During my time at ICCAD, I interviewed senior executives from several companies in EDA, IP, design services, and foundry, and had conversations on eight aspects, including the development and prospects of China's semiconductor industry, as well as possible obstacles that may be encountered during development.
At the 2017 ICCAD Summit, Professor Wei Shaojun, Chairman of the IC Design Branch of the China Semiconductor Industry Association, gave an authoritative summary of China's semiconductor IC design industry in 2017.
Statistics show that as of November 2017, there were 1,380 IC design companies in China, an increase of 18 from 1,362 in 2016, and the overall change rate was not large.
Professor Wei Shaojun believes that the substantial increase of more than 600 design companies in China in 2016 is both inevitable and accidental. The inevitability is reflected in the huge driving effect brought about by the release of the Outline, while the accidental nature is reflected in the unsustainability of such a substantial increase in number. The fact that the number of companies remained basically the same in 2017 reflects that the development of this industry has returned to rationality to a certain extent, and as time goes by, a reasonable reduction in the number of companies in the future is also foreseeable.
In terms of sales, according to statistics from the IC Design Council: the industry's sales in 2017 are expected to be RMB 194.6 billion, up 28.15% from RMB 151.8 billion in 2016. This is the fastest growth year for China's IC design industry in recent years, with a growth rate of 23.04% in 2016. Based on the exchange rate of 1:6.65 between the US dollar and the RMB, the annual sales reached US$29.263 billion, and the proportion of the global integrated circuit design industry increased significantly to 30%.
In terms of IC product structure, compared with 2016, the number of companies engaged in communications, multimedia, navigation, power and consumer electronics has increased, while the number of manufacturers in the three fields of smart cards, computers and simulation has decreased. Among them, the number of outstanding communication chip design companies increased from 241 in 2016 to 266, and the corresponding total sales increased by 31% to RMB 90 billion, while the second-ranked consumer IC products have shown a significant upward trend compared with other categories, widening the gap.
The data and results are exciting!
During my time at ICCAD, I interviewed senior executives from several companies in EDA, IP, design services, and foundry, and had conversations on eight aspects, including the development and prospects of China's semiconductor industry, as well as possible obstacles that may be encountered during development.
Discussion 1: Macro Outlook and Market Hotspots of China's Semiconductor Industry
AI, automotive electronics, IoT, security and 5G are the consensus of most guests.
Lin Rongjian, Synopsys Global Vice President and President of Asia Pacific, first expressed his views: "I think there are two topics that are more attractive in China from 2017 to 2018. One is the extensive application of AI, and the other is the automotive industry. There have been many applications in the past two years, and these two categories have great opportunities."
VeriSilicon Chairman and President Dai Weimin gave a more detailed explanation of AI applications and the automotive industry, believing that one of China's rigid needs in the field of AI is monitoring, that is, AI applications in the security field. In the automotive field, electric vehicles will be China's strength, and many famous car companies have also begun to intervene and introduced Chinese standards. "Overseas car companies still use Chinese standards, which is amazing. We have mastered the standards. Some Chinese companies are also involved in AI algorithms in a comprehensive manner. The question is how to turn them into chips." He said.
TSMC is the world's largest semiconductor foundry. Luo Zhenqiu, the company's vice president of business development, said: "I think China is really on a good track from a macro perspective. IoT, automotive electronics and security are now developing rapidly in China and around the world. It's just that China did not have the ability and financial resources to catch up in the past few eras. In the future, China will start at the same time as other countries and regions in these three applications, and even have some advantages because we have a large market. Automotive electronics will be China's largest market, and IoT applications are related to the population base. China will definitely have different results in the IC field of these applications in three to five years.
"In the field of communications, China has begun to take a leading role and formulate 5G standards. Combining communications with IoT will create a huge market. Now, a lot of computing needs to be done at the terminal, and there is a huge demand for semiconductors, memory, and processors. Add to that the requirements for security, and the overall market is huge. This is not just about general IoT, the Internet of Vehicles industry is a part of IoT. China will take a leading position." This is the view of Lee Chu, vice president of sales at Kilopass.
