In the first half of his career, it was the "low period" of domestic chips. In his memories of many chip developers, at that time, the industry was filled with a kind of admiration for foreign chips. "I feel that I can use overseas." If you can use better, use people's, why do you spend your time working on chips?" During this period, high-end chips were controlled by foreign capital, and domestic companies could only kill in the low-end market, not only working more busy, but also lower profits.
But in recent years, Sun Chao has clearly felt that China's chip industry is picking up.
This is a fact. Since the establishment of the 100 billion-level government chip industry support fund in 2014, various preferential policies for chips have emerged in various places, and the number of chip companies has increased. By 2016, the number of chip design companies has doubled compared to 2014, reaching more than 1,300.
Although so far, there have been problems such as shortage of talents, low profit margins, and obvious gaps in many industries in the development of the chip industry over the past decade. However, in the eyes of many chip makers such as Sun Chao, with the funds Enter, the market will increase the recognition of domestic chips, and these problems will have an opportunity to be resolved.
"If there is enough money and effort to develop, I believe that domestic chips can reach the international level in the next decade." He gave this judgment.
"Lost" for 10 years, talents leave
In 2000, China's first modern chip manufacturing company, SMIC, was established. Soon after, Sun Chao graduated from Fudan University and became one of the R&D personnel.
At that time, it was at the peak of the founding of chip companies. Today, SMIC, which ranks among the top in domestic chips,HuaweiMore than ten companies including Haisi, Spreadtrum Communications, Zhongxing Weicheng and Zhuhai Actions were founded between 2000 and 2004.
After more than a decade, when Sun Chao recalled this past, he was most impressed by the scenes of colleagues in the company who worked hard at the time. "The company's overall environment is aggressive, and we feel that we have to catch up with TSMC, even if we can't do it. The world's top chip manufacturing company has to become the backbone of domestic chips, and there is such a belief."
However, the fortunes were not good. In 2003, TSMC sued SMIC in California for infringement of patent rights and theft of trade secrets. The reason was that SMIC founder Zhang Yujing had hired a large number of engineers from TSMC when he founded the company. As a result, SMIC and TSMC began a "competition" that is well known in the chip industry for six years.
The battle ended with the failure of SMIC. In six years, SMIC lost to TSMC twice in the United States. Even after the Beijing High Court counterclaimed TSMC, it still lost the case, accumulating compensation for TSMC's $375 million and 10% of the shares.
After losing the case, SMIC also suffered a performance crisis. In 2009, the company's net profit was the largest annual loss since its establishment, and the net loss reached 960 million US dollars.
"If things happen today, public opinion should give SMIC support. Even if it is not comprehensive, it will not make SMIC fail so badly." Sun Chao said with emotion that in the current environment, domestic chips were not optimistic. , SMIC is helpless.
He said that before the lawsuit, SMIC was about one generation behind the TSMC in the chip process, and it is expected to catch up with them. "Since then, it has been two or three generations behind."
Not only is SMIC's vitality hurt, it is also a good thing for the domestic chip industry. Subsequently, many engineers, including Sun Chao, gradually left the domestic chip industry, and more than two-thirds of them in their class flowed into foreign companies.
Since then, domestic chips have suffered a “lost decade”; in contrast, the mobile Internet has ushered in a “golden decade” of rapid development.
In 2008, with the launch of Apple's App store, the mobile Internet industry ushered in the limelight. China Internet Network Information Center statistics, in the 10 years from 2008 to 2018,China MobileThe number of netizens has increased from 100 million to nearly 800 million. According to CNNIC data, more than 70 mobile Internet companies were listed between 2014 and 2018.
Along with the rise of the mobile Internet, there is also a programmer profession. Eight or nine years ago, the monthly salary of the first-time Internet companies such as Tencent and Ali had exceeded 10,000. In recent years, the annual salary was 200,000 and 300,000.
"The same is the technical staff, the software wages are high, the way out is good; the hardware is not good, not only hard, the wages are still low." Wu Zihao feels like this, he is a key university professor in the field of chips.
