The Pain of ZTE Incident: The Emperor's New Clothes for China's Whole Machine Industry

On the morning of March 7, a friend called me and asked, "Did you know that ZTE has been suspended? The US government has banned ZTE from purchasing." At this time, my attention was still focused on the new banners for this year's Women's Day, and I replied disapprovingly, "I saw the report, and I guess the US government is just pretending." However, two days later, the incident fermented. First, a ZTE netizen broke the news that in addition to not allowing the purchase of chips, US suppliers have completely stopped technical support for ZTE: no longer replying to emails, and when I called, the other party said, "I will pretend that I didn't see your email, and don't call me in the future, otherwise I will be in trouble." Then, I saw that ZTE announced that it was cooperating with the US government to apply for an export license, although such applications are usually rejected. Later, I heard that ARM, a British company, was also forced to stop supporting and cooperating with ZTE because most of its research and development is in the United States. All these made me take a deep breath. It seems that the United States is really playing this time.

Regarding this incident, some people think it is no big deal. Even if Nubia loses Qualcomm, it still has ZTE Microelectronics. Why not use its own chips? Some people think that it is best to impose a complete ban. Now is a good opportunity for domestic chips. However, I think that if the US government's sanctions of cutting off supply last too long, it will bring disaster to ZTE and even the entire machine industry. As the saying goes, if the skin is gone, where will the hair be attached? For domestic chips, if they lose the support of domestic machine manufacturers, what development opportunities will there be? Therefore, the current priority is to get the US government to lift the embargo as soon as possible, get through the current difficulties, and then plan for the future.

Although the domestic integrated circuit industry has developed by leaps and bounds in recent years, and the self-sufficiency rate has increased year by year. Huawei HiSilicon's latest Kirin chip can compete with Qualcomm Snapdragon 820; Loongson has accumulated more than ten years and can finally go to space with Beidou satellite; if you take apart a Bluetooth speaker, set-top box, refrigerator or washing machine, most of the core chips inside are domestic brands. But the current situation that cannot be ignored is that the successful application of these domestic chips is mostly in the consumer field. In large-scale applications in communications, industry, medical care and military that require high stability and reliability, domestic chips are far behind the general international level. In particular, some key components with high technical content: high-speed optical communication interfaces, large-scale FPGAs, high-speed and high-precision ADC/DAC and other fields are still completely dependent on American suppliers.

Entering the second decade of the 21st century, the Wassenaar Arrangement, which restricts the export of high-tech products to China by Western countries, is still in effect. The above-mentioned chips are the hardest hit by export restrictions. If you want to see the true level of China in these areas, just check the updates of the Wassenaar Arrangement every year. In modern phased array radars, they are all necessities and can only be obtained through "you know" channels. For every domestic oscilloscope produced, the ADC inside needs the consent of the US government to be imported, and at the same time, it must promise not to be transferred for military use. Open the base stations produced by ZTE and Huawei. Except for a few digital baseband chips on the circuit board, there are no domestic suppliers for RF, PLL, ADC/DAC and even peripheral chips for measuring power supply voltage on the communication link. Although the whole machine factory has mastered the core algorithm through the self-produced baseband chip, it cannot solve the problem of being "stuck" by foreign chip suppliers. People who understand the whole machine industry know that if a base station has 100 chips, as long as one of them is banned, the entire base station cannot be delivered. Even if they find a team to redesign, according to the inherent rules of IC R&D, it takes at least one year for a chip to be designed, tested and mass-produced, and high-reliability industrial-grade chips take even longer. If the sanctions last for one year, all of ZTE's products will be out of stock, contracts will not be fulfilled, and there will be no income at all. The result is self-evident. The lips and teeth are cold, so even if domestic ICers make chips for ZTE a year later, what's the use? This time, the US government has a hand on ZTE.

