The Four Challenges, Four Opportunities, and Four Paths of the Chip Industry

The ZTE incident revealed the current situation of the domestic chip industry structure: domestic chips are mainly used in the low-end field, and the self-sufficiency rate of the high-end general chip market is almost zero. As an industrial investor, I hope to objectively discuss with you how to develop more effectively in the future under the premise of respecting the law?

In addition to market demand, another big constraint in the field of general chips is that developers do not use them. In the final analysis, making a good chip is only the first step, and the software and hardware ecosystem supporting the chip is not perfect, even if the chip is made, Made in China will still not become the priority choice.

Regarding the future development strategy, Lu Yongchang of Galaxy Internet believes that in addition to government financial support and driving demand, chip companies are also key to identifying the market and choosing the right path.

Why can some domestic chip companies tear out 5% of the market share in a market monopolized by giants, some chip companies can even dominate the global market to obtain high gross profit margins and high net profits, while some companies are still barely maintaining state funds and investor funds after more than ten years of development. In addition to the problems of technology and talents left over from history, there are also important reasons for the grasp of ecology and market destiny, the card position of competition, the rhythm of product research and development, and the skillful design of commercial operation.

Guest introduction: Lu Yongchang, partner of Galaxy Internet Internet of Things Division. Master of investment management from Peking University, he has served as the general manager of Analysys E-commerce Consulting Center and the investment director of Sunshine Insurance, and has successfully invested in a number of TMT projects, with more than 10 years of experience in SaaS, Internet of Things, and Internet consulting and investment in traditional industries.

The current situation of the domestic chip industry: the self-sufficiency rate of high-end general-purpose chips is almost zero

The Internet of Things industry chain is divided into eight major links, the first four parts include sensors, computing and control systems, communication networks, and front-end platforms, and the last four parts include PAAS platforms, management platforms, general capabilities, and industry application products/solutions.

Six of the eight links must use chips, only pure software PaaS platforms and management platforms are mainly engaged in software services, and chips are used in other fields.

On the whole, the current composition structure of domestic and imported chips is roughly as follows:

The proportion of chip localization is very low, less than 20%, and most of them need to be purchased abroad, and there are about 40 GM in the world.

Domestic chips focus on a few fields, such as some communication chips, display processing chips, power management chips, discrete devices, MCUs, positioning and navigation, etc., most of which belong to the low-end, with low chip unit price and low profit margin, while mid-to-high-end performance chips and key devices are basically monopolized by foreign manufacturers. From the perspective of the whole process of the chip industry, domestic enterprises are not weaker than foreign manufacturers in the field of packaging and testing, some digital chip design fields do not lag behind foreign manufacturers, analog chip design is still far behind foreign countries, and chip manufacturing equipment and processes are still lagging behind foreign IDM or foundry enterprises 1-2 generation technology.

The domestic relatively advanced has the opportunity to become a platform-level 2C-end product in the future, and the general chips used are basically purchased from foreign manufacturers (the Kirin SoC used by Huawei mobile phones is all HiSilicon's IP of arm, designed by itself, and manufactured by TSMC). Common chip applications in the 2C and 2B fields mainly include:

2C-end application

The first category, smart homes. A variety of home smart items improve convenience and save manpower, even if an operation saves a few seconds, it can also form irreversible user habits in the future.

The second category is wearable devices. It is a smart product that can be carried on the human body.

The third category is smart medical equipment. For example, you can use electronic devices to measure blood sugar at home, or medical equipment to monitor various vital signs.

The fourth category is all kinds of vertical life application smart items, such as intelligent transportation, sports intelligent equipment, etc.

2B-side applications

The 2B field includes smart cities, industrial Internet, etc., among which sensors, various control units, edge computing modules, communication modules all need to use chips, among them, from filters, power amplifiers, to embedded microprocessors, most of which come from foreign manufacturers, of course, domestic independent embedded microprocessors, MCUs, Wifi chips, Bluetooth chips, 5G communication chips, storage control chips and other dozens of subdivisions are rising and occupying a small market share in the past 20 years.

Four major challenges, four opportunities and four paths in China's chip industry

Domestic chip manufacturers face 4 mountains, but the trend of high-end chip localization is irreversible.

The ZTE incident revealed the problem that national manufacturers have been having, that is, more than 80% of the chip components are actually still dependent on foreign manufacturers, and even domestic manufacturers in some fields are completely unable to design, which is to unveil the prosperity that is relatively virtual in nature.

