On January 10, 2006, Apple’s annual Macworld conference was held, but it was not Jobs that grabbed the spotlight at the conference, but Intel CEO Paul Otellini wearing a lab coat. In the smoke of dry ice, Otellini slowly walked towards Jobs, solemnly handed over a silicon chip to the latter, and said loudly: **"Report Steve, Intel is ready!"** The whole speech lasted about 2 hours, Jobs Announced the first iMac and MacBook Pro with Intel Core series processors, and the 14-year honeymoon period between Apple and Intel kicked off. <a target="_blank" href="http://www.baidu.com"></a> <div style="width:100%; text-align:center;" > Jobs and Otellini at the conference, in 2006</div> Although Apple Computer had a market share of only 3% at the time, in Jobs' plan, the cooperation with Intel was still only temporary, and the chips had to be designed by themselves. Four years later, Apple released the first generation of iPad. Jobs introduced excitedly: The iPad is equipped with the most powerful chip we have ever made, so powerful that it makes people scream. This is **Apple A4Chip**. Although the chip that was named Apple for the first time did not shine, and Wall Street analysts poured cold water on it, Apple's plan to build the core did not stop. After another ten years, Cook, who took over the mantle of Joe’s gang, announced Apple’s "One More Thing": **the first self-developed M1 chip for Mac computers** in the early hours of Double Eleven in 2020. This chip encapsulates nearly 16 billion transistors and integrates CPU, GPU, NPU and other components together. It can complete 11 trillion operations per second. It is different from the market’s indifference towards A4Chip ten years ago. M1 allows the industry shocked, and it gave Apple enough confidence to break up with Intel. Whether it's from the brink of bankruptcy to AMD, or the "Ni-Liu Controversy" of China's business history, China and the United States have never lacked challengers coveting Intel's dominance in the past three decades. Since 1994, the Chinese IT industry has launched an attack on CPUs. From Ark, Shenwei, Loongson, to Zhaoxin, Haiguang, Feiteng, and to Huawei’s breakthroughs, Chinese companies’ obsession with domestic CPUs has never faded. However, after the "independent faction" and the "introduction faction" took turns to battle, the industry cycle has experienced several ups and downs. These companies and products all add up and it is still difficult to shake the pattern of the CPU industry. Behind the emergence of super projects, the semiconductor industry represented by CPUs has become a hidden pain in a manufacturing country. **In ten years, Apple's CPU has completed the leap from A4 to M1, spurring Intel; and China has been fighting the road of CPU for more than 20 years. After ups and downs, how can it get out of the quagmire?** ### Son of the Chosen ------------ At 6 o'clock in the morning on August 10, 2002, with the word "login:" flashing on the computer, a burst of cheers broke out in the Institute of Computing Technology of the Chinese Academy of Sciences-the Loongson 1 CPU finally worked. The engineers who hadn't closed their eyes all night were even more excited, and they took a taxi to Tiananmen Square and queued for more than an hour in front of the leader's memorial hall and gave the leader a report on Godson's work [2]. This CPU was named "Xia 50" by Hu Weiwu, 32, in commemoration of his doctoral supervisor, "Mother of Computers in China" Xia Peisu's 50th anniversary of his computer career. More than a year before the advent of Godson, Li Guojie, the leader of the computing institute who advocated the launch of GM's domestic CPU, pulled out 1 million yuan from the tight scientific research funding, and raised the banner of CPU. Faced with this hot potato, Hu Weiwu issued a military order: **If it can't be done, come and see you.**  <div style="width:100%; text-align:center;" > Hu Weiwu, Chief Designer of Loongson</div> The success of Loongson not only means a breakthrough in domestic CPU design, but also a result of chip manufacturing. It was the famous SMIC who taped out Loongson back then. In addition to the engineers who participated in the initial creation, **the Loongson's short-term success also benefited from the "limited autonomy" principle of domestic CPUs at that time.** As the name implies, the so-called "limited autonomy" refers to self-research based on overseas streamlined instruction sets. At that time, there were two instruction sets in the CPU field (similar to the command method). One was a simplified instruction set, which could be understood as a command that only did one thing; the other was a complex instruction set represented by Intel, which was expensive and technical complex and powerful. Under the conditions at the time, all instructions were self-developed from 0 to "completely autonomous", and would face a situation where there was no compiler, no operating system, and no software available. It was neither realistic nor efficient. It was even more useless to benchmark Intel’s Complex instruction set. Therefore, domestic companies have focused their attention on the streamlined instruction set overseas, and are doing independent research and development on this basis. In addition to Loongson, Feiteng developed the first CPU as early as 1999, using the SPARC architecture. Shenwei is developed