Ms. Xu Yun, General Manager of Cadence China, said that Cadence is optimistic about the development of the entire Chinese integrated circuit industry, especially since the next round of big funds has been approved, and the investment in this industry will become larger and larger. "As for popular applications, we see several key applications that will be focused on in the next few years: the first is AI, the second is the Internet of Things, including the Industrial Internet of Things. The Industrial Internet of Things may have a better application space; in addition, there is automotive electronics. The concept of smart cars is very popular in China, and we also see a lot of realistic opportunities." It is reported that Cadence has already started extensive cooperation with some advanced smart car manufacturers.
Zhang Jingyang, CEO of Moore Elite, is optimistic about some small-scale smart hardware and wearable devices, and believes that these points will bring segmented chip demand. "In the past, the chip industry was market-oriented, and then the chips were made to serve the market, but now we see some new trends that the terminal is first, and many system manufacturers will rush into the chip field. For example, DJI has been pursuing the autonomy and control of its core chips, as well as the security and stability of the supply chain since its earliest drones," he said. "At the same time, the models of some chip companies are also changing. The core of the entire company is the chip, but the product behind it is the system, which is directly facing consumers and the terminal market. This application may not have a particularly large shipment volume, but it will bring many new ideas and directions to the semiconductor industry."
He Ying, product director of Innosilicon, shared his views from the perspective of IP: From the perspective of IP suppliers, we are optimistic about the data encryption industry. There are other technologies behind digital encryption that deserve attention, such as blockchain; at the same time, we are also concerned about the encryption field of chips related to national security, and use domestic IP to make domestic chips.
Some of the guests attending the meeting.
Second Discussion: How can China's domestic IC industry reach or surpass the international leading level?
Professor Wei Shaojun pointed out in his keynote speech at ICCAD that it is difficult to achieve the growth target for 2020 set in the "Gang Yao". To achieve the target, first, there must be a major breakthrough in major products such as CPU/DSP, memory, FPGA and other product areas, and second, the intensity of product innovation must be increased.
Lin Rongjian took the market size and application scenarios as examples, and believed that China has a foundation in the fields of mobile communications and monitoring. At the same time, he is very optimistic about the corresponding applications of AI, which can be divided into two categories: one is the IP of AI chip security, and the other is the application of AI. He believes that these two points complement each other and will have great growth in 2018 and even in the next few years. In the past few years, China has grasped the second half in the mobile field, but now China has greater applications and can drive the development of more technologies.
"There are more than 1,300 IC design companies in China, most of which are still small and medium-sized with their own characteristics. Is there any way for these different companies to concentrate on one point instead of making the same thing? Based on Taiwan's experience, many manufacturers make similar things at the beginning, but they are gradually classified and have a certain degree of communication. Everyone seizes a hot spot and moves forward without overlapping with each other. Once a topic becomes popular in China, everyone moves together for a while, and some people can get capital to swallow up another company. However, this process cannot guarantee that you will have a truly competitive company, because technology needs to be accumulated and studied, and it will be difficult to expand it. Companies must have vision. I am worried that if they start with a price war instead of gradually accumulating technology, they will lose a good opportunity in history." Lin Rongjian gave his advice.
Dai Weimin answered this question from the perspective of the industry, believing that there are two dividends: one is capital dividend, which is the availability of funds; the other is manufacturing dividend. In his opinion, the dividends that Chinese IC design companies will see in the next two years will be huge. At the same time, he called on funds to dare to invest in high-end, long-cycle companies.
Luo Zhenqiu responded with the customers that TSMC has contacted. "In China, new companies have been contacting us all the time, and the products they make are also varied. This is what I mean by passion. You will pursue long-term and ideal goals. Now I see that the products that are more powerful, passionate, well-funded, and developed earlier are all consumer products. The consumer category covers a wide range. Security is considered a consumer category, which is a big hot spot, and AI is also one aspect," he said. "360-degree 8K image quality is integrated in one system, and the light difference can be compensated. I am very impressed to see it. TSMC has a wide global coverage, and we probably know what new products will be made. China has accumulated technology for more than ten years and is slowly launching it. Now it just lacks industrialization. The ability of domestic enterprises to industrialize and the knowledge and experience required for industrialization are not enough. This may be the next step to find ways to strengthen it."
Lee Ch believes that industrialization must cooperate, but in China, everyone is still in a competitive role. There are not enough cooperation platforms, or the exchange of resources or cooperation between each other is not strong enough, so resources cannot be combined. "I see a huge market for microprocessors and sensors, and I want to see how to combine them. According to the customers we have contacted, this is also a new field. General consumer products are not necessarily high-end, many are 30 euros, and memory and sensors are combined. This market is very large, and there will be more and more products in the future."