Any technology product, Internet of Things device, requires a lot of chips and a lot of control units. The entire chip industry can be subdivided into thousands of subdivisions. In these segments, foreign manufacturers have generally experienced more than 30 years of R&D accumulation, while the development of China's integrated circuit enterprises has been in the past 20 years. Intel has monopolized almost all the PC server CPU market, and has been spending a lot of money for many years, investing in new chip designs, investing in new production lines, and cultivating a huge ecosystem with its own x86 system as the core, constantly deepening barriers, which is difficult to break through.

The four mountains faced by domestic chip manufacturers are:

1. Accumulation and high tolerance for long-term R&D investment. It is reflected in the microarchitecture design, the lack of design capabilities of the underlying operating system, and the lack of its own microarchitecture for general-purpose CPUs (most domestic PC/server operating systems are still based on Linux, in these aspects, foreign manufacturers such as ARM actually broke out after more than 20 years of R&D accumulation), or quickly introduced and snatched up excellent chip design talents.

2. Realize the positive cycle of heavy capital investment and high output.

3. In the short term, it surpasses foreign opponents in terms of performance and stability.

4. Cultivation of hardware developer ecology. Intel and MS have developed curriculum systems, certification systems, and ecological cultivation systems in domestic universities for many years, and domestic enterprises rarely have such cross-level strategic operations.

However, in the future, new fields may break the inherent system in biology, chemistry, physics, optics, such as bionic chips, artificial intelligence chips, etc., many of which are just beginning to be done in Europe and the United States, and domestic manufacturers still have many new opportunities to stand on the same starting line.

I firmly believe that the trend of localization in the future high-end PC, mobile phone, and server market is irreversible. This development requires the market to be inclusive of domestic chip companies and allow domestic manufacturers to make mistakes.

Nowadays, domestic demand-side enterprises and domestic chip suppliers are growing together, 2018 may be a turning year for the chip industry, first from industry customers, in some special equipment to start using, gradually improve the performance and stability of products in the process, and then extend the C-end field. According to this path, in the future, in the field of general chips and high-sales chips, domestic chips will gradually be replaced.

At present, domestic chip-related market opportunities are divided into these four categories.

First, general-purpose chips

There are also start-up companies doing general chips in China, but the performance, yield rate, product quality, and stability are far from foreign chips, so no one has bought them, and they are always in a state of experimentation. A big feature of the chip field is that it needs to find problems and improve problems in the mass production process to improve performance. Domestic chips have not been mass-produced and have not met customer requirements, so they have not been developed.

In addition to market demand, there is another big constraint in the field of general chips that developers do not use them.

A chip must be developed by a group of developers, no one will use your chip, and everyone will definitely not purchase it. Why are NVIDIA GPUs so popular? Its price is high, and it is not a chip with a better neural network computing model, but why do so many companies still use its GPU architecture for training and inference, because its developer system is very complete, there are a series of development tools for developers to use, and it has accumulated a lot of developers in the past, which is a positive cycle.

Domestic chips lack mature development tools and operating systems, and if they cannot be used together, no one will use them. And no one uses it because no one in the end product company buys its chips, which is a vicious circle.

If you are preparing for employment now, you have to spend time learning a new tool, after learning this tool, no one may buy it, you will definitely not learn, you must learn the operating system or the tool that matches the chip with the universality is better, the whole industry is using the operating system or the tool that matches the chip.

In the final analysis, making a good chip is only the first step, and the software and hardware ecosystem supporting the chip is not perfect, even if the chip is made, Made in China will still not become the priority choice. If Made in China wants to meet the world's top chip standards, it also needs an operating system that cooperates well with Chinese chips.

Recently, we can see the trend of increasing the promotion of the state, and the role played by the state:

The first aspect is to provide a large amount of funds, and at the same time have enough space in the shareholding ratio of the founding team of the chip company to maintain the team's control over the enterprise and retain scarce talents. Chip design and chip manufacturing require a lot of capital investment, such as 28nm chips, but also need to invest one to two hundred million yuan to do chip design, tape-out, these all require cost, need to invest a lot of money, if it is chip manufacturing, the investment will be larger, it is tens of billions of dollars of investment, the general enterprise cannot afford to invest at all, so the state does some equity investment, low-interest loan financial support and export tax rebate and other policy support. At the same time, it is appropriate to maintain the team's shareholding ratio in the enterprise, and not sacrifice the control of the founding team due to the investment of funds.

The second aspect is that the national policy guides state-owned enterprises to use PCs, servers, mobile terminals and other products with domestic chips as the core, and state-owned chip enterprises force the use of domestic chip production equipment and services to drive demand. Only by driving demand can we drive the development of the industry, when the demand rises, Loongson, Shanghai Microelectronics and other manufacturers gradually take their products out of the laboratory and continue to iterate products.