by Wuxi 56 with a long history and adopts the ancient Alpha architecture. The Loongson of the Chinese Academy of Sciences is based on the open source MIPS architecture. Although MIPS is now unknown, as a product of the micro-architecture scuffle in the 1990s, MIPS64 is not only the first commercialized 64-bit architecture, but also a complete micro-architecture design and relatively complete software chain support. In the same period, with the support of many national funds such as 863 and Nuclear High Base, domestic CPU ushered in a spring breeze: basically maintaining the speed of producing one chip every two years, and spare no effort in catching up with performance. The Loongson II in 2004 has been able to match the Pentium III in 1999. However, in a seemingly good situation, domestic CPUs are also facing three problems: **Performance problems**：Although domestic CPUs have made great progress, their opponents are also improving, especially Intel’s entry into the new century. Poor;**ecological problem**：the x86 patent wall is towering, and the domestic independent CPU adopts a unified bypass strategy, but it is difficult to adapt to the Microsoft system;**mass production problem**: users have adapted to the Microsoft operating system, a barren desktop, It is not attractive, so how about mass production. At the beginning of the domestic CPU, the research groups of Godson, Feiteng, and Shenwei were originally small scientific research workshops in the national research institutes. They belonged to a product left over from the planned economic system. After 20 years of reform and opening up, they have been impacted.**One is that the budget is extremely limited, and the other is that the supporting industrial chain has long been lost**. Under two major constraints, domestic CPUs can only strive to complete R&D tasks assigned by superiors within the tight fiscal framework of the budget. In this case, the streamlined instruction set architecture has become a stopgap measure: due to the narrow market, their license prices are often so cheap that people can't believe it, and Feiteng originally purchased SPARC licenses for only $99. Architectures such as SPARC and Alpha are mostly built behind closed doors by many large companies that have clashed with Intel. They are well-stocked and are more willing to draw alliances to compete with Intel at low cost. The MIPS architecture adopted by Loongson also has a special advantage: as an independent architecture that does not rely on any major manufacturer, MIPS is more fond of embedded devices such as game consoles, optical discs, and set-top boxes. These scenarios start at a fast rate and come in quickly. At the time when the design industry in Taiwan was surging, MIPS' business on the island was doing well, and the license fee was soft. However, like most foreign technology companies, MIPS faces difficult intellectual property issues in the mainland. Previously, Loongson had been working on its own MIPS-based computer CPU without authorization, but since there was no commercialization, the two sides were also in peace. But in 2005, Loongson and Microsoft cooperated to develop Windows CE, and when Loongson boards were about to be delivered to the United States, MIPS sent a fatal email to Microsoft: **As long as Loongson set foot on the land of the United States, it would violate the patent rights of MIPS**. Loongson sees the handy Microsoft buff, which is very easy to beat. The warning letter of MIPS revealed an almost default lid in China’s scientific and technological circles: at the beginning of the reform and opening up, foreign companies paid little attention to the Chinese market and did not apply for patents simultaneously. If you apply for a patent within one year, you will not be able to apply for it later. This creates an unprecedented opportunity-but the fly in the ointment is that this foundation only works within China. Without the protection of the mainland patented pot cover, Loongson suffered a series of patent snipers: In July 2005, the semiconductor research organization In-Stat released an independent analysis report saying that the Loongson processor architecture and MIPS are similar to 95%, due to unauthorized. Once Loongson begins to market, it may cause intellectual property disputes. As a key project of the National 863 Program, the Institute of Computer Science of the Chinese Academy of Sciences urgently refuted the rumors. As the research team leader, Hu Weiwu also came forward to clarify that Godson was independently developed by the Chinese, and the instruction system deliberately avoided foreign patents. But MIPS obviously cannot be satisfied with this statement. After several negotiations, MIPS offered a sky-high price of $1 million. In 2006, the infamous Hanchip incident broke out in the chip industry. With SMIC's surrender in the TSMC lawsuit, the entire public opinion carried out undifferentiated rhetoric against the semiconductor industry, and negative reviews were overwhelming.  <div style="width:100%; text-align:center;" > Chen Jin, who was holding the "Hanxin No. 1" chip in his hand, showed off the chip</div> In 2003, his eyebrows were burning and the crisis had to be resolved. It coincided with the exchanges between China and France in the spring of 2007. Taking advantage of this political opportunity, the Chinese Academy of Sciences, through its partner STMicroelectronics, allowed Godson to indirectly obtain all the patents of the MIPS architecture. Under the leadership of the leaders of China and France, Loongson and STMicroelectronics also held a press conference in the Great Hall of the People. It was announced that on the day of cooperation, the large headline of "People's Daily" wrote: <span style=”color:#CE0000;”>China Loongson faces the global market</span> However, this emergency strategy has natural flaws: using STMicroelectronics to buy patents, tape out films, and sell chips, who is the final chip property right? Hasn't Godson become an outsourced design studio of STMicroelectronics? Two years later, in order to solve the problem once and for all, the Chinese Academy of Sciences cut and twisted and signed a contract with MIPS directly. **But unexpectedly, shortly after the signing of the contract, a certain media published an article entitled "Loongson's helpless purchase of U.S. company patent authorization, "Chinese core" halo faded out", Loongson's behavior of spending money to buy MIPS architecture patents was infinitely magnified**. For a time, the voices of "Loongson fraud", "Loongson cheating", "Loongson is not self-designed" and other voices one after another. Coupled with the bad influence of the Hanxin incident before, public opinion has reached a culmination in the question of domestic chip independent research and development capabilities. The domestic CPU is the same as the domestic operating system. Because the core technology is controlled by others, there is a natural fear of "stuck neck". Regarding independent CPU chips, there were always three groups of opinions in the industry at that time: One group believed that domestically-made CPUs should be started from scratch, completely "autonomous", and abandon the civilian market to serve the military and government only; the other group believed that there was no hope for technology to catch up. Funds are saved to tackle other areas, **"Isn't there only one Intel in the world? Isn't the whole world safe except the United States?"** Hu Weiwu and Loongson represent the third faction: complete independent property rights are too difficult in the short term, and independent innovation is not necessarily guaranteed. Start from scratch. Faced with the controversy caused by the purchase of patents, Hu Weiwu had to give the media a helpless and sad computer principle lesson: **buying a chip architecture is not the same as buying a chip; with the architecture, a lot of independent research and development work is still needed.** However, under the negative shadow of "Skinning Hanxin" a few years ago, Godson, as an early bird, still suffered from all kinds of criticisms. During the time when domestic CPUs were rushing together, the industry also gradually reached a consensus on the future development direction: **technically, complete independence is not realistic in the short term, and only independent innovation based on overseas patents; ecologically, peripherals and applications support users’ experience, in the market, the civilian market is difficult, temporarily led by government procurement, focusing on the independent information market.** Among these questions, the market route is the most difficult to answer. In the field of government procurement, there is often a "chicken and egg" paradox-the government's requirements are easy to use and can be used, that is, products that have been verified in other markets. However, domestically-made CPUs were not used at the beginning, and they also hoped to support one from the government and then shift to the civilian market. On the other hand, domestically produced CPUs want to open up the market, provided that they are market-oriented first, and their role is shifted from "research group" to "company". In 2010, the "Eleventh Five-Year" plan was coming to an end, and state subsidies were tightened. The three domestically produced CPUs ushered in a turning point in their fate: Shen Wei was hidden in the system and the supercomputer market and did not attack downwards; Feiteng broke out a dispute, the two factions are arguing in front of the government leaders whether to abandon the original establishment. Hu Weiwu gritted his teeth and decided to let Godson take the road of marketization, focusing on the military and government markets and the low-end desktop system market. Subsequently, the Loongson research team collectively resigned from the Chinese Academy of Sciences, gave up its identity within the system, and formed the Loongson Zhongke Company, trying to make blood through marketization and industrialization.  <div style="width:100%; text-align:center;" > One of the Loongson 3 chip series</div> he Loongson 3B1000, which was developed in 2008, is the last Loongson chip in the Chinese Academy of Sciences. Loongson 3B1000 is based on 65nm process, the target main frequency is 800MHz-1GHz, and the eight-core structure is at the world's leading level at the time. The film was taped out in June 2010, and the first batch of chips was delivered at the end of November. During the research and development period of the Loongson 3 represented by Loongson 3B1000, it happened to catch up with the end of the 863 project, and the "Nuclear High Base" has not yet started. In order to ensure the development of Loongson, the Academy of Sciences has allocated 5 million yuan, and the Ministry of Science and Technology has also urgently arranged 20 million yuan in the 863 plan. At the middle-level cadre meeting of the Institute of Computing Technology, the former director Li Guojie specially set up a topic for Godson without a funding limit, telling him: **"Hu Weiwu, the computing office is to sell iron and support the research and development of Godson."