Xu Yun believes that China has encountered resistance in overseas mergers and acquisitions, and local IC design companies are working hard to improve their main business. "First of all, there is a data for reference: in the field of AI, China is the world's second largest AI economy, in three aspects: one is investment, two is the number of companies, and three is the number of patents, and it is far ahead of the third place. Take HiSilicon as an example. The Kirin 970 chip launched by HiSilicon has been widely recognized in the entire market, and Cadence's DSP is used internally." She shared, "In addition, the domestic direction of independent control has been advocated. There are also many opportunities for Chinese local companies in this field. For example, AI, image recognition, and security are all combined. China has many advantages in the system direction. For example, Hikvision and Dahua are leading companies in the global intelligent monitoring system. How to make good use of our advantages in the system to develop our chip field is also our future direction."
He Ying mentioned the shared bicycle, one of the four new inventions of China, and believed that China's current development of shared bicycles is better than any other country in the world, but the chips currently used are still foreign solutions. Chinese local companies have begun to design miniaturized smart chips. As long as there is a market and application, they can catch up with advanced foreign companies and technologies in this regard.
Discussion 3: Where will EDA companies go in the AI era?
Currently, major EDA companies are expanding their service coverage, extending to other services in addition to traditional businesses such as EDA tools/IP. At the same time, what development trends will EDA tools have in the era of artificial intelligence? Will the tools become more intelligent?
"I remember very clearly that about two years ago, at an IC conference in Tianjin, many familiar friends asked me why your company was not doing its main business and why it was doing software from EDA and IP. After continuous communication, we looked at this issue more positively after the meeting," replied Lin Rongjian of Synopsys. "I think this direction is right. It seems that the scope has expanded, but our thinking is the same, following the needs of the industry. The development of IoT involves security. When you talk about design, you are missing a big piece if you don't talk about security. Our expansion is moving forward in line with the logic of business needs. Our business balance is also relatively simple. We invest more in areas with greater demand."
"Our view on security is this: if we wait until it has been applied to the system before talking about security, it will be too late, because there is no way to completely block external risks after the system is online. You must consider it from the source, and when you start developing software, you must include the system architecture. Our solutions cover the source to the entire software, and there are corresponding security solutions. In this way, no matter what scenario it is used in, there will basically be a certain degree of security guarantee." He explained Synopsys' security software business.
"Synopsys has grown rapidly in China in recent years, and we are very clear that we cannot do everything, especially since our partners do some things very well. We cooperate with VeriSilicon on design services or overall support for many users, and we have cooperated with TSMC on many R&D projects in the early stage. We do the parts that we are best at, and we should fully cooperate on other aspects."
Lin Rongjian believes that AI is an intelligent process, corresponding to different applications, EDA is also one of the applications, making EDA more intelligent. "We have invested a lot in AI, even because different regions and different product departments have to do AI. Our IT department has developed a public platform called 'Common Machine Learning Platform'. The purpose of forming the platform is to eliminate the need for repeated designs in the development process of different products. EDA itself constantly generates huge amounts of data during the operation of different software. Using such a platform, it can be quickly brought to the back for machine learning, which has helped our development. Many important users use our solutions for design. After machine learning, the design that used to take several months now only takes 1-2 weeks."
Another area is designing AI chips. "AI chips will definitely use heterogeneous architectures. A large amount of computing, plus the ability to run heterogeneous designs, usually requires special tools. The second is verification. In the past, chip verification was really just verification, but now AI will be used in a variety of different products, including cars and autonomous driving. We must be able to ensure that there will be no life-threatening situations in autonomous driving scenarios. Therefore, for new AI applications, the difficulty of verification has increased," he said.
