China’s economic upgrade is primarily focused on two key sectors: the automotive manufacturing industry and the integrated circuit industry. Conversely, to ultimately defeat Western powers, we must achieve a turnaround in these two super industries.
In the coming decade, we must prepare for a long-term battle; if we succeed, China will become a true superpower. If we fail in these two areas, the rejuvenation of the Chinese nation cannot be deemed successful.
What are the billion-dollar profit companies in the most powerful country on Earth, the United States?
Excluding energy and finance, there are about 17 such companies.
Facebook, General Electric, IBM, Cisco, Intel, Procter & Gamble, Johnson & Johnson, Microsoft, Apple, Walmart, Disney, General Motors, Oracle, Alphabet (Google’s parent company), Verizon, AT&T, Comcast
To be more specific, there is 1 in retail, 2 in daily chemical industry, 1 in entertainment, 1 in automotive, and 1 in other manufacturing (General Electric). A total of 6.
In the ICT sector, there are 11! Out of 17 billion-dollar profit companies, 11 are in the ICT sector, thus it can be said that the ICT industry has become the cornerstone of the United States.
Google, Amazon, Facebook, Cisco, IBM, Intel, Microsoft, Oracle, Verizon, and AT&T contribute enormous profits to the U.S.
What supports America’s strength? These great companies. Conversely, without these companies, the powerful United States would not exist. In 1945, Japan and Germany still had tens of millions of people; the Japanese army lost only over two million on the battlefield. Why were they completely unable to resist? Because their domestic industrial companies were destroyed, industrial production could not continue, and thus their powerful force ceased to exist.
We can see that in the ICT sector, which is the foundation of the U.S., China is clearly the biggest challenger. Companies like China Mobile, Huawei, ZTE, Alibaba, Baidu, Tencent, NetEase, Xiaomi, Inspur, and Unisoc not only gained advantages in the domestic market but also expanded abroad. Most of these companies have net profits exceeding one billion USD; globally, there are only about three or four hundred companies with net profits exceeding one billion USD.
This is also why the U.S. is so concerned about China’s integrated circuit industry.
Integrated circuits are the upstream technology of the ICT sector and the cornerstone of hardware. If China’s integrated circuit industry rises, the biggest impact will be on the U.S., as this is one of America’s core sectors.
Look at the following: the three giants of the global integrated circuit industry: Intel, Samsung, Qualcomm. Just Qualcomm and Intel alone can bring in eight to nine hundred billion USD in revenue for the U.S. each year, supporting tens of thousands of American engineers and generating over a hundred billion USD in net profits.
In addition to Intel and Qualcomm, the U.S. also has Texas Instruments, Nvidia, Broadcom, Micron, and many other semiconductor companies. The importance of this industry to the U.S. is self-evident.
Currently, China imports over 500 billion USD worth of industrial goods annually in only two categories: one is automobiles and parts, with imports totaling 74.61 billion USD,
and the other category exceeding 500 billion USD is the well-known integrated circuits.
China’s imports of industrial goods lead all others, with integrated circuits ranking first, with an import value of 227.1 billion USD, three times that of the second category, automobiles and parts, at 74.6 billion USD.
Of course, China once had another industrial product that exceeded 500 billion USD in imports, which was LCD panels. Just five years ago, the imports of LCD panels were about 51 billion USD.
However, with the rise of domestic giants like BOE, the imports of LCD panels have been declining in the context of growing domestic market share. By 2016, imports had dropped to 31.85 billion USD. As all domestic panel manufacturers are frantically increasing their market share and profits this year, we have reason to believe that in the coming years, imports of LCD panels will continue to decline. The battleground has shifted to the OLED panel sector, where the only competitors are from South Korea.
Returning to integrated circuits,
let’s look at the following set of numbers: how has our import substitution progressed in the past seven years? The answer is regrettable.
Unlike other industries where our manufacturing capabilities have gradually increased and trade deficits have decreased, the trade deficit in integrated circuits has been continuously rising over the past seven years. From 127.74 billion USD in 2010, it has risen to 165.7 billion USD in 2016.
In 2010, integrated circuit imports were 156.99 billion USD, exports 29.25 billion USD, with a deficit of 127.74 billion USD.
In 2011, integrated circuit imports were 170.2 billion USD, exports 32.57 billion USD, with a deficit of 137.63 billion USD.
In 2012, integrated circuit imports were 192.06 billion USD, exports 53.43 billion USD, with a deficit of 138.63 billion USD.
In 2013, integrated circuit imports were 231.34 billion USD, exports 87.7 billion USD, with a deficit of 143.64 billion USD.
