns and high risks

   What are the most careful pitfalls in terms of investment rhythm and periodicity?

1. Opportunity:Is it possible for China to rise in the semiconductor industry?

I believe that the friends who come here today, more or less have some understanding of the semiconductor industry, so those publicly reported investment boom, national policy support, and so on, we will not say much here.

Let's discuss a little practical question here: is it possible for China's semiconductor industry to rise?

I believe many friends are worried. After all, after half a century of development, China's semiconductor industry is still weak, with a global market share of no more than 3% in almost every link of the entire industrial chain.

As the representative of the high-end manufacturing industry in the world, is it possible to throw money at it?

It doesn't seem that simple, does it?

If you look at the Hanxin incident and the Red Core browser incident in the past, subsidies and counterfeiting go hand in hand like a twin.

Let's take a look at the industrial robots in the past few years. After the bubble, the share of domestic robots still hovers around 20%.New pine robotShare prices have fallen 75% since 2015.

How can it not be worrying?

Now that we're here, let's put aside our emotional prejudices and explore in depth.

First of all, we all know that the biggest obstacle to the development of China's semiconductor industry is the United States, which has the most developed and complete semiconductor industry ecology in the world.

In front of the American semiconductor industry, we are simply hitting a stone with an egg.

How developed is the semiconductor industry in the United States?

Maybe you only have a vague impression, but you don't have a clear concept.

To put it this way, as of October, there were 75 listed companies in the four major industries (machinery, automotive, aerospace and military industry, and IT hardware) with a market capitalization of more than $13 billion.

Among them, IT hardware industry ranks first with 30 companies, accounting for 40%.

Of the 30 companies, 10 are equipment manufacturers in downstream computers, communications and other fields, and 20 are upstream semiconductor industry chain companies.

They include--

Digital chip:Apple(series A chips),Intel Corp 、 NVIDIA Corp 、 IBM(supercomputing chip),Qualcomm Inc, AMD, Sailingsi(data center chip),Meguiar and Western Digital(acquisition of SanDisk)

Analog chip:Texas Instruments Inc, Broadcom Ltd, Yadno, Microcore Technology, Meixin Semiconductor, Sijiaxun Semiconductor

Equipment:Applied Materials Inc(integrated equipment),Lam Research Corp semiconductor(wafer equipment),Corey semiconductor(detection equipment)

Software:New thinking of science and technology(IC design software),Cades.(IC design software).

There are Apple Inc, Intel Corp, NVIDIA Corp, IBM, Qualcomm Inc, Texas Instruments Inc and Broadcom Ltd, seven top companies with a level of 100 billion US dollars.

Among the major industries in the United States, the only ones with so many hundreds of billions of dollars of companies are the software Internet, semiconductors and pharmaceuticals.

There is no doubt about the strength of the US semiconductor industry, so what about other countries except the United States?

In fact, there are also large-scale semiconductor industries in Europe, Japan, South Korea, Taiwan and other places. Let's take another look:

Europe--

Equipment area:ASML Asmaier(Philips of the Netherlands split), accounts for 80% of the lithography machine market.

Analog chip:Enzhipu(spin-off of Philips, Netherlands),Infineon(split of Siemens Germany),Italian semiconductors(split in Thomson, France), accounts for 30% of the analog chip market.

Japan--

Memory and analog chips:Toshiba(exited),Fujitsu(exited),NEC+ Hitachi + Mitsubishi = Renesas, Sony Group Corp(CMOS sensor).

Equipment area:Nikon + CanonThe lithography machine is in decline.Tokyo Electronics(etching equipment)

Field of materials:Xinyue chemistry(Silicon wafer material),JSR(photoresist),JX(target),Hitachi Kasei, Asahi Kasei, Sumitomo ChemistryAnd other dozens of companies, accounting for 50% of the global semiconductor materials market.

Korea--

Samsung, SK HynixAccounts for 80% of the memory chip market.

Taiwan--

Taiwan Semiconductor Manufacturing Co Ltd, Sun and MoonIt accounts for 50% of the wafer manufacturing and downstream closed testing market.

MediaTekIs the second largest independent mobile phone chip company, and Wenmao is the world's largest gallium arsenide compound semiconductor company.

It seems that the semiconductor industry chain in Europe, Japan, South Korea, Taiwan and other regions is not weak, and it is entirely possible to compete with the United States.

Then, under the circumstances that the ecology of the semiconductor industry in the United States and Imperial is so perfect, how do the semiconductor companies in Europe, Japan, South Korea and Taiwan grow up tenaciously?

Why should they?

To find the answer to this question, we must start with the history of the development of the semiconductor industry.

How did semiconductors come into being?

The first generation of semiconductor components are called tubes.

Look at this picture:

These components that look like light bulbs are electronic tubes.

It is said that more than 100 years ago, Edison invented the electric light bulb. in the process of research, in order to find the best filament material, he inadvertently discovered the "Edison effect" of semiconductors, which is one of the theoretical origins of the semiconductor industry as a whole.