It is true that the level of China's electronic equipment industry has made great progress in recent years. Huawei surpassed Ericsson to become the world's largest communications equipment company. Several other companies were forced to merge continuously, and finally ZTE was able to squeeze into the top four in the world. The product level of domestic large-scale medical equipment such as United Imaging and Mindray is close to that of giants such as GE and Philips. Domestic radars have completed a leapfrog leap in active phased arrays, and high-performance weapons such as 052C/052D and J-16 have been put into service. Their radar standards and performance have already approached those of the United States, surpassing Europe and Russia. Just when military fans are celebrating and their pants are as red as they can be, the fact that cannot be concealed is that the key to the lack of "cores" has actually been in the hands of Americans.

Looking back at history, the breakthrough of China's electronic equipment industry is actually the result of the evolution of electronic technology and the changes in the world's division of labor. The core of electronic equipment is algorithms, software and hardware. Algorithms and software have their own characteristics, and Chinese people can easily catch up with them by relying on their intelligence and diligence. If a customer needs a feature, Huawei can send engineers to work overtime overnight to write it; for 4G base stations, Huawei can complete the configuration with one click, while its competitors need to complete it step by step according to the operating manual. In the early years, Huawei gradually established its market advantage by relying on these. As IC technology develops, hardware integrates more and more functions in chips, and in fact, the technical content is concentrated in the chip. In the past, there were hundreds of components on a circuit board, and debugging and yield were both thresholds, but now it has become one or two chips. As long as you can buy a chip and design it according to the reference circuit, 80% of it can be used. In addition to high-end military chips, Huawei and ZTE can buy almost the most advanced commercial chips in the world. Despite the Wassenaar Agreement, American suppliers are also helping us find ways to bypass restrictions in the face of huge profit temptations. So, after buying first-class chips, we have first-class hardware, and with diligent software and smart algorithms, it is only natural to defeat the lazy European telecom operators. So China became the world's factory and has the world's largest semiconductor consumer market. But on March 7, the US government's sanctions came, and we found that the world's leading complete machine industry was actually built on a foundation like sand, and the emperor's new clothes were stripped off.

In the Internet, we have BAT to compete with Facebook/Google, and in the electronic equipment industry, we have Huawei and ZTE to compete with Cisco and Ericsson. Why is it that in the IT industry, only integrated circuits cannot compete with the United States? This has to start with the characteristics of the IC design industry. Compared with the Internet and complete equipment, IC design has two important characteristics: high trial and error costs and difficulty in debugging. When making an app on the Internet, one version can be released every day. Some bugs are okay, they can be fixed the next day, and the cost of trial and error and modification is almost zero. The circuit board design cycle of the complete hardware is between 1 day and 1 month, and the production cycle is between 3 days and 2 weeks. If there is an error, the cost of re-boarding is between hundreds and thousands, and at most tens of thousands of yuan. IC design, not counting the architecture design, starts from circuit design to wafer production, at least half a year. It usually takes 2 to 3 months for the wafer to be sent to the factory for processing and production. The most important thing is that the cost of a wafer production is at least hundreds of thousands of yuan, and the advanced process is as high as 10 to tens of millions. Such high trial and error and time costs require a very high success rate. The process has to be prolonged and repeatedly verified, requiring close cooperation among multiple types of work. If one person in the team makes a mistake, the chip returned three months later may be a piece of stone. After another round of revisions, it took another three months.

The high cost of trial and error is accompanied by the difficulty of debugging. When writing a software on the Internet, you can set breakpoints almost anywhere in the program to check the current state of the variable or print a log. On the hardware circuit board, almost any signal line can be pulled to the oscilloscope to view the waveform. However, a chip the size of a fingernail contains hundreds of millions of transistors, but only a dozen to a few hundred signal lines can be measured on the circuit board. It is self-evident how difficult it is to infer which transistor design error is based on this pitiful amount of information.