In 2014, integrated circuit imports were 217.62 billion USD, exports 60.86 billion USD, with a deficit of 156.76 billion USD.
In 2015, integrated circuit imports were 230.7 billion USD, exports 69.31 billion USD, with a deficit of 161.39 billion USD.
In 2016, integrated circuit imports were 227.1 billion USD, exports 61.38 billion USD, with a deficit of 165.72 billion USD.
By 2017, this trend had not reversed, with imports of integrated circuits (chips) from January to May amounting to 95.48 billion USD, a year-on-year increase of 17.9%, while exports were 25.66 billion USD, a year-on-year increase of 11.3%, and the deficit continued to widen.
Today, we will discuss the core of this article,which is to recognize both the gap and the progress.
First, we need to know how large the global market capacity for integrated circuits is? Different organizations have different data, with slight discrepancies; we will take IHS data as an example. In 2016, the global semiconductor market size reached 338.93 billion USD, with a slight year-on-year increase of 1.1%.
Of course, the global market size data varies among organizations, with some reporting over 330 billion USD and others over 340 billion USD, but it is basically within this range.
Everyone should compare this number with China’s annual integrated circuit imports of 227.1 billion USD; our annual import amount accounts for 67% of the global market. This is not surprising, as 90% of laptops, 90% of smartphones, and a large number of other electronic devices are manufactured in China. The title of “world’s factory” is not just a boast; it is a reality.
Therefore, until China’s integrated circuit industry reaches a global market share of 70%, we will continue to import integrated circuits.
China imports a lot of integrated circuits, and this trend is continuously growing.
On one hand, this is because China’s integrated circuit industry lags significantly behind the U.S., Europe, South Korea, and Japan.
On the other hand, it is due to the rapid development of our downstream manufacturing industry and domestic brands, which are capturing a growing share of the global market. For instance, in 2017, Chinese brands of smartphones accounted for about 50% of the global market share, while Apple and Samsung smartphones are mostly manufactured in China.
Based on this reality:
China is not only the world’s manufacturing center but also shows a trend of concentration in the share of downstream consumer electronics brands towards Chinese brands. Therefore, for a considerable period, we will maintain a high import trend for integrated circuits.
Thus, in the report “Made in China 2025” released by the State Council in 2015, it was stated that by 2020, China’s chip self-sufficiency rate should reach 40%, and by 2025, it should reach 50%. This is indeed a very high goal because it means that by 2025, China’s integrated circuit industry will account for 35% of the global market, surpassing the U.S. and ranking first in the world. Of course, this refers to overall output value; in terms of industrial structure, the U.S. remains in the high-end, while China is in the mid-to-low end and some high-end segments.
If we follow the more aggressive planning of the Ministry of Industry and Information Technology, by 2025, the goal is to achieve 70% chip self-sufficiency, meaning China’s integrated circuit industry will account for 49% of the global market. What does this mean? By 2025, China will become the world’s largest integrated circuit industry in terms of output value, capable of meeting domestic needs while also capturing a significant portion of the global market.
However, we must also note that whether it is the State Council or the Ministry of Industry and Information Technology’s planning, foreign-funded enterprises in China are counted as domestic production. In other words, even if the 70% self-sufficiency target is achieved by 2025, and we account for 49% of the world, a portion of this will still be completed by foreign-funded enterprises in China.
Can we achieve this goal? Or rather, what level are we currently at? The following chart from IC Insights shows that in 2016, none of the top 20 global semiconductor companies were Chinese, which is a very rare phenomenon. In almost any other large industry on Earth, it is difficult to find a situation where there are no Chinese companies among the top 20.
There are 9 companies with revenues exceeding 10 billion USD, and the threshold for the top 20 is 4.455 billion USD.
However, we should not be disheartened. The largest semiconductor company in China is Huawei’s HiSilicon, which had sales of 30.3 billion RMB in 2016, equivalent to about 4.45 billion USD, which is almost the same as the 20th place threshold of 4.455 billion USD. Therefore, it is expected that this year, HiSilicon will become the first Chinese semiconductor company to enter the global top 20. This will be a historic breakthrough. Of course, this means that another unfortunate company will be pushed out of the global top 20; such is the cruelty of life.
In 2016, the sales revenue of China’s integrated circuit industry reached 433.55 billion RMB, a year-on-year increase of 20.1%,
with the sales revenues of the three segments—design, manufacturing, and packaging—being 164.43 billion, 112.69 billion, and 156.43 billion respectively, with growth rates of 24.1%, 25.1%, and 13%.