Later, the scientist Fleming invented the electronic tube, and a huge industrial ecology was born.

Electronic tubes are components used to amplify electrical signals, on this basis, downstream telephones, radios, radars and other electronic products appear one after another.

It can be said that the semiconductor industry is the foundation of the modern electronic industry. Before World War II, the whole electronic industry was built on electronic tubes.

Second-generation semiconductor components: transistors.

The glory of the tube lasted for half a century, until after World War II, it was finally replaced by more advanced transistor technology.

These are transistors that are plugged into the circuit board.

Transistors are packaged with solid materials instead of glass, with more stable quality, smaller size and more rich functions.

This is one of the greatest inventions of the 20th century, invented by Shockley of Bell Laboratories in the United States.

It is said that Shockley, who won the Nobel Prize, was a godfather in the semiconductor industry and later went into business and started a company near Stanford University.

The company did not grow big, but trained a group of apprentices, two of whom founded Intel Corp and one of them founded AMD, which is the source of Silicon Valley.

But that was 20 years later.

In the era of transistors, there are several key branches to be carried out in detail.

The first discrete device.

Discrete devices, mainly power semiconductors, account for a small share of today's semiconductor industry chain, with a scale of about 5 per cent and US $20 billion.

But in the early days, discrete devices were all semiconductors.

Including diodes, transistors, capacitors, resistors. Wait.

At that time, the entire electronics industry was built on this basis.

In Europe,Philips, Siemens, ThomsonThe industrial giants dabbled in one after another, and then they spun off the related business.ASML, NXP, InfineonSuch as today's well-known semiconductor companies.

Naturally, there are also some in the United States.Texas Instruments Inc 、 Broadcom LtdAnd other giants, but before World War II, the strength of European industry was able to wrestle with the United States, and both sides started at the same time, naturally carving up the market of this era.

This tells us a truth:When a technology or an industry is born, it is very important to do it early, which is called first-mover advantage.

But today, the pattern of the industry has changed dramatically:

First of all, diodes, transistors and other markets are gradually eroded by China's industrial chain. The representative enterprise isYang Jie Science and Technology, Hua Microelectronics, Suzhou solid Technetium.

The core competitiveness of Chinese enterprises is low cost and cost performance. In generalIndustries with low technical barriers, without exception, are difficult to escape such a fate.

European and American manufacturers, however, stick to the high-end market: more complex components such as IGBT and MOSFET.

Do not think that power semiconductors mature early, there is no future, this is actually a misunderstanding, because these emerging technologies such as IGBT are about to shine again in the 5G era.

IGBT is the core device of energy conversion and transmission, commonly known as the "CPU" of power electronic devices.

In the 5G era and artificial intelligence era, which terminal equipment has the most imaginative market prospect? It's new energy + driverless cars.

In the traditional fuel truck era, discrete devices will also be used, mainly in automotive air conditioning and central control systems and other places, but the overall proportion is not large, not as much as communication equipment.

However, in the era of new energy vehicles, the power of cars has been converted from chemical energy to electricity, and the use of semiconductors has been greatly increased.

It is estimated that the semiconductor use of a new energy vehicle will be about three times higher than that of a fuel vehicle, and the use of discrete devices such as the IGBT will increase by about five times.

This is the biggest growth opportunity for the semiconductor industry brought by the whole new energy vehicle market.

After 2020, with the commercialization of 5G and the maturity of self-driving technology, the amount of semiconductors used in each smart car is expected to increase by about 4-5 times.

In this process, the growth of discrete devices is quite remarkable.

In this field, the current four leaders are from Europe.NXP, Infineon, Italian and French Semiconductor, and JapaneseRenesas Electronics.

In fact, it is easy to understand why it is them, because Europe and Japan are the largest automobile industry bases in the world, and it is precisely because of the huge demand downstream that these giants are born.

In the face of the wave of new energy and self-driving cars, how does the industry respond?

Qualcomm Inc of the United States wants to buy NXP, but this acquisition of nearly 50 billion US dollars has been solemnly rejected by our government, and the acquisition is aborted.

China has a 50% market for new energy vehicles in the world, and is at the forefront of self-driving. Of course, this piece of meat is reserved for our own children.

The second sensor.

Sensing devices include photosensitive, sound sensitive, pressure sensitive, thermal sensitive, gas sensitive, magnetic sensitive, humidity sensitive, fluid sensor and so on.

The traditional market in this area is small, with only about $10 billion worldwide, about half the size of power semiconductors.

However, there has been a rise in this field in recent years.

Because of the combination of different sensors and micromechanical power, it can be upgraded to MEMS sensor, that is, MEMS, which is amazing!

Do you know, in the era of self-driving cars, what are the prerequisites for a car to drive safely?

In addition to the chip + operating system, 5G communications + sensors, the former gives the car the brain, the latter gives the car the nervous system and sensory system.

not a single one can be omitted.