According to the introduction of a number of chip practitioners, the graduates are engaged in chip research and development, and the wages are generally only 5 to 70% of the Internet industry. "In the last one or two years, for example, master students have entered a good chip company, probably about 1.2 to 15,000 yuan per month." Research and development engineer Lin Bin said that the growth rate of chip R&D engineers is slower than that of programmers. After the start of the work, the average student has to go through 4 or 5 chip project cycles, and each cycle can be “starting on the side” from 6 months to 2 years. If you want to go to the next level of technology, it will take several years. "The growth rate is slow, the salary increase will be slower, and the programmer can't compare."
In the laboratory of Wu Zihao’s faculty, there have been several incidents of “defection” by students. A well-qualified student, soon after entering the lab, said with the tutor: "This is too hard, I have to go to learn software." Another graduate student experimented with self-learning while experimenting, and directly applied for programmers after graduation. post. “Graduate and doctoral students graduate less than 20% of the profession,” Wu Zihao said. “Most people went to Internet companies.”
As the chip industry is less attractive to talent, the talent gap is gradually emerging. According to the "China IC Industry Talent White Paper (2017-2018)" forecast, before and after 2020, China's IC industry talent demand scale is about 720,000, but the existing talent stock is about 400,000, the talent gap will reach 320,000.
The Red Sea kills the low-end market
"The requirements for chip technology talents are actually very high. They require the deposition of basic disciplines such as physics, mathematics, and chemistry, as well as the accumulation of many industry experience." Sapphire King is a foreign-funded chip company engineer with 7 years of experience. He believes that the moment The mismatch between the talent quality of the chip and the salary is “a situation that is limited by market conditions”.
In the electronics manufacturing industry, chips belong to the upstream industry, and their profits are not high. The chip's top-level gross margin can reach 40% to 60%, but the profit at the low-end level is further compressed.
In the ten years since the domestic chips deviated from the focus of attention, they also ushered in some "window periods." "The chip is essentially tied to electronics." Sapphire remembers that in 2012 and 2013, along withiPadWith the fire on the tablet market, the domestic ushered in a wave of "white card" tablet PCs. In view of the low performance requirements of such products, many chip design manufacturers have entered the market.
The first to seize this "opportunity" is Zhuhai's Quanzhi Technology. In the second half of 2011, Quanzhi Technology, which saw the opportunity, first launched a tablet processor chip A10 based on ARM architecture, which was upgraded to 1G. Simply put, the emergence of this chip, the low-end tablet from the previous "MP4 large version" to a basic tool available.
With the A10 and the subsequent launch of the A13, Quanzhi Technology became a dark horse in the chip field, and its shipments in the low-end market jumped to the top in the country, accounting for about one-third of the total domestic chips, and achieved revenue growth in 2012. Over 400%. At that time, the gross profit margin of Quanzhi Technology reached 56%.
The founding and management team earned a lot of money. An anonymous user who claimed to be a Quanzhi technology employee revealed that most of the early employees of Quanzhi Technology had stocks. After the listing, they went to the stock market high, and many people realized the freedom of wealth.
However, Quanzhi Technology has also opened up the price war of domestic processor chips in the following years.
In the launch of the A13, in order to seize more market share, Quanzhi will position this chip in the 7-inch tablet market with the largest and lowest market, and price it at $5 per piece, which is less than the average market price. half. "In 2011, a tablet chip could still sell for more than a dozen dollars, but the price war was a dozen. In less than two years, the chip has slipped to $3,5," recalls an electronics manufacturer general manager.
Many chip practitioners introduced that although the chip sounds "high-end", in fact, in the low-end chip design field, the technology and capital threshold are not high, and the performance level of the same level chip is similar, and it is impossible to form an absolute competitive barrier. With the increasing number of participants, the cottage with lower technical thresholdMobile phoneThe market for chips and LED chips has finally moved toward price wars.
The price war of Quanzhi Technology finally burned back to itself. With the popularity of A13, a large number of chip manufacturers such as Lingyang, Yangzhi, Junzheng, Ankai, and Ruixinwei joined the market, and the competition quickly entered a fever. According to the prospectus disclosed by Quanzhi Technology when it was applied for listing in 2015, the advantage of 56% high gross profit margin was only maintained for one year. In 2013, the company's revenue reached the highest peak of 1.65 billion yuan in history, but the gross profit margin fell to 41%. . Since then, the revenue and gross profit margin of Quanzhi have not returned to the high point.