If you don't drive demand, if you just give money to support and do research on topics, you will never be recognized by the market. Because in the laboratory stage, as long as a sample can be produced, a demo can implement the basic functions to get the final payment, but there is still a long way to go before it can withstand the test of the market.

Third, the government leads the industrial pattern and retains market autonomy. Statistics show that the state has invested hundreds of billions of funds in the integrated circuit industry, and our government supports the independent research and development of important technologies, but the effect has been little for so many years.

In the private market, some manufacturers of domestic operating systems and domestic chips have cooperated to develop together. In addition to state support and requirement for state-owned enterprises to use them, more and more private enterprises will also use state-owned chips in the future. Integrate scattered resources, form a leading pattern with government support in key areas and large projects of chips to promote the localization of chips, while retaining market power, small projects are independently determined by the market, and the survival of the fittest is achieved through competition.

Incidents like ZTE have made domestic companies realize that chip constraints in the United States or abroad will definitely break out at a certain time node in the future, so that the achievements accumulated by enterprises over the years will be lost overnight, so the market will gradually pay attention to state-owned chips, which is a consensus, and no one wants to become the next ZTE.

This is a long-term development, but this road must be taken, and it is not too late to make amends.

Second, ASIC chips

The ASIC chip path starts from the end product and finally makes a special chip. The risk of directly producing chips is relatively large.

Products like sensors, sell products first, do not make chips, based on FPGA pilot market, it is more cost-effective when the initial volume is not large.

Another idea is to make modules based on heterogeneous chips first, and then ASIC after the future volume is large.

FPGA + main control chip + other components to form a SOC, first make a chip that can be used in the terminal, do artificial intelligence on the terminal, and put some reduced algorithms on the terminal chip. This chip is a system module that combines a variety of chips, and the solution is first made for downstream end product manufacturers.

Many security camera manufacturers, with this module, can make the camera recognize actions, a camera can recognize thousands of faces, can recognize simple objects, can meet daily needs, and can directly iterate intelligently. It is a better path to sell to manufacturers first, enter the market, and then make special ASIC chips after the equivalent gradually rises.

From the perspective of development path, many artificial intelligence companies are taking detours, but there is one company that has a clear thinking path of Bitmain.

Bitmain specializes in ASIC chips and mining machines for mining, and it makes products that can be used directly, rather than making early semi-finished products for customers. Selling algorithms in the AI field does not work, and large Internet enterprise platforms can provide algorithms to users for free, but entrepreneurial enterprises have to survive. Microsoft's strategy of selling operating systems and office software licenses is the first time users use Microsoft pirated products, I don't care, everyone uses it first, and when user habits have been developed, users will transfer to other software and the transfer cost will be very high.

In addition, the algorithm of artificial intelligence in the early days is constantly iterating, and there is still a lot of work that only license customers need to invest in, and only users who have become a terminal form are willing to pay and can really use it, and can gradually build an ecosystem.

Third, vertical segmentation of field chips

The overall sales volume of the chip industry in various fields shows significant long-tail characteristics. The top high-sales chips include general-purpose chips, memory chips, AD/DA and other general-purpose configuration chips, while a large number of various types of chips/SOC/SIP/MCUs in various industries including acquisition, transmission, and control still have thousands of market opportunities. Every market opportunity will not produce giant chip companies, but may produce invisible money-making companies and listed companies, and giants in these fields have no time to take care of them, and the market's focus on products is more about demand matching and services. There is no doubt that the leading players in these segments can form a monopoly advantage.

Chip design in the vertical subdivision field, from the foreign situation, more than 40 chip companies in Europe, the United States, Japan and South Korea are doing some small departments, but such foreign enterprises in the domestic support team and customer service capabilities are relatively poor, and some manufacturers in the Chinese market have gradually used domestic chips, because the nano-level and process level requirements of chips in this kind of subdivision are not very high, and may be 55nm chips can also be used by domestic terminal manufacturers, but they are more sensitive to cost and power consumption, and manufacturers pay more attention to the ratio of cost and performance/power consumption. To achieve such high performance, the price of chips has gone up and no customers have bought it.

Fourth, industrial chain opportunities

In the future, there will be many chip companies in China, and chip manufacturers need a lot of industrial services. Wafer processing alone includes more than 30 links, around chip design, processing, cutting, packaging and testing need to use a lot of equipment and services, among which there are opportunities for domestic enterprises to innovate, Galaxy Internet will continue to pay attention to.