As we said earlier, the biggest beneficiaries of the upgrading of fuel vehicles to the era of new energy vehicles are power semiconductors, while the biggest beneficiaries of upgrading from new energy vehicles to the era of self-driving are chips and sensors.

In the field of automotive sensors, most of the market is also monopolized by several European and Japanese manufacturers mentioned above, and there is no way. After decades of accumulation, the technology is too deep.

Then our company will not have the opportunity?

Of course not, this market is very large, with many branch applications, and some small European and American manufacturers with accumulated technology are being acquired by listed companies in China one after another in recent years.

Like 2016.Niwei Science and Technology700 million yuan bought the SwedishSilexThis is the fifth largest MEMS wafer foundry in the world and has risen to the third place in the industry this year.

To put it another way, the two largest MEMS sensor companies in China are in the field of electroacoustics.AAC Technologies Holdings Inc.和Gore shares。

The third plate is optoelectronic devices.

Optoelectronic devices can be divided into two categories:Light to electricity, electricity to light。

Light to electricity: photovoltaic cells, fiber optic cables, CCD and CMOS (cameras, cameras).

Electric light conversion: LED lighting, flat panel display, laser equipment, infrared remote sensing, etc.

Which sector of our country has had the most tenfold shares in the last ten years? There is no doubt that it is an optoelectronic device. FromLongji shares到Hengtong Optoelectronics, FiberHome Communications, Zhongtian Science and Technology从San'an Optoelectronics, Liad, Huacan Optoelectronics到Large family laserTen-fold shares simply emerge one after another!

Why are optoelectronic plates so hot?

A very important point comes from the semiconductor material revolution in the last 20 years.

The earliest semiconductor material, called germanium, but due to its own defects: small reserves, high production costs, active chemical properties, unstable quality. As a result, it has not been widely used.

Later, silicon was used in semiconductor materials and soon became the mainstream of the market.

We know that silicon is one of the most common elements on earth, second only to oxygen. The sand, rock and crystal on the ground are full of silicon, which makes its production cost low enough to become the basis for large-scale use.

On the other hand, silicon has stable chemical properties and good insulation, which makes it a basic material for high-density memory and high-performance computing chips.

Until today, more than 95% of the materials in the semiconductor device and chip market are still made of silicon, which is almost equated with the chip.

So, does silicon have any disadvantages?

Of course, the biggest problem with silicon is its poor performance in photoelectric conversion, high frequency and high power performance.

So in the last 20 years, more and more compound materials began to appear, such as gallium arsenide, gallium nitride and silicon carbide.

In the field of poor performance of silicon, compound materials can show their skills and shine in the fields of optoelectronic devices, power devices and radio frequency devices, and technological revolutions emerge one after another.

To take a simple example, gallium arsenide diodes emit red light, gallium phosphide diodes glow green, silicon carbide diodes glow yellow, and gallium nitride diodes emit blue light.

It can be said that the whole LED industry is based on compound materials.

In the past decade, not only LED lighting, LED flat panel display, but also a series of technological advances such as optical fiber broadband and laser equipment have led to a ten-fold prosperity.

In the next decade, the coming explosion opportunities of IGBT power devices in the new energy vehicle market, optical fiber and radio frequency devices in the 5G communications market will be visible to the naked eye.

More far away, quantum communication and quantum computer technology are based on compound materials.

At present, crystalline silicon is the mainstream of photovoltaic cells, and the main advantage is low cost, but in terms of photoelectric conversion efficiency, it is not as good as thin film cells based on compound materials.

It can be said that the past century is the era of silicon, while the next hundred years may be the era of compound semiconductors.

In this field, the most noteworthy company is San'an Optoelectronics.

In May 2014, San'an Optoelectronics invested 3 billion yuan to set up San'an Integration, which is mainly engaged in compound semiconductor integrated circuit business, and has completed 6-inch gallium arsenide and gallium nitride production lines. At present, the company has five largest compound semiconductor R & D centers in the world, with the accumulation of more than 1000 chip patents.

The Quanzhou Core Valley project, with an investment of 33.3 billion yuan, was launched in December 2017. all the projects are expected to be put into production within five years and full production within seven years, mainly engaged in the design and manufacture of compound chips, lasers, radio frequencies, filters and power devices.

It is expected that when the project is fully produced, the total revenue will be three times that of the current one.

Having said so much before, in fact, it has not yet entered the core.

As we said, the core material of semiconductors is silicon. In the transistor era, more and more silicon components have been replaced by compound materials, but there is one area that may never replace the core position of silicon.

This is the integrated circuit.

Of course, as we said earlier, the future quantum computer and quantum communication technology are based on compound materials, but because the production cost of the latter is high, it may not be able to replace silicon for decades. Therefore, it can only be used in high-performance computers, national defense communications and other places where there is no shortage of money.