Previously, the "Investor News" had disclosed that the chip industry's gross profit margin averaged more than 40%. According to the domestic chip earnings report of Quanzhi, Jingchen Semiconductor and Ruixinwei, the gross profit margin is slightly higher than 30% from 2015 to 2018. In contrast, during the same period, Broadcom, Qualcomm and Intel's gross profit margin ranged from 40% to 60%.
"In the United States, Qualcomm and Intel engineers will not be lower than Google and Amazon," Wu Zihao said. "But in China, because the overall profit rate is relatively low, the concept is not hot, and the chip engineers are really lower than the programmers." too much."
When the domestic mobile phone just emerged, there was a view that this is an opportunity for domestic chips to enter the high-end market. But in the end, the facts show that this field is still the world of Qualcomm, and even the low-end and mid-range machines are mostly occupied by MediaTek.
"After the tablet boom, domestic chips graduallyTVBoxes, smart speakers, car entertainment systems are converted in several directions. "Sapphire King said that the above several types of product use scenarios are relatively simple, but because the market has ushered in the wind, it has become the main shipping direction of domestic chips in recent years.
Industry rejuvenation, getting together design
In 2014, the chip industry, which has been on the edge of wandering for several years, has ushered in spring.
In September of this year, China established the 100 billion-level chip industry support fund, aiming to increase the self-made rate of the chip domestic demand market to 70% within 10 years. In 2030, it will catch up with the world's leading companies in technology. In contrast, according to Morgan Stanley estimates, in the second half of the 1990s, the government invested less than $1 billion in the chip industry.
The chip industry has revived, and the number of chip design companies has grown by leaps and bounds. In the statistics of China Semiconductor Industry Association IC Design Branch, in 2015, there were still 736 domestic chip design companies, which doubled to 1362 in just one year.
Not only is the number of outbreaks, the same period, the chip industry also ushered in a wave of mergers and acquisitions. From July 2013 to June 2015, Ziguang completed two mergers and acquisitions of Spreadtrum Communications and R&D Microelectronics, and acquired 51% of the shares of Hewlett-Packard's Xinhua Group. At the same time, Changjiang Electronics and state-owned enterprises Jianguang Asset Management The company has also launched chip-related acquisitions and acquisitions.
However, the rapid growth of the chip industry has gradually formed a "partial" phenomenon. “A lot of resources have come together,” Lin Bin said. “There are many companies with large market and high input-output ratio. On the contrary, some unpopular areas are still blank.”
From the perspective of the complete industrial chain, the chip goes from upstream to downstream, including equipment, raw materials, R&D, manufacturing, and packaging testing. Among them, R&D and design are the most “light assets”. Overseas Qualcomm, MediaTek, Broadcom, domestic HiSilicon, and Spreadtrum are all “fableless design companies,” that is, they are only responsible for the design process, and manufacturing, packaging and testing are handed over to other companies. Since 2016, the number of chip companies that have doubled their growth has also belonged to the design field.
Chip industry chain schematic all-weather technology mapping
However, since the dispute between SMIC and TSMC, Chinese companies have been slow to advance in chip manufacturing. "The establishment of an advanced process chip production line requires a lot of capital investment," said a chip salesperson. "To maintain production line operations, we must ensure sufficient orders, and we must lead the process. We also need a lot of research and development funds. Many manufacturers simply Can't afford to play."
In the chip boom of the time, various places have invested in the construction of chip factories, but unfortunately, after a short period of time, including Chengdu Gexin, Fujian Jinhua and other projects, there have been "pause" news. Two years later, Gexin announced that the project was shut down.
Today, SMIC is still the largest and most advanced chip manufacturer in the Chinese market. However, if SMIC recently announced that it will mass produce 14nm process in the second half of 2019, it is still more than TSMC. The level of trial production of 5nm process is 2 generations behind.
In other words, the 7nm process required for the two most popular Qualcomm Snapdragon 855 and Huawei HiSili 980 chips in smartphones is still not available.
In addition, domestic enterprises are seriously missing in the two links of equipment and raw materials in the chip industry chain.
Taking the lithography machine called "the most critical field of integrated circuits" as an example, as early as 20 years ago, various kinds of news of China's independent research and development of lithography machines were continuously transmitted, but so far, it has not been able to produce advanced Process lithography machine. “This is one of the easiest links to be 'kicked neck',” Lin Bin said. “Without manufacturing equipment, there is no way to make a chip.”