** Wait until the first "nuclear" of Godson. When the funding for the "Gaoji" project arrived, the Loongson research team had already advanced nearly 80 million yuan in funding from the Institute. **But no one thought that the Loongson 3B1000, which was placed high hopes, would not even be able to start the operating system in the end.** Loongson adopts a multi-core R&D strategy and wanted to complete its work, but the Loongson 3B1000 just caught up with Loongson's shift from a research group to market-oriented operations. Originally, Hu Weiwu, as the leader of the research group, undertook research and development. Each chip looked at from the beginning to the end from structural design, functional verification to physical design. However, after the corporatization operation, Hu Weiwu's energy was implicated by the company's operations, and the R&D team had no leader, and technical problems were exposed on a large scale. It wasn't until 2013 that Godson's technical problems were gradually resolved, and as a result, they encountered the foundry process Waterloo. Some competent departments and application units have long been unable to withstand such a long "problem-solving-problem-reproducing" cycle. During the same period, rivals on the other side ushered in a period of rapid development. **From 2006 to 2013, the single-core performance of the Loongson CPU only increased by 50%, while the x86 CPU in the same period increased by 5 times. The gap between the two has increased from one to ten times.** The nature of the divergence of domestic CPU lies in the market transformation stage, the decoupling of academics from the industrial world. Under the architectural inertia of the "research group", domestic companies basically relaxed the improvement of single-core performance during the "Eleventh Five-Year Plan" period, but switched to multi-core. The number of cores mostly beats Intel and AMD, but the single-core performance is weak. So that domestic CPUs could hardly find any living space in the autonomous information market that was in full swing at that time. Hu Weiwu later concluded: “**In terms of chip R&D, we overemphasized the technological development trends of multi-core and SOC, and paid little attention to the performance improvement of single-core itself. In terms of performance management, we copied the performance appraisal method of mature large enterprises, not only did not improve. On the contrary, it dampened the enthusiasm of sales staff.**” In May 2013, Loongson suspended the tape-out of the 16-core Loongson 3C processor that had completed its main design, completely stopped the development of the 16-core processor, and started the development of the quad-core 3A2000 processor. But at this point in time, both the policy and the market have lost patience with "domestic independence". Since 2013, "NuGao Ji" basically abandoned the independent CPU research and development route, and instead supported the route of introducing foreign CPU technology. The "introduction group" in the market has also overshadowed the "independent group", and overseas CPUs represented by IBM, AMD, VIA, and ARM have poured in in all directions. For Godson, which has just broken away from the market-oriented operation of the system, this is the last situation they want to see. The introduction of the "introduction" has become the mainstream of the industry after 2010, and then affect the direction of decision-making, there are factors that domestically produced CPUs themselves pull the hips, but more of the changes in the global industrial structure. After 2006, Intel’s old enemy AMD made mistakes one after another. First, it acquired ATI’s GPU business for two-line operations, and it was difficult to cope with toothpaste factory and leather man Huang Renxun at the same time; then, its own wafer foundry, Grofand, overturned in the gutter, and its business survival space was narrow. The superposition process upgrade was blocked, and the stock price fell from more than $40 at the peak to just over one dollar. In 2012, AMD suffered a loss of more than $1 billion and layoffs 15% of its employees. Wall Street handily sentenced AMD to death: "There is no investment value." In the midst of the storm, Huang Renxun's distant relative, Su Zifeng, took over AMD's command. And as frustrated as AMD, there is VIA Electronics founded by Tiger Queen Xuehong, the father of Formosa Plastics Wang Yongqing.  <div style="width:100%; text-align:center;" > AMD’s new head, Su Zifeng</div> In the 1980s, VIA began to erode Intel’s territory from around the CPU. In the 1990s, Wang Xuehong had acquired two American companies that developed CPUs and owned many x86 architecture patents. The region sees the hope of industrial upgrading. But after the new century, Intel began to develop its own chipset. Coupled with Wang Xuehong's main focus on HTC phones, VIA has slowly become a tasteless one. The frustration of the two big players in the x86 camp gave the domestic industry hope. In the new round of route conversion of domestically-made CPUs, local governments have become a supporting force that cannot be ignored behind emerging forces. In 2013, Shanghai State-owned Assets and VIA jointly established Zhaoxin. The Wang family has a deep connection with Shanghai. Wang Xuehong’s brother came to Shanghai at almost the same time as Zhang Rujing to establish a wafer foundry. VIA established a foundry in Shanghai as early as the beginning of the century. Point, the mainland design team has been developing for more than ten years, Zhaoxin stood on the shoulders of VIA and saved a lot of effort. Like Loongson, who was born under the Chinese Academy of Sciences, Zhongke Shuguang chose to establish Haiguang with AMD in 2016. Behind Haiguang is the strong support of Chengdu's state-owned assets. Li Guojie, who "selled iron to support Loongson's research and development", led the research and development of Sugon server in the 1990s. In the following two decades, Sugon became a domestic server manufacturer that rivaled Lenovo. For China, AMD's Zen architecture is powerful and an attractive option. After paying about US$300 million in cooperation fees, Haiguang was finally able to sell products in mainland China through 11 process authorizations of the two-tier company. **Although the introduced chips were largely castrated by AMD after joint research and development, their performance still far exceeds that of many domestic CPUs. The architecture inherited by Haiguang five years ago is still the first echelon player today.** The cooperation between AMD and Haiguang provided Su Zifeng with a generous start-up capital. On the day the news was announced, the company's stock price rose irrationally by 50%. Zhaoxin’s transaction with VIA is also killing three birds with one stone. The CPU R&D team, x86 architecture patents, and chipset industry chain; Haiguang backs to the listed company Sugon, borrowing from AMD’s operations, plus the national secret algorithm, and has a large number of commercial servers. More importantly, the initial funding of the two is 2 billion, compared with the Godson that spent less than 800 million in 15 years, and the introduction of ammunition is sufficient. Zhaoxin achieved mass production two years after its establishment and sold more than 10,000 sets. In the first half of 2020, Haiguang has a revenue of 270 million yuan and a net profit of more than 60 million yuan. The introduction route seems immediate, but there is a fatal risk-changes in the international situation. In 2019, the Sino-US trade frictions have been diving. "The Wall Street Journal" suddenly gave Haiguang a "military background" hat. In the same year, Haiguang was included in the list of entities by the United States, and AMD subsequently announced that it would no longer authorize the new-generation Zen architecture. I bought the patented Zhaoxin from VIA and received a warning from Intel early: Do not cross the minefield. As the saying goes, now and then, compared to the turbulent wind and rain, AMD at this time has gone through the long valley of death and no longer needs foreign blood transfusions. Moreover, at Lawrence Livermore National Laboratory in the Bay Area, AMD chips are used to power the US nuclear deterrence. AMD knows which one is more important. In the end, the line of "usageism" proved to be fundamentally flawed. For a long time, the mainland's x86 talents are extremely weak, and whether "usage doctrine" can evolve to "absorb innovation" is a big question mark. **If you can only stay at the x86 hardware level copy and paste, it is difficult to engage in in-depth research and development, and to meet the high security level and the needs of the consumer market.** This kind of difficulty in absorbing the x86 architecture can be glimpsed from Zhaoxin. In 2010, VIA brought x86 to Shanghai, and the technical person in charge led hundreds of people in a small black room. After looking at the source code of the CPU core for two years, I finally figured it out. After Zhaoxin started to engage in CPU, the R&D team of four to five hundred people worked hard for three or four years, and only maintained one and a half CPU project development. The heavy historical burden of x86 is a huge drag on "absorbing innovation". Fully autonomous, it needs to start from the bottom, it is almost impossible; the introduction of x86 architecture, but also faces the fear of being cut off and unable to update. The next step for domestic CPUs has to return to the most basic problem: **ecology**. ### Ecological exploration ------------ For domestic CPUs, the best teacher is always Intel. The difference is that the introduction party adopts the idea of "importing" and has to copy a localized mirror of Intel in China to ensure the security of supply in the industrial chain. What the independent faction needs to do is to learn from Intel’s historical play, and finally form an industrial chain that does not rely on external supply in technology. In short, the introduction of school form, independent school of God. On the other hand, learning from Intel is not a problem, but which year you learn from Intel is a big problem-when all purchasing units encounter Godson, they will ask a question that hits the depths of the soul: **how far are you from Intel?