Integrated circuit, abbreviated as IC, is made by compressing a large number of transistors, integrating them into a semiconductor chip, and then encapsulating them.

The first generation of chips is called analog chips.

There are two categories--

General purpose chips (operational amplifiers, data converters, filters)

Special chips (signal processing chip, RF chip, power management chip).

The leaders of this industry are Texas Instruments Inc and Broadcom Ltd of the United States. Thanks to the huge demand and profound accumulation of the computer and communications equipment industry in the United States since World War II, there are not many seed players with bright spots in our country.

The second generation chip is called digital chip.

It is divided into two categories: logic chip and memory chip.

First of all, logic chips, this industry since World War II, giants continue to emerge, can be said to be the mainstream battlefield of the semiconductor revolution.

From IBM power series chips in the mainframe computer era, Intel Corp and AMD in the personal computer era, Apple Inc A series and Qualcomm Inc in the mobile phone era.

In the digital age, the iterative evolution of computing power is the main theme, and the technical threshold is extremely high, so the United States has always occupied the dominant position of martial arts by the United States.

For half a century, Europe, Japan and South Korea have only envied and hated in this field, but they have never acted improperly.

The senior management of our country once hated and made great efforts to subsidize policies and engage in independent control, but it was a pity that fraudulent projects such as "Hanxin" were controlled, and then they came to an end and stopped thinking.

So, is there a chance for China?

In fact, there are, logic chips to be big, the key is two points: get up early, good technology.

For example, in the era of mobile phone chips, Android chips use the architecture technology designed by the British company ARM. This technology has low power consumption and is naturally suitable for mobile devices with prominent battery bottlenecks.

So when did ARM start working on this technology?

In 1985, it was only a few years later than Intel Corp.

Unfortunately, in the PC computer era, ARM chips can only be used in some low-end electronic devices, and the light was covered by Intel Corp, but it was precisely because he got up early and the technology was mature that when the world entered the mobile Internet era, Intel Corp could not catch up.

This is how the change of dynasty took place.

In the traditional chip field, the technology has been mature and perfect, our country wants to catch up, in fact, it is very difficult, there is almost no opportunity.

However, facing the future, the era of artificial intelligence is coming, the traditional technical architecture has been outdated, is destined to be difficult to adapt, there are actually a lot of new opportunities.

For example, deep learning requires a lot of data training, NVIDIA Corp's GPU chip has been widely used, and the stock price has soared 14-fold in the past three years.

Another company, Cyrus, whose FPGA chip is widely used in data centers, has a head start in cloud computing, and its share price has doubled in three years.

Another Israeli company, Mobileye, whose chips have a 90 per cent market share in smart driving assistance chips, was acquired by Intel Corp for $15 billion in 2017, also doubling in three years before the acquisition.

In the era of artificial intelligence, the biggest feature is that the data is massive, but the application market is scattered, and customized special chips are needed in almost every application field.

This gives rise to a lot of opportunities.

In addition to some of the American start-ups we mentioned earlier, China has also emerged one after another in recent years.

such asCambrian periodThe chip architecture has been integrated into Huawei Hayes chip, becoming the world's first artificial intelligence dedicated communication chip.

such asBit continentIts blockchain chip technology monopolizes 80% of the market in the industry, alone in the world.

For example, the unicorn startup horizon's self-driving chip, Yunzhisheng's voice recognition chip.

In particular, there are several unicorn companies in the field of face recognition algorithms in China.Shang Tang, Xiangshi, Yuncong, YituWith current valuations of more than 15 billion yuan, it is one of the fastest growing artificial intelligence startups in the world.

Due to the attention of our government, the flooding of capital, and the huge downstream application market, the leaders of China's AI industry have obtained excessive valuation in the chip research and development of artificial intelligence, with a strong speed of development, and are extremely prosperous in finance, security, mobile phones, automobiles and other different industries.

On the contrary, because the US science and technology industry is monopolized by Apple Inc, Alphabet Inc-CL C, Amazon.Com Inc, Microsoft Corp and Facebook Inc, the financial support for start-up companies is actually decreasing, and the momentum of innovation is weakening.

At present, not many start-ups in the AI industry in the United States get high valuations, and they are mainly concentrated in the field of self-driving cars, as well as related chips, software and parts.

The status of the two countries in the industry is already neck and neck.

Back to the A-share market, since leading AI companies are generally not listed yet, it is worth paying attention to some vertically segmented companies, such as smart speaker chips.Quanzhi science and technologyAs a security video chipFu Han Wei, make fingerprint identification chipHuiding Technology.

Another area of digital chips is memory chips.

This is not only our highlight today, but also the main battlefield of this vigorous wave of semiconductor investment in China.

Why is the main battlefield here?

Two reasons:First, there is a strong demand, and second, our government sees an opportunity.

As mentioned earlier, 20 years ago, because of the pain of the chip, the Chinese government tried to make a force on the digital logic chip, but failed in the end, why did it make a comeback today?