The eagerness of the lithography machine has even caused a "Oolong event" with a lot of movements. In December 2018, a "Domestic Super Resolution Lithography Equipment Passed Acceptance, Processable 22nm Chips" article detonated the entire market. In this article, this machine can be used up to 10nm process, only one level difference from the top level. In the second generation, a lot of media have written articles on topics such as "Chinese lithography machine finally appeared" and "the chip's most critical areas have finally broken through".
Unfortunately, the news was clarified the following day: This lithography machine developed by the Chinese Academy of Sciences was not used to manufacture chips, but only for other fields. Today, the lithography machine developed and produced by Shanghai Microelectronics Equipment is the most advanced machine in China's domestic market. The highest level of mass production of this lithography machine is 90nm.
The best era?
Despite the existence of various problems, in the eyes of Sun Chao, Lin Bin and others, the present is the "best era" of domestic chips.
"I said that the chip is now on fire," said a family sitting in Zhangjiang, a gathering place in China's chip industry.coffeeInside the museum, Lin Bin smiled. "In the past, there was very little media to pay attention to our industry." He said that in the past one or two years, there have been many more headhunting calls. "Chip talents are more valuable."
Good news from the chip industry has swarmed. In the past two years, Beijing, Zhejiang, Henan, Guangdong, Shanghai, Shenzhen and other places have introduced implementation opinions and plans to promote the development of integrated circuits. Among them, Zhejiang Province clearly stated that by 2020, the province's integrated circuit and related industry business will exceed 100 billion yuan.
Since 2017, the tax reduction policy for the chip industry has been issued once a year. The most recent time was on May 22 this year, the Ministry of Finance and the State Administration of Taxation issued the latest corporate income tax policy announcement on integrated circuit design and software industry, which wrote: qualified IC design companies and software companies, since December 2018 Before the 31st, the preferential period will be calculated from the profit-making year. The enterprise income tax will be exempted from the first year to the second year, and the enterprise income tax will be halved at the statutory tax rate of 25% from the third year to the fifth year.
In addition to the external environment, according to the introduction of a number of chip practitioners, the chip itself is also enjoying a new life cycle.
"In the PC era, Intel occupied the mainstream position of the processor chip; in the era of smart phones, Qualcomm became the main force," Lin Bin said. "At the beginning of each era, when the chip company reshuffled, the industry is now ushered in. In the era of AI chips, no company has an absolute advantage in this product."
Although the "partial science" of domestic chip development has caused many gaps in the industry chain, on the other hand, it has also made domestic chips advance rapidly in the design field. Most practitioners believe that Huawei HiSili has already competed with Qualcomm on the mobile phone Soc chip; in the latest AI chip field, Haisi, Cambrian and other enterprises also have a highly competitive design level.
"NVIDIA has a first-mover advantage on the AI chip, but it has not yet formed an absolute moat. It is now an alternating period. For domestic chips, this is an opportunity." Lin Bin said.
The AI chip has formed a new vent, and in 2017 alone, dozens of startups including Yunzhisheng, Cambrian, and Horizon have announced their entry into the AI chip field. According to statistics, in 2016, the global AI chip scale was 2.388 billion yuan, and by 2020 it will grow to 14.616 billion yuan. “A lot of money is coming in, and there are indeed some scams,” Lin Bin shrugged. “However, there are some companies with very good standards.”
In the global AI chip company rankings given by research firm Compass Intellingence, Chinese companies occupy 7 of the 24 companies, of which Huawei ranks 12, ranking first among Chinese manufacturers.
Outside the design field, in the fields of manufacturing and equipment where the gap is more obvious, in the view of Sapphire King, although there may be a state of being “card necked” at present, it is not without hope. “Once it really appears In extreme cases, if there are big companies and big forces to do this, they will also have the strength to do it."
In the question about "Do you think that domestic chips can have a chance to achieve first-class levels?", each engineer's answer is "yes"; in the years they give, the shortest is considered to be 3 years, and the longest is It is 10 years.
(Wenzhong Sun Chao, Lin Bin, Wu Zihao, and Lan Baowang are all pseudonyms. All-weather technology writer Yang Fan also contributes to this article)
Author | Yao Xinwei Editor | Luo Lijuan