These two characteristics lead to extremely high quality requirements for IC practitioners. The long trial and error cycle requires a rigorous and meticulous work attitude, and the difficulty of troubleshooting requires a set of scientific experimental methods. These two aspects are precisely the weak points of domestic education. Too much emphasis is placed on the memorization of knowledge, while logic and methods are ignored. So when software engineers rely on their own intelligence and diligence to continuously and quickly iterate, ICers often encounter the confusion of pig teammates, which leads to spinning in circles. There can't be more overtime, but there are still delays again and again, and the time to market is still behind. There are also many bugs whose causes cannot be found, and repeated chip experiments have no results, and finally the project can only end in failure.

Behind the difficult industry are huge benefits and commercial values. Integrated circuits are known as the food of the electronics industry. In addition to their great significance to national and industry security, their profit margins have also risen with the technological content. The material of the chip itself is silicon dioxide, which has extremely low cost, and the technology condensed on it determines the profit. Consumer chip products generally have a gross profit margin of 30% to 40%, and industrial products generally have a gross profit margin of more than 50% to 60%. What's more, the American Linear Company, which mainly produces high-performance analog chips, has an average gross profit margin of 90%! Many chips that we cannot design, such as high-end switching chips, have a gross profit margin of more than 99%. Once China and the United States go to war, even if there is no embargo, it will take only a few minutes for the US government to use administrative means to reduce the batch cost of its electronic weapons to 1/10 of ours. It is terrifying to think about it.

Try to break through

We kept working hard and never gave up.

Colleges and universities. Some high-performance key component chips are not large in scale, and it seems that colleges and universities are the main force for breakthroughs. But over the years, it is generally recognized in the industry that the level of colleges and universities is not as good as that of the industry. This is not unique to China, but also the case in the United States. This is closely related to the characteristics of the aforementioned integrated circuit industry. The advantage of colleges and universities is to come up with new ideas, which is quite suitable for fields such as algorithms. The simulation experiment can see the results quickly and accurately. If the simulation is effective, it will basically be effective in practice. At most, the implementation complexity is too high. The cost of chip trial and error is high, the process is long, and there are many types of work involved in the collaboration. If there is a problem in any link, there will be no good results. It is not easy to make a chip at the general level of the industry without falling into the pit. It requires years of accumulation. Students have little accumulation. Even if they have good ideas, they often cannot avoid the countless hidden pits on the road. They die halfway before they see the effect of the idea. The specialty of the school is to do more cutting-edge research, which is suitable for overtaking on the curve. However, integrated circuits are not easy to overtake on the curve, especially analog chips. If you don’t solve the 100MHz problem, those problems will still exist when you reach 200MHz.

Imitation and plagiarism. Military fans are proud of their ability to copy, that is, to see what the US military has, and then we copy it. Integrated circuits can also be copied, which is scientifically called reverse engineering. Although the chip is very small and the circuit density is extremely high, it is still possible to obtain all its layout information through microscopy, photography, etc., and then copy a copy and send it to the factory for production. It seems that you can get exactly the same product. In fact, it is not the case. The layout is equivalent to the machine code compiled by the software. It has poor readability and cannot understand its principles and architecture. There are physical errors and human errors in the layout extraction itself, especially for high-performance analog mixed signal chips, which are very sensitive to the process. Any slight inconsistency may lead to huge differences in chip performance and yield. At this time, the designer cannot understand the principle, and the positioning error is like a blind man looking for a needle in a haystack. Military research institutes generally use this method. Each reverse is like a gamble. Sometimes it is best to make it OK, but once there is a problem, there is basically nothing you can do. Therefore, over the years, except for RF and power amplifier chips with relatively simple circuits, there are few examples of successful reverse engineering of the above-mentioned high-performance PLL, ADC and other key components that can be mass-produced equipment.