On the demand side, we'll talk about it later. Where are the opportunities first?

Opportunity can be divided into two aspects: one is ability, the other is threshold.

Today's China, compared with 20 years ago, the ability is no longer under Wu Ameng, whether it is the ability to mobilize funds, or the accumulation of talent, has been many times as much as in the past.

This is gradually accumulated through the success of large-scale infrastructure projects, aerospace science and technology projects and high-speed rail projects over the past 20 years, and it is also the strength that we dare to advance into the large aircraft and semiconductor industry.

On the other hand, our government realizes that compared with logic chips, the technical threshold of memory chips is relatively low, and it is more of a capital-intensive industry.

This is why in the field of storage, the strength of the United States is relatively weak, and the industrial chain is more occupied by East Asia.

In the capital-intensive industries, the Chinese government looks around the world and almost kills gods and Buddhas, and has never met a rival.

This is precisely the confidence that our government dares to invest a large number of troops and concentrate on tackling key problems in this field.

It can be said that we already have the strength and confidence.

Now, let's explore further--

How does the history of the rise of the memory chip industry in Japan and South Korea evolve, and why is there a huge demand and opportunity here?

In fact, the development of memory chip has been going on for half a century, and it was first invented by Intel Corp.

At that time, the chip was called RAM, dynamic random memory chip, the data will disappear as soon as the power is off, the most important one is called DRAM, which plays an auxiliary role in improving the computing speed on the computer, commonly known as memory.

After Intel Corp invented memory technology at that time, the Japanese government soon realized its great value, so it organized five major domestic companies to focus on tackling key problems.Fujitsu, Hitachi, Toshiba, Mitsubishi, NEC.

The turning point here is that Intel Corp is a start-up company. although he invented the technology, it is not mature enough after all. Five Japanese companies come in together, and their financial strength and R & D investment are much stronger than him.

So in less than two years, the technology and manufacturing capacity of Japanese companies greatly exceeded that of Intel Corp.

As a result, Intel Corp was forced to withdraw from the price war of memory chips and turned to the market of CPU logic chips, casting another resplendence.

The Japanese Legion, from the 1970s to the late 1990s, basically monopolized the market, the ecological chain was mature, and the upstream equipment and materials industry was extremely prosperous.

Now, let's sum up the main points of the rise of Japan's semiconductor industry chain: get up early and spend a lot of money.

Since you can do it with money, and since there are more than five players in an industry, the technical threshold for this industry should not be too high, so in the 1990s, latecomers from South Korea and Taiwan began to enter.

They don't enter in the same way.

Let's start with Korea.

South Koreans' competitive weapons are simple: throwing money, expanding production capacity, driving down costs and fighting a price war.

The capital structure of Korean chaebol is different from that of Japanese consortia.

The controlling stake of the Japanese consortium is in the hands of the banks. To ensure profitability and the rate of return on investment, banks will choose to give up and withdraw from the industry when the price war is fierce and the pressure of losses is huge.

South Korean chaebol, whose controlling shares were generally still in the hands of the founders at that time, had a strong desire to expand, had a strong will, and had a deep relationship with the government due to historical reasons, so they were able to obtain a steady stream of low-cost financial support.

From 1990 to the end of the war in 2012, more than 20 years later, Japan's former semiconductor Wuhu has all withdrawn from the semiconductor business, and three of them (Hitachi, Mitsubishi and NEC) spun off their memory business and finally sold it to Micron Technology Inc of the United States.

This is the pattern of today's memory market: South Korea's Samsung and SK Hynix have taken the top two positions and monopolized more than 95% of the market with San Meiguang.

We can see that the rise of South Korea started later than Japan, and those who come from behind can catch up with only one thing:Spend a lot of money, use losses and time to kill competitors.

Besides, Taiwan.

Taiwan's financial strength is obviously not as strong as that of South Korea, so the path to success mainly depends on the division of labor and specialization.

The production process of a chip includes three links: upstream design, equipment and materials, mid-stream wafer manufacturing, downstream packaging and testing.

Different links have different requirements for technology, capital and human resources.

For example, the upstream design, equipment, materials and other links have higher technical requirements.

In the middle reaches of wafer manufacturing, the capital requirements are higher in the early stage, and in the later stage, with the increase of storage density, the requirements for technology are getting higher and higher.

Downstream packaging testing is mainly labor-intensive.

In the early days, when the scale was small, chip companies did everything and did every step themselves.

Later, with the expansion of the market scale, the competition becomes fierce, and the newcomers tend to cut in from a certain subdivision and rely on specialization to obtain stronger competitiveness.

This is the way the Taiwanese think. There are MediaTek in the design, Taiwan Semiconductor Manufacturing Co Ltd and United Microelectronics Corp in the wafer manufacturing, and sunlight and moonlight in the closed test.

Each company focuses on one point and continues to improve its competitiveness in order to compete with the giants of Japan and South Korea for the market, thus growing step by step.