There is both good news and bad news here. The good news is that the growth rate of 20.1% is very fast; it doubles in four years and quadruples in eight years, which is quite astonishing.
The bad news is that this 433.55 billion RMB sales figure includes foreign-funded enterprises in China. Therefore, to understand the true strength of China’s integrated circuit industry, we need to focus on domestically owned companies.
We will discuss the design, manufacturing, and packaging segments in order. This article will focus first on the design segment.
First, in chip design, it is well known that during the PC era, Intel and AMD dominated the CPU market, and the gap between Chinese companies and them can be likened to a galaxy apart; we could only look up at extraterrestrial civilizations from Earth.
In the early 21st century, the only notable chip manufacturer in China was Zhuhai Jieli Technology, which produced MP3 chips and became one of the world’s largest MP3 chip suppliers. Let’s take a look at the original news from that year:
“In 2005, Jieli’s SoC chip shipments reached 49.8 million units, a 308% increase from 12.2 million units in 2004. The sales of the company’s MP3 player SoC chips also grew from 54.1 million USD in 2004 to 147.9 million USD.”
For a Chinese chip company to achieve this level over a decade ago was quite remarkable; however, Jieli’s rise was also predicated on the fact that domestic MP3 brands occupied a large market share. At that time, MP3 players were very expensive. Not to mention Apple and Samsung’s MP3 players, I remember that the MP3 player I bought from Beijing Patriot in 2004 cost 900 RMB, which required me to save for a long time.
While unable to compete in the consumer electronics field, China’s chip industry unexpectedly benefited from the rise of the telecommunications industry. Due to the emergence of giants like China Unicom, Datang, and Huawei in the telecommunications industry, China gradually began to achieve autonomy in telecommunications equipment manufacturing.
The rise of the telecommunications equipment manufacturing industry unexpectedly gave birth to China’s chip industry alongside the rise of the telecommunications sector. Huawei’s HiSilicon, ZTE Microelectronics, and Datang have become some of the largest chip manufacturers in China.
As we entered the era of smartphones and tablets, the emergence of domestic smartphone and tablet manufacturers provided development opportunities for companies like Spreadtrum, Allwinner, and Rockchip.
The following chart from IC Insights shows the top 50 pure chip design companies globally. In 2009, China had only one, which was Huawei’s HiSilicon, and in 2016, this number grew to 11, including HiSilicon, Spreadtrum, ZTE, Datang, NARI, Huada, RDA Microelectronics, ISSI, Rockchip, Allwinner, and Montage.
It can be seen that the top three, HiSilicon, ZTE, and Datang, all originated from the telecommunications industry, while NARI provides chips for smart grids, and Allwinner, Rockchip, and Spreadtrum provide chips for smartphones and tablets.
This shows that the development of China’s telecommunications, power grid, and consumer electronics industries has a driving effect on the upstream. Thus, the development of a country’s industry must be holistic; if the downstream does not develop, the upstream cannot develop either. The Chinese chip industry cannot develop independently of the overall industrial upgrading environment in China.
By 2016, China had 160 chip design companies with sales exceeding 100 million RMB, indicating that chip design is flourishing across China. The two largest chip design companies, HiSilicon and Ziguang Zhanrui, have already entered the global top ten, with HiSilicon ranked sixth in 2015 and Ziguang Zhanrui ranked tenth. Moreover, HiSilicon and Ziguang Zhanrui have the fastest growth rates among the top ten.
In the first quarter of 2017, the overall growth of China’s chip design industry reached 23.8%, with sales reaching 35.16 billion RMB, and the global market share of domestically designed chips reached 8%, with the Chinese market share exceeding 13%.
According to the rankings published by the China Semiconductor Industry Association, there are differences in the top ten Chinese chip companies compared to IC Insights’ rankings. The following chart shows the rankings of Chinese IC design companies published by the China Semiconductor Industry Association in 2016. It can be seen that through the integration of Ziguang Group with Spreadtrum and RDA, China has already produced two local IC design giants with revenues exceeding 10 billion RMB: HiSilicon and Ziguang Zhanrui.
Looking at this list, I believe everyone is familiar with the top three:
The first is HiSilicon.
Many of you have Huawei smartphones that contain numerous HiSilicon processors and baseband chips. Additionally, the smart TVs and security systems you buy also have HiSilicon chips. HiSilicon will remain China’s largest chip design company for a long time, and its growth will be in line with that of the Huawei Group.
Globally, there are only seven manufacturers that produce smartphone processors: Apple, Samsung, Huawei, Xiaomi, Qualcomm, MediaTek, and Spreadtrum.