** After leading Godson to experience the fiasco of "refining multi-core", Hu Weiwu gradually realized that domestically-made CPUs should not learn from Intel now, but Intel in the 1980s and 1990s: single-handedly carry many competitors and take advantage of the people "War" is a revolutionary who wins the battle of the market. In 1985, Intel was beaten by Japanese manufacturers in the DRAM market. CEO Grove stood out from the crowd and turned to the "non-mainstream market" at that time: CPUs used in personal computers. At first, Intel was at a technological disadvantage, but adopted three strategies: **(1) Starting from the low-end market**, creating useful CPUs, and establishing a solid "revolutionary base"; **(2) Maintaining forward compatibility**, although carrying the burden of history, but able to win over a large number of loyal customers; **(3) Open the x86 architecture**, formulate peripheral standards, unite all forces that can be united in the industry chain, and mobilize the masses to fight landlords. These three strategies are Intel's dragon slaying technique to complete the unified war from computer to server CPU. Comprehending Intel’s winning secrets, since 2012, Godson has adopted a three-step strategy: **First, start with the embedded CPU and apply Godson to door locks, learning machines and satellites.** These scenarios were originally natural scenarios of the MIPS architecture, which ensured cash flow; **then stopped multi-core R&D and focused on single-core performance**. This is very important for the daily use of ordinary users, and it also ensures that the Loongson test will not pull the hips; **finally; Take the initiative to unite with all system vendors, software vendors and ODM factories that can be united, and serve them well like a waiter.** The three-step approach has gradually achieved results, and Loongson's revenue exceeded 100 million in 2015, initially breaking even. After passing the ghost gate, Hu Weiwu discovered that for the Chinese who have hardly established a modern information ecology on their own, the ecological work of domestically-made CPUs is intricate and can only refer to another learning object, Apple. After Intel’s business war of the century, Apple actually inherited its mantle, but Apple’s self-developed chip road actually faced two special problems: First, from the top-level application to the bottom-level chip, there are many ridges. Which are the grabbing hands? Secondly, Apple's self-developed ARM chip is extremely ugly at first. If the chip is dragged down, how can the experience of the terminal product be guaranteed? With the deepening of government application testing batch by batch, Hu Weiwu realized that the specific problems faced by domestically-made CPUs in constructing the ecology are Apple-style problems. Under the Chinese environment, they can be summarized into two problems: **First, the industry is discrete and puzzles.**For a long time, China's ICT industry chain is based on the foundation of foreign architecture, with more people doing applications and fewer people making wheels. There are millions of programmers who are proficient in java, but when it comes to java virtual machine, there are very few. The same embarrassment also exists in API interfaces, QT libraries, and even low-level computer languages. **The second is that the system has more redundancy and less system optimization.**The x86 and Microsoft systems carry decades of historical baggage and become more and more heavy and redundant. The introduction and use of these systems has made the information system overwhelmed. But in fact, in the newly-built domestic CPU+ operating system, it can be greatly optimized to enhance the experience. According to these two logics, Loongson attacked on both sides. On the one hand, Loongson simply built its own Java virtual machine, API interface and other intermediate layers, and used these key wheels to connect the past and the next. On the other hand, for the key applications of the party, government, and military, Godson eliminates redundancy and improves efficiency. After optimization of a command system application, the running time of Loongson has increased from 3 frames per second to 30 frames, which is 50% higher than that of Intel products in the same period. With the establishment of ecology and the improvement of user experience, Godson has responded to the challenge, and the domestic CPU industry chain has gradually established two principles: **one is competition within the "fence wall" defined by independent and controllable policies; the other is " The nationwide system under the conditions of a market economy"**. The former is aimed at the R&D intensity of domestic CPUs, while the latter is aimed at the construction of the entire industry chain under fair competition. With the passage of time, these two principles have become a very instructive framework for the development of domestic CPUs. In this framework, the semi-open ARM architecture gives ample room for independent research and development, and there are a large number of mobile applications that can be easily transplanted. With these two advantages, the ARM line has risen rapidly, and finally ushered in a heavyweight. At the beginning of 2019, Huawei released the first ARM architecture server CPU "Kunpeng", and established Kunpeng Computing Industry