Today, Taiwan Semiconductor Manufacturing Co Ltd has a market capitalization of more than 200 billion US dollars, almost monopolizing the middle and high-end wafer manufacturing market, especially the top 7nm technology, which neither Samsung nor Intel Corp can do.

It can be said that Taiwan is the creator of the contract manufacturing mode in the semiconductor industry, which has surpassed Japan mainly through the division of labor and specialization.

In fact, the rise of ASML lithography in Europe is also the logic of division of labor and specialization.

The early lithography machine, the industry leader is Japan's Nikon and Canon, we know that the two companies are aggregators of optical technology, in cameras, photocopiers and other optical fields have a strong product line.

But it is precisely because of the numerous product lines and the inevitable dispersion of energy that the technical investment in lithography machine is not as good as that of ASML.

The latter, originally the semiconductor equipment department of Philips, was established in the 1960s and has considerable technical accumulation. after its separation in 1984, it focused on polishing the technology of lithography and gradually surpassed its Japanese competitors.

To sum up:

In the evolution of the memory industry for half a century, the market pattern has been turned upside down.

Through early entry and large capital investment, Japan established a complete memory industry ecology in the early stage.

South Korea, on the other hand, rose in the capital-intensive midstream and downstream markets by means of capital subsidies and price wars, eating up the share of Japanese companies.

Taiwan, on the other hand, has made a single point breakthrough in all aspects of the chip industry chain through division of labor and specialization, especially in wafer manufacturing and closed testing.

The rise of European ASML in the field of lithography is also achieved through division of labor and specialization.

Entering early and rising through high capital investment, there is no shortage of cases in China, such as LED industry, photovoltaic, optical fiber broadband and new energy vehicles.

Today's compound semiconductors and artificial intelligence chips are also following this path.

As for the memory chip industry, it is not too early for China to enter, and there are only two ways to learn: South Korea's capital subsidy strategy and Taiwan's division of labor and specialization strategy.

For these two strategies, it is clear to those who have a little common sense that our government and enterprises are exactly experts in this.

In 2014, the "China National Integrated Circuit Industry Investment Fund" was established, commonly known as a large fund, raising more than 130 billion yuan, officially opening the charge of China's semiconductor strategy.

Why did the National Semiconductor Strategy be launched this year?

The amount of China's oil imports over the years:

The amount of semiconductor imports in China over the years:

It is obvious that China's semiconductor imports soared rapidly for three consecutive years from 2011 to 2013, and surpassed crude oil, the largest import commodity before.

For three consecutive years from 2015 to 2017, the gap between China's semiconductors and oil imports has even widened to the order of hundreds of billions of US dollars.

With the change of data, semiconductors have replaced oil and become the core interest affecting the strategic security of our country from 2013.

The abnormal surge in the amount of semiconductor imports from 2011 to 2013 is mainly due to the popularity of smartphones around the world and the rise of China's smartphone industry chain.

Compared with PC computers, smart phones are not only the expansion of the industry brought about by a mobile phone, but also the explosive expansion of the demand for internal memory chips.

How do you say that?

It turns out that early computers mainly used a memory chip, that is, memory DRAM.

As mentioned earlier, the data will disappear after a power outage, mainly to assist in improving the speed of CPU operations.

At that time, the storage of data mainly depended on floppy disks, hard disks, optical discs and other mechanical storage means, and semiconductors were powerless.

Entering the era of smartphones, the shortcomings of mechanical storage are exposed. After all, there is almost no compressed space in the huge volume. How can you make a slap-sized mobile phone look like bulky mechanical storage?

As a result, flash NAND FLASH is introduced.

Flash memory is a kind of semiconductor storage technology that will not lose data without power outage. in the early days, due to immature technology and limited capacity, flash memory was mainly used in U drives and MP3 players.

But there is one advantage that flash memory cannot be ignored: its size can be very small, and its density and capacity can continue to increase with the progress of technology.

Since the introduction of mobile phones as storage carriers, the market scale of flash memory has begun to grow explosively, and after technological advances, it has further penetrated into notebook computers, tablets and other fields.

A huge industrial chain was born from then on!

Unfortunately, flash memory was first invented by Toshiba in Japan, but today's Toshiba, flash memory business has been sold, dropped out of the industry competition, just like Intel Corp in the memory era.

In 2017, China imported US $260 billion in chips, of which total imports of memory chips totaled US $88.617 billion, accounting for 1/3, an increase of 38.8% over the same period last year.

Of this total, the import volume of chips made in South Korea reached US $46.348 billion, up 51.3% from the same period last year and accounting for 52.3% of the total imports. Among them, the dual leaders of storage chips, South Korea's Samsung and SK Hynix have become the biggest beneficiaries.

In 2018, China's imports of memory chips reached US $123 billion, an increase of 1,188.9% over the same period last year.