Cadence's Xu Yun believes that basically the major mainstream EDA companies are expanding their service areas in response to the needs and changes of the entire industry and customers. "In order to catch up with Moore's Law, many design companies have not accumulated all the IPs as they did before, and they can complete them through their own design processes. Now the complexity of chips is getting higher and higher, and we must solve the problem of how to cooperate with the development of Moore's Law. Under this premise, several EDA companies are expanding into more fields. One is IP. We have also seen that all important customers basically need cooperation in IP. Another is verification. Now the coordination of software and hardware, including the demand and complexity of software development, is getting higher and higher, which is also the direction we are all moving in," she analyzed. "In addition, the system is becoming more and more complex, which has raised the threshold of other directions, such as packaging and PCB full system design, which will bring more challenges and new development directions. For Cadence, in addition to cooperating with the development of Moore's Law, the second major direction is that the system is driving the development of the entire chip design. We must consider the overall optimization of software and hardware in the whole system and the overall integration of various directions. The accumulation of experience in system applications and the design services for system applications are the general direction of Cadence's development in the future. In line with Cadence's development strategy, the company recently established an independently operated company in Nanjing, focusing on IP and system design services.
In response to the needs of the AI era, Cadence can provide solutions in both CPU and DSP, including DSP IP with neural networks specifically for AI. In terms of AI chip verification, Cadence provides a multi-engine solution to ensure that the final verification can be completed effectively and completely, and it also has hardware acceleration simulation equipment. "How does EDA itself develop in the AI era? Whether it is implementation or verification, or the traditional advantages of analog and mixed-signal circuits, we use deep learning and machine learning methods to optimize the algorithms of our own software. More importantly, if everyone's design data can establish a public big data foundation, we have a good chance of achieving a new breakthrough in EDA and turning it into a learnable model, so that new companies or new design projects can learn from the experience of previous projects and other friendly companies. One problem is how to share data, because these data are the intellectual property of each design company. In this regard, we also want to use our capabilities as much as possible and provide help with the cooperation and permission of customers."
Some of the guests attending the meeting.
Fourth Discussion: The Next Development Roadmap of China's OEMs
In recent years, both overseas and local OEMs have been frequently building factories in China, which can be described as a turbulent trend.
"It was the same in Taiwan 20 years ago. The most outrageous thing at that time was that a pig farmer wanted to build a factory. This was a crazy situation in Taiwan at that time. In China, because there is the National Development and Reform Commission and the government leading the way, the whole situation is much better. It is not difficult to build a factory, but what process to use and where your customers are are the difficulties," said Luo Zhenqiu of TSMC. "In our factory in Nanjing, all the circuit diagrams were designed before the construction started. How to lay out the electricity, how to run the gas, how to drain the water, and what process to use were all very clear. But I found something special in China. How many pieces of 12-inch wafers should be built in the factory? He said 100,000 pieces of 12-inch wafers would be enough. What process and what product to use? Anyway, it is 100,000 pieces. Finally, I need to know where the customers are. If China wants to make an industry solid and strong, it must be clear what to do, what process to use, what to distinguish, what to do next, and who to sell the products to when investing funds, manpower and material resources. These are some suggestions for domestic peers."
Regarding TSMC's future development in China, Luo Zhenqiu used the metaphor of "food and grass go first before the army moves." Food and grass refer to orders. Why did TSMC build a factory in Nanjing? It is because China's integrated circuit design has reached this stage. TSMC now has more than 20 companies in China using 16nm and 12nm projects. It needs a base to serve these customers, and it is the most efficient time to provide local services. "This will naturally lead to the next step. Our next step depends entirely on what kind of good products we can provide to Chinese design companies, and whether Chinese design companies can use them to make very competitive products. This creates demand."
Founded in 2010, Shanghai Huali Microelectronics (HLMC) was one of the first companies to enter the 12-inch foundry business in China. Shu Qi, vice president of the company, shared Huali's construction process: HLMC's first phase has been successfully completed, and the production capacity of 35,000 pieces is even fully loaded until the first half of next year. This demand does exist. When we first built the factory, China's design companies were very weak, but now they have reached the second place. China's future development is indeed good for us, so our second phase is 40,000 pieces, mainly focusing on 28nm. Phases 3 and 4 are also being prepared, and the piling of Phase 3 has been completed, also 40,000 pieces.
"I think it's a good thing for places like Nanjing where a hundred flowers bloom. If the competition becomes stronger and bigger, it can create a better platform for domestic design companies and give them more choices. The challenge for us is that with more choices, we have to improve the process," said Shu Qi. "When I visit customers, they usually have active and passive requirements. The active requirement is that the customer's new product requires such a process to be achieved, such as power saving, smaller size, and faster speed; the passive requirement is that this node has been done very well, but why should we move forward? The cost of each node is very high, but if we don't move forward, our competitors will move forward. Passive movement forward is also a kind of motivation. If there is no 14 or 16nm, many people will not move forward, and the entire development may be limited to 28nm, and there will be no motivation to develop."