Scientific research projects. For more than a decade, the country has been supporting key components through national scientific research programs such as 863/973/Core High-tech, with huge investments. In the later stage, the industry and the whole machine factory were also required to join in to solve the problem of application disconnection. However, over the years, there have been few results that can be truly mass-produced and transformed into actual products. The reason is that the goals are disconnected. There is a saying in the IC industry that passing the laboratory test is just a small step, and there is still a huge workload to mass-produce. Scientific research projects only need to provide a few samples during the review and demonstrate the required performance to get the final payment. Industrial-grade applications need to maintain stable performance under various temperature and environmental changes, as well as solve the yield problem of mass production. How to ensure mass production needs to be considered from the beginning of the design. The architecture chosen by some scientific research units itself determines that the results can only be delivered, not mass-produced. The second is the disconnection of indicators. The project-setting units of scientific research projects do not consider the actual domestic level and blindly catch up with the world's leading level. Regardless of whether the projects in the previous cycle are completed, this year's indicators must be one step closer. The applicants competed maliciously, without considering their own strength. They bid for the indicators when applying, and whoever submitted the highest indicators got the project, regardless of how they would deliver the project two years later. Under such a system, according to the existing technical accumulation, 100MHz could be mass-produced, but after the indicator bidding was completed, the target became 500MHz, and in the end no one could handle it.

Talent introduction. Around 2000, the country used talent policies to attract overseas students to return to China to start businesses. During this period, Chen Datong and Wu Ping returned to found Spreadtrum, and Wei Shuran returned to found RDA and other domestic IC design companies. These companies may have had the ambition to make industrial-grade products and key components at the beginning, but they soon found that the industrial environment was not suitable. China's complete machines were not as strong as Huawei and ZTE today. The market capacity was small, the technical reliability requirements were high, and the design-in cycle was long. Therefore, the companies that successfully survived in this group relied on the consumer market and a wave of Chinese knockoff mobile phone craze around 2008 to complete the original accumulation and enter a virtuous circle. However, it was not so smooth sailing for the introduction of industrial-grade and key component talents.

First of all, there are very few suitable candidates. For example, in the United States, due to the restrictions of the Wassenaar Arrangement, Chinese people cannot enter the core ADC product R&D departments of companies such as ADI/TI. Even in their R&D centers in China, mainland engineers can see most of the parent company's designs through the Internet, except for high-performance ADC products. This is simply another replica of the 1990s when the Meteorological Bureau was locked in a glass house with a supercomputer.

In 2009, a Dr. Li returned from ADI in the United States. He brought back the high-performance DAC product map through illegal means, which immediately improved the performance indicators of domestic DAC products. However, the incident was exposed in 2013 and was protested by ADI and the US government. The Li incident led to the US government's stricter control over Chinese people's participation in the research and development of key components, and stricter scrutiny of international students traveling to and from China. The Vanchip plagiarism of RFMD and the FBI entrapment of Professor Zhang Hao of Tianjin University were successively discovered. Regardless of whether it was true or false, it caused psychological trauma for overseas students who returned to China to engage in the research and development of key components. In order to maintain their personal freedom to travel to and from the United States, many people gave up the opportunity to participate in the research and development of high-performance key components in China.

At the same time, in the international market, Huawei and ZTE need to abide by the rules of the game of international intellectual property rights. Li's methods and products cannot be adopted by regular machine manufacturers, and they have not actually solved the problems in the industry. On the contrary, Huawei and ZTE are more worried about the intellectual property rights of introducing domestic suppliers, and require domestic manufacturers to prove their innocence. Some even go to the exaggerated point of requiring the founders of domestic suppliers not to have ADI/TI resumes, which further leads to a serious lag in the progress of domestic substitution. Finally, in the military field where there is no intellectual property problem, due to the impact of the exposure in 2013, Dr. Li's products have not been equipped and used so far, which is a pity.

The whole machine factory should work hard on its own. The only domestic company that has really made a breakthrough in the field of high-performance key components should be HiSilicon, a subsidiary of Huawei. HiSilicon has made breakthroughs in high-speed optical communications and switching chips because of Huawei's investment at all costs. It is well known that Kirin has made achievements. It has also made breakthroughs in high-speed optical communications and switching chips, and has been slowly eroding the technical barriers built by Broadcom and others for many years from the low end. However, in a previous speech by Ren Zong, HiSilicon was positioned as a spare tire. Ren Zong required Huawei to use the best devices to provide customers with the best performance. HiSilicon would not be used if it could not achieve the best performance. In fact, I think this idea is problematic. There is a characteristic of the chip industry. Many problems cannot be measured in the laboratory. They must be discovered, improved and improved when they are applied on a large scale. If you don't use it when you see that the indicators are not good, you will never have the opportunity to find problems. Then this spare tire will always be a paper paste and will break once it is on the road. In fact, it was Huawei's terminal department that was required to use K3V2 even if it pinched its nose that made today's Kirin.