With the strong demand growth and national security considerations, memory chips have become the main battlefield for the growth of China's semiconductor industry, in which flash NAND FLASH is the key.

Let's look at the following table:

▲ Note: the bar chart of the fab is marked with the start time of construction and the first production / mass production.

This is the main investment project and planning of semiconductor industry in our country.

As you can see, the largest projects are the following:Ziguang + Yangtze River Storage, Hefei Changxin, Fujian Jinhua, Semiconductor Manufacturing International Corporation.

Among them--

Semiconductor Manufacturing International Corporation works as a pure wafer foundry, and because of the big gap with Taiwan Semiconductor Manufacturing Co Ltd in technology (SMIC 14 nm, Taiwan Semiconductor Manufacturing Co Ltd 7 nm), he can only drink soup with him in the long run.

Fujian Jinhua cooperates with United Microelectronics Corp, the second largest faber in Taiwan, to enter consumer electronic memory products with low technical barriers, and the market prospect is limited.

The two most noteworthy projects are the Yangtze River Storage and Hefei Changxin.

Let's start with Hefei Changxin.

What Changxin does is the mobile phone memory product, the growth is not as good as the flash memory business of Changjiang Storage.

However, as far as the market pattern is concerned, memory is mainly composed ofSamsung, SK Hynix, MeguiarThe three monopolies, the giants have a strong desire to control prices, so the risk of price decline is low, as long as the scale of production capacity, the technology gap is small, it is possible to achieve long-term profits.

Changxin's DRAM memory project, with an investment of more than 7.2 billion US dollars (49.5 billion yuan), is divided into three phases, integrating the tripartite forces of Hefei government, big fund and Zhaoyi innovation.

The goal is to launch 8Gb LPDDR4 in the third quarter of 2019 and to have a production capacity of 20, 000 tablets a month by the end of 2019. The construction of the second plant will begin in 2020 and 17 nanotechnology research and development will be completed in 2021.

If the progress goes well, the monthly production capacity of 125000 pieces of 19nm memory in 2020 is expected to account for 8% of the global memory production capacity, with an annual output value of US $5 billion.

Besides, the Yangtze River storage.

If Changxin is only a local team, Changjiang Storage is the real national team, integrating the strength of the big fund and Ziguang Group.

In terms of scale, Changjiang Storage is expected to invest US $24 billion (160 billion yuan), three times larger than the Changxin project, and will also be built in three phases, including three factories in Wuhan, Nanjing and Chengdu.

In terms of technology, the Yangtze River has the strongest storage strength, with a research and development team of 1000 people, which will further expand to more than 2000 people in the future, with an annual R & D investment of US $1 billion.

The progress is also the fastest, and a new 3D NAND architecture, Xtacking, has been released to increase the speed of Istroke O to 3Gbps (currently the most advanced speed value in the industry is 1.4Gbps).

According to the plan, Changjiang Storage will mass produce 64 layers of 128G products in the second half of 2019, with a monthly production capacity of 20,000 pieces.

By 2023, Changjiang Storage aims to achieve full production of 300000 pieces of 64-tier 3D flash memory per month, with an annual output value of more than US $10 billion, which is expected to account for more than 20% of the global flash memory output value, and technically release 128layer 256G of the most advanced products in the industry.

Integrating Changjiang Storage and Hefei Changxin's investment plan, 2020 is a critical point.

Before, it is mainly factory construction, equipment procurement, R & D testing; after that, it will enter the stage of large-scale mass production, yield climbing and technology transcendence.

If we say that the investment in the downstream fab is a hard business, it is realized by financial subsidy + enduring time, which is the embodiment of the strong will of the country.

Then, the upstream demand driven by downstream demand expansion-equipment and materials-is the greatest opportunity for our investors.

Any company that can firmly implement the Taiwan-style division of labor and specialization strategy and achieve technological breakthroughs will have the opportunity to stand out and become a "tenfold share" in this wave of investment expansion.

There are many subdivisions of semiconductor equipment and materials, such as:

Equipment, including lithography, etching, film deposition, process control, testing, cleaning and other subdivided areas

Materials, such as silicon wafer, mask, target, photoresist, wet electronic chemicals, silicon powder, etc.

For investors, the main concerns are the following:

First, whether there is a large market space in this field to ensure the sustainable growth potential of the target company.

Second, whether there is the possibility of technological innovation in this field. If there is no technological innovation, it is very difficult for local companies to rely on quality (which is usually the weakness of local companies) to get orders.

Third, in terms of investment rhythm, before 2020, the main focus is the equipment demand brought about by downstream production capacity, and after 2020, the main focus is the material demand brought about by downstream mass production and production expansion.

Fourth, the most important observation point is which company entered the Changjiang Storage, Hefei Changxin's purchase list, and the share of orders.

Let's take the winning bid for equipment orders of Changjiang Storage as of April 2019 as an example:

Changjiang Storage has purchased a total of 212 domestic equipment, accounting for 12% of the total equipment purchases, and is also the largest single buyer of domestic semiconductor equipment. In contrast, other domestic chip factories still purchase less than 10% of domestic equipment.