Fifth Thesis: New Requirements for Design Services
As China's IC design industry develops, new requirements are being placed on design services. As a representative of design service companies, Dai Weimin from VeriSilicon attended the interview.
"First of all, we will never compete with our customers or make products secretly. We have perfect IP cores, and in some areas we are better than American companies," Dai Weimin explained. "Now the companies providing cloud services are getting stronger and stronger. This is a change. We used to serve many chip companies, but now we need to serve companies such as Alibaba, Tencent, Baidu, Amazon, and Google. They are very eager to make chips. They are leading the industry and need customization. Now the cloud is full of videos. Who is leading the entire industry? Some customers are also moving towards the system level, which is also in line with this trend."
Lin Rongjian agreed with the above view: "It is a trend that the share of system manufacturers is increasing. The entire electronics industry and semiconductor industry used to be well-organized and each did their own thing, but now they are expanding to the periphery. For example, the automobile industry is a very traditional industry. In the past decades or hundreds of years, it was very clear what it did, but with the popularization of automotive electronics, automobile manufacturers are also moving upstream. We are all manufacturers that provide solutions and must adapt to this situation. The share of systems will increase, but ICs must eventually be made, and the corresponding supporting industries also need to move forward together. At present, the most urgent problem to be solved is that systems may not understand ICs, and ICs may not understand systems. We also have many system companies. How can we use some technologies and methods to enable these two groups of people to communicate and exchange in the early stage? This is why we are talking about this very new and important change now. The scope of this type of application in the future will be longer."
As for whether this will disrupt the original IC ecosystem layout? "It is already happening now, and it is difficult to distinguish what kind of company you are," said Lin Rongjian.
"Huawei and Xiaomi are also making ICs, and so is Lenovo. They all have IC design departments. It just depends on whether they can make good use of their IC strengths. It is very simple for most companies to make ICs. If they do it just to save IC costs, the effect will definitely be poor. If they do ICs for differentiation, they will quickly succeed. If a successful system company in China enters the IC industry with the purpose of differentiation, then its chances of success are very high, and it has the scale to support it," Luo Zhenqiu added.
Dai Weimin concluded: "Chip companies move upwards, first turning a board into a chip, writing software, and doing a lot of system work. When chip companies move upwards, system companies move downwards, but they don't have chip teams, so they look for us, a design service company. Both upward and downward companies may look for us."
Sixth Discussion: Opportunities for IP Manufacturers in China
The demand for IP involves all aspects, and here we take memory and hybrid IP as an example.
Kilopass is currently promoting memory IP in China. Lee Chu of the company has seen that the demand for memory chips has increased from simple storage to security in recent years. "In the past, it may have only been used to store simple keys, but now there are systems for storing keys. In addition to processors, memory is an indispensable and important component. China has paid more attention to flash memory in recent years." He said, "In China, this part has not been studied, mainly because it requires a lot of long-term technical accumulation, which is also relatively weak in China. IP is not very pleasing. Unlike R&D products, which can be sold for millions or tens of millions, IP is a basic skill that needs to be developed slowly. I hope to do it more deeply in the future."
Innosilicon Technology provides hybrid circuit IP. He Ying from the company believes that the hot spots of IP demand in the next one to two years include the following: On the one hand, IoT chips require lower power consumption for some new process nodes, and there are new IP requirements. Such demand reflects the driving force behind it. We must respond ahead of the market and prepare the IP; the second aspect is high-performance applications. Now we know that whether it is AI or other high-performance computing fields, a relatively large bottleneck is bandwidth. The current approach adopted by everyone is to seek direction on DDR, but in new application fields, this speed can no longer meet the requirements, and a faster interface is needed. It is now only used in graphics cards. This is why some companies can lead most other companies in AI because they have mastered key technologies, which is also the direction we should work hard; finally, it is the cost-effectiveness of our IP and our services.
Seventh Theory: China's Differentiation Path for Semiconductors
Will differentiation be transformed into the core competitiveness of each enterprise?