The state also saw the above problems. In September 2013, Vice Premier Ma Kai of the State Council investigated the integrated circuit industry, and then the state issued a new integrated circuit industry revitalization plan. Instead, an industrial fund was established to help integrated circuit companies through equity investment. At the same time, domestic private capital led by Tsinghua Unigroup combined with government funds to start a crazy shopping spree in the international market. Spreadtrum, RDA, OmniVision and other companies were acquired one after another. But in 2015, Tsinghua Unigroup and the big fund series began to encounter obstacles in their shopping spree. The acquisition of Micron and Western Digital in an attempt to break through the nandflash industry was frustrated, and China Resources' offer to Farchild was rejected. Even the acquisition of Philips' lighting business was terminated because the US government was worried about the leakage of power semiconductor technology. Looking back, except for Spreadtrum, which was originally a domestic company, and OmniVision, which was originally a Chinese company, the state did not purchase genuine core technologies through acquisitions, not to mention key device technologies such as RF and ADC that can be used for military purposes. It can be concluded that Americans are unlikely to sell them.

How to break the deadlock?

The author believes that there are three factors that contribute to breakthroughs in high-performance key integrated circuit components: sufficient financial support, the cooperation of complete machine manufacturers, and a patient team.

How to solve the problem of funds? The 18th National Congress of the Communist Party of China requires that the market play a decisive role in resource allocation. Chip research and development is high-investment and high-risk, and only good luck can lead to high output. The government currently uses large funds to decide on resource allocation, but it may not always be able to select the final winner. In addition, there is pressure to maintain and increase the value of state-owned assets, and moral risks, which will cause its investment behavior to distort. In addition, large-scale investment of state-owned assets will also cause a crowding-out effect, reducing the amount of private capital invested in the industry.

The author believes that the best way is to attract private capital. As long as there is a large enough market and large enough profits, private capital will be tempted. At the beginning, the technical level of the team is not up to the advanced level, and it cannot participate in global competition, so the market scale that can be captured is bound to be not that large. At this time, the state should step in, through subsidies and rewards for complete machine manufacturers, without damaging the cost competitiveness of complete machine manufacturers, to allow domestic chip manufacturers to sell at a high price in the early stage, obtain excess profits, make up for the huge investment in research and development, and attract private capital to enter. In the later stage, according to the number of cumulative chip equipment, subsidies will be gradually reduced, and finally market pricing will be reached to enter the international market to participate in competition. The advantage of this policy is that the money will definitely be spent on competitive market players. Whoever is finally equipped and whoever makes something that can be used will be subsidized. Of course, we must pay attention to preventing fraudulent subsidies.

As for the R&D risks and the risks of choosing the wrong product, let private capital bear them. Private capital will naturally choose the team carefully when spending their own money, and will be able to accept the failure of R&D calmly. In this way, one national subsidy can attract multiple private investments, and as long as one of them is successfully mass-produced, the country will make a profit.

Of course, all of this is based on the premise that we still have a strong, internationally competitive complete machine industry. Only they have the motivation to try out domestic ICs that are still in their infancy. In the process of promoting domestic ICs, the author was most moved by this group of technicians from complete machine manufacturers. They did not need any profit drive, but were willing to help domestic ICs from the bottom of their hearts. Sometimes their superiors allowed them to give up, but they still worked overtime to help domestic suppliers find problems.

A blessing in disguise. Perhaps looking back many years later, the ZTE incident will be a turning point for the domestic IC industry. But no matter what, I sincerely hope that ZTE can overcome the difficulties.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.