This order has had a huge impact on the revenue of domestic chip equipment companies.

Medium and micro semiconductors:

From 2017 to April 2019, the procurement of equipment from Changjiang Storage was 5, 11 and 26 respectively. The revenue contribution of these 26 devices to medium and micro semiconductors is about 760 million yuan, compared with 240 million yuan in 2018.

North Huachuang:

In 2018, there were only 5 sets of equipment purchased from Changjiang Storage, contributing about 100 million yuan in revenue. Up to April 2019, there were 27 sets of equipment purchased, contributing about 200 million yuan in revenue.

All right, let's sum up:

In the last decade, the smartphone revolution driven the great development of the semiconductor industry, especially the high demand for memory chips, and brought a wave of rising prices, national security concerns and a great leap forward in China's production capacity.

In the next decade, the demand for semiconductors will continue to be driven by technological revolutions such as 5G communications, new energy vehicles, self-driving cars, artificial intelligence and cloud computing, and the Internet of things.

Driven by the growing demand and fuelled by policy subsidies, there are many investment opportunities in optoelectronic devices, power devices, chip design, equipment, materials and other upstream links in China's semiconductor industry.

If we say that the early success of China's semiconductor industry mainly occurred in closed testing, diodes, transistors and other labor-intensive links with low gross profit margin.

Then in the future, the main thing is to focus on the innovative R & D link with high gross profit margin, and whoever can make a breakthrough in technology will be able to run out.

In the course of the outbreak of the smartphone industry chain in the past decade, the competitors of the United States, Japan, South Korea and Taiwan were too strong for us to match, but a number of decimal companies, such as Kang Dexin, Xinwei Communications, Dazhou Laser and so on, still take advantage of all the available space to grow.

In the future semiconductor industry, the same truth, we believe that there will be more outstanding companies stand out.

two。 Risk:Industries with high returns and high risks

High growth and high risk go hand in hand, and for the semiconductor industry, the most important risks are two:

1. The periodicity of the industry

2. Disturbance of subsidy policy

Let's start with the first point.The periodicity of the industry.

Semiconductor has always been an industry with a double spiral of growth and cycle.

Growth comes from the expansion of demand, while the cycle mainly comes from the pulse expansion of supply.

How to understand this sentence?

Semiconductor manufacturing links, the scale of investment is huge, a project from the establishment of a project, engineering construction, trial production, yield climbing, full production, the cycle can reach 2-4 years.

Usually, when the downstream demand is strong and the price is high, it will attract a large number of new competitors to set up projects. once the projects are put into production, it will inevitably lead to excess capacity and lead to a price collapse.

This will lead to the withdrawal of some weak competitors, coupled with the growth of downstream demand, prices enter a new round of upward cycle.

This forms the cycle law of the semiconductor industry.

In fact, after the 2017 boom, memory chip prices fell in the second half of 2018.

However, this round of downward cycle has little to do with the wave of investment in China's semiconductor industry.

Because in the first half of 2019, several large local companies such as Changjiang Storage and Hefei Changxin have not yet started mass production.

The main reason is that after several years of technological progress, the mainstream process yield of Samsung, SK Hynix and Micron is constantly improving, and the previously expanded factories are also going into production, with supply growing by about 20% a year, while the smartphone market is saturated and demand is stagnant.

As soon as there is an increase and a decrease, there is an imbalance between supply and demand, and a new cycle comes.

However, this is only the situation of the downstream industrial chain, and for the upstream design, equipment and other links, the trouble does not exist at present.

After all, for the projects already under construction, there is no turning back, and the equipment to be purchased is still to be purchased.

When local production capacity is mass-produced after 2020, with the arrival of a new round of technological revolution such as 5G and self-driving, demand will break out and the cycle will rise again.

And then the second point.The disturbance of policy subsidies.

At present, a large part of the driving force of semiconductor investment boom in China comes from policy subsidies.

And the policy, there is going in and out, once the slope, or exit, will have a huge negative impact on the entire industry.

Investors should be able to feel the lethality of the policy of photovoltaic and new energy vehicles in China this year.

It is difficult to say when the policy will exit, but it is usually closely related to the maturity of the industry and its position in China's economic strategy.

For example, the retreat and withdrawal of photovoltaic and new energy vehicle policies are closely related to the fact that the two industries have established a dominant position in the world, which makes our government believe that even if the policy withdraws, the industry can still gain a comparative advantage in global competition.

In the semiconductor industry, it is clear that there is still a long way to go to achieve a similar status, so there is no need to worry about the risks for the time being.

Generally speaking, the future opportunities of the semiconductor industry far outweigh the risks, which is one of the main directions of high-end manufacturing upgrading in China, and it will also be one of the key research and tracking areas of our King's Landing. Welcome your attention.