"Speaking of differentiation, how can Internet companies increase customer stickiness with their applications? Many of them are based on hardware systems. Many system design companies have new requirements for hardware. These requirements are also due to the changes in the applications themselves, which have led to more and more customized demands for hardware. This is a mutual effect. This actually has a very big impact on our industry, which is why we put forward the big strategy of 'system design implementation'. For us, the thinking of the entire company's future development has indeed been affected, and we are also making corresponding adjustments and changes."
"Take the automotive field as an example. In addition to basic work, we also need to accumulate experience in applications. We have cooperation with the most advanced smart car companies in the world, and we have also specifically acquired a lot of application technologies related to automotive electronics or smart cars. We are now working in this direction. Although there are certain bottlenecks, we also hope to provide some system-level support for the entire automotive electronics industry." said Xu Yun of Cadence.
Self-control can also be understood as a way to achieve differentiation. In this regard, Shu Qi shared his views. "Self-control means that you have to master the core technology. Some small and medium-sized companies that have just started do not have so much energy and so many independent intellectual property rights in the early stage. They have to rely on some powerful partners to help them. In this way, many technologies and aspects are in the hands of others. It is more difficult to continue to develop after the product is completed, but this is a process. It is the same at the beginning. China's development will one day achieve self-control. For example, HiSilicon was not completely self-controllable at the beginning, and many technologies relied on some manufacturers. However, after a period of running-in, its technical strength has accumulated, its business has been done well, and its investment in research and development has become larger and larger. In the end, it can make its own replacement chips, and it has mastered this technology. This is also the case for us. We must master core technologies and develop our own self-controllable brands, so as to support the booming IC design companies in China."
Zhang Jingyang also shared his views: If you want to surpass the previous company, you may never be able to catch up by following the same path. Replacing Qualcomm is definitely not the same way to replace it. We must make good use of our home advantage. Since almost all electronic products in the world are made in China, we must find opportunities for differentiation here. Mobile phones have already made the industry chain very flat, which is the most convenient for everyone to use, but in fact there is room for optimization in each application, which requires domestic manufacturers to do more than simply replace foreign products. This road is easy at the beginning, but it becomes more and more difficult as you go. You will encounter patents and bargaining with others later. At the beginning, you have to innovate and define new products by yourself. This may be difficult to do, but it is often easier to do difficult things than to do them. If you can start doing something that is difficult and make some progress, and persist, there will be barriers behind it, and it will be difficult for foreign companies to surpass us.
Eighth Discussion: The most pressing issues that need to be addressed for China’s semiconductor industry to maintain sustainable development
Below are the elements that the guests believe need to be addressed most or are currently lacking in order to ensure the sustainable development of China's semiconductor industry in 2018 and beyond, as well as action suggestions.
Dai Weimin: I think it is a talent dividend. There should be funds specifically dedicated to investing in start-up companies.
Lin Rongjian: It is not easy to make technological breakthroughs now. Technology is created by people, and people are a big limiting factor. This needs to be solved. We need to plan the entire education system and start from the bottom.
Luo Zhenqiu: I think we have the support of the Chinese government, funds, a lot of people, and excitement. We have all the necessary resources. Everyone must have the spirit of an entrepreneur. When doing something, we must focus on it, do it well, and make it strong and solid.
Lee Chu: In addition to passion, we need to be patient and keep going, see how to cooperate with universities, and hope that the industry can continue to support university research and development. Many cutting-edge technologies are still developed by universities, and I hope to pay more attention to them.
Xu Yun: From the current perspective, the development trend of China's semiconductor industry is unstoppable. The most fundamental driving force for China's potential development is the huge market and the data that will be the biggest development in the future. This is the biggest value that distinguishes China from the United States and any other country in the world. We should increase investment in AI or big data to truly transform our data into resources that are useful to us. Increasing investment in this area has the opportunity to create a good opportunity to surpass Europe and the United States.
He Ying: The most urgent issue to be addressed is how we can become game makers. We need to participate more in the formulation of international standards and rules so that we can have a say on the world chip stage.
Shu Qi: 2018 is indeed a very good opportunity, both in terms of the market and the opportunities. Therefore, all chip companies, including our company, should do their own things well. Research and development should not only be invested, but also done in a down-to-earth manner. I think the media should give more suggestions to the country. From the national level, we should see the existence of this opportunity and provide more policy support, especially for some start-ups. The country should support small and medium-sized enterprises, especially those that have just been established and have good ideas, and should not be afraid of their failure, because many times success comes after failure.