Alliance with various provinces and cities. In May 2020, Kunpeng CPU entered China Telecom's procurement list. Huawei, as it did in the program-controlled switch market in the past, binds local telecommunication bureaus and co-exists with each other. Finally, relying on technological upgrades and channels, Huawei has completed the localization of switches. On the one hand, "Kunpeng" has brought new breakthroughs to domestic CPUs; but on the other hand, "Kunpeng" has also allowed the phrase "**only Huawei to hit**" to sweep several industries, and cover a new field. There is pride and embarrassment. ### Ending ------------ In 2018, the ZTE incident broke out, and the domestic CPU market began to rise in large quantities. After leaving Lenovo for more than two decades, Ni Guangnan has been running for domestic CPUs. The dispute between Ni and Liu became a good story in the history of China's science and technology. The dispute between "Trade, Industry and Technology" and "Technology, Industry and Trade" has been a topic on social media to this day. Liang Ning, Ni Guangnan's young assistant at the time, described Ni Guangnan as a lofty idealist after the ZTE incident, which caused heated discussions at the time, and the Chinese people all sighed [6]. In 1994, with the removal of trade barriers, Intel, Microsoft and other overseas companies entered China in large numbers. Ni Guangnan rushed to Beijing and Shanghai to plan the establishment of a computer CPU design center, but the attempt failed. Subsequently, Ni Guangnan made another "Ark CPU", similar to the extremely advanced idea of today's Google Netbook. As a result, civil servants were assigned and did not want to use it, and the second attempt failed. The development of many high-tech industries in China has followed these few steps: **the leaders are eagerly caring-the ministries and commissions attach great importance to it-financial resources are invested-enterprises focus on key research-technological breakthroughs**. Most fields rely on this production-university-research model with Chinese characteristics. However, this model has two key prerequisites-**without considering cost and yield**. Today, many of the achievements of China's semiconductor industry that have "reached the international level". **Either can only serve individual niche markets, or rust and dust after review and acceptance.** In the 1990s, changes in the international situation aroused public opinion about technological autonomy, and the slogan of "independent research and development" resounded across the sky. But when Hu Weiwu was leading the research and development of Godson in the research team, the deep environment was stretched scientific research funds, almost groundless consumer market and outdated manufacturing system. What's even worse is that semiconductors are born to be a "**one step ahead, step by step lead**" industry. In other words, until the second decade of the new century, the infrastructure supporting domestically-made CPUs is hardly tidy. As the crown of industry and technology, what CPU needs is not just full of enthusiasm and hard work, but also a competition of industrial supporting facilities, economic structure and even comprehensive national strength. **It is not for nothing, but it is impossible for it.** There is a very interesting question: **Intel, Apple or Huawei, who has the most R&D expenses? Who has the least?** The answer may surprise many people: **Counting from 2006, when Huawei began to publish its annual report, Apple’s R&D expenditure has not been higher than that of Huawei. In 2018, Huawei's R&D investment exceeded Intel for the first time.**  But as strong as Huawei, the launch of its self-developed server CPU is a matter of nearly two years, even earlier than Apple's M1 chip. In Intel's main computer and server CPU base camp, except for AMD, which has been in the hip for a few years, it has been lacking a strong competitor. Some time ago, Hu Weiwu’s “**remarks that 14nm chips are enough**” sparked controversy again. Many people only noticed the remarks “Don’t make 5 nanometers in the United States, we will do 5 nanometers”, and ignored the second half sentence: “**90.9% of the applications can use 14nm. As long as the system is optimized, it can be very competitive in the market**.” Compared with the 30 years ago, China has a market, policies, urgency, and slogans. But for domestic CPUs, it requires idealistic enthusiasm, awe of market rules, and generosity, tolerance, and patience of scientific researchers. **Many industries in China are not afraid of repeated battles and defeats. What they are afraid of is the swing of "today is a big hand, it is better to buy tomorrow.**。 The full text is over. Thank you for your patience to read ###### Reference materials: ###### [1].Steve Jobs Biography, Walter Isaacson ###### [2].Godson's Footprints, Hu Weiwu ###### [3].Embarrassing China Chips: The Difficult Journey of Godson CPUs, Financial Digest. ###### [4].Hanxin fraud case series, 21st Century Business Herald ###### [5].Chief Designer of Loongson: Arming the Loongson Research Group with Leadership, Finance and Economics Digest ###### [6].A piece of the past about the domestic chip and operating system, Liang Ning ###### [7].Loongson branch official public number <span style="font-size:24px;"> </span> <span style="font-size:24px;"> </span>