HiSilicon is currently the most likely Chinese company to reach the top of the world, aiming to challenge the positions of Samsung, Intel, and Qualcomm. Of course, the gap is still significant. The leading pure IC design company, Qualcomm, had a revenue of 15.4 billion USD in 2016, which is 3.5 times that of HiSilicon.
However, from a long-term perspective, I believe it is not a big issue for HiSilicon to surpass Qualcomm, as companies like Apple, Samsung, and Huawei are all moving towards chip self-sufficiency, while Qualcomm’s market space is continuously shrinking.
The second is Ziguang Zhanrui, which was established after the merger of Spreadtrum and RDA, and is currently the largest supplier of smartphone processors and baseband chips for Samsung, apart from its own products. Ziguang’s task is to outperform MediaTek. Ziguang is the only domestic company besides HiSilicon with revenues exceeding 10 billion RMB.
Please note that the Samsung phones you buy, especially the mid-to-low-end series, primarily use Ziguang Zhanrui’s chips. We are still a chip-exporting country, and Ziguang Zhanrui’s chips are sold alongside Samsung phones worldwide. Isn’t that a bit exciting?
Additionally, Ziguang Zhanrui is the largest smartphone chip manufacturer in India, as mid-to-low-end phones widely use Ziguang’s chips, and the same situation applies in Africa.
The third is ZTE Microelectronics, which mainly produces chips for its own telecommunications equipment, while smartphone chips are still purchased externally.
The fourth is HwaCom, which is a bit interesting as it is rarely heard of by the public but is actually the fourth-largest chip design company in China.
HwaCom is the national team of China’s IC design and is part of CEC (China Electronics Corporation), which is a Fortune 500 company and the national team of China’s electronics industry. It is the only enterprise in mainland China that covers the entire integrated circuit industry chain, including design, manufacturing, packaging, testing, and EDA tools.
Of course, it should be noted that while CEC appears to be involved in everything, it does not excel in any of them, and its scale is relatively small. However, CEC’s existence is to ensure the country’s integrated circuit capabilities from a national level.
What products does HwaCom offer? As a national team, it is responsible for everything related to national security, including the second-generation resident ID cards, Sinopec fuel cards, social security cards, and tax control IC cards, all of which contain chips designed by this company. Additionally, there is a satellite navigation chip, including the BeiDou system. There are rumors online that China’s second-generation ID card technology comes from Japan, but just laugh it off; the chips in the second-generation ID cards are all designed, manufactured, and packaged independently by China, led by the national team.
Another very noteworthy point is that HwaCom has the world’s fourth-largest EDA design software provider—HwaCom Jiutian. EDA tools are essential software for integrated circuit design. Currently, the mainstream in China is still using foreign EDA tools for chip design, such as Protel, but HwaCom Jiutian’s EDA also holds a place, which preserves China’s independent capabilities.
The fifth is Beijing Zhixin Microelectronics, which is relatively unknown among the public. This company is under the State Grid, and it provides various chips for the smart grid, including those for electric vehicles and charging piles.
The sixth is Goodix Technology, which I mentioned in a previous article. This company has become the second-largest in the world in fingerprint recognition chips.
It is quite an achievement for a Chinese chip company to rank second in a specialized field. Many of us have likely used the fingerprint recognition modules on our phones.
In 2016, Goodix’s revenue was 3.079 billion RMB, a year-on-year increase of 175.04%, and its net profit attributable to shareholders was 856 million RMB, a year-on-year increase of 126.46%. The revenue from its fingerprint recognition chips was 2.312 billion RMB, a year-on-year increase of 788.66%, accounting for 75% of its total revenue. In the first quarter of 2017, Goodix completely surpassed its old rival, Sweden’s FPC, and is now only second to AuthenTec, which supplies fingerprint recognition chips to Apple.
The seventh is Hangzhou Silan Microelectronics. This company’s LED lighting driver IC is one of its main sources of income. Additionally, Silan is one of the few IDM companies in China, meaning it handles design, production, and packaging.
In addition to its main business of LED lighting driver ICs, Silan also provides variable frequency motor control chips to home appliance manufacturers. Other products include MEMS sensors and IGBT products, which have also been successfully developed and brought to market.
The eighth is Datang Microelectronics, which is a national team. Its main product direction is chips and solutions for identity cards and financial social security cards. Datang’s market share in financial social security cards is second, and its financial IC security chip shipment volume is about 200 million units. In addition, Datang also provides industry certification cards, such as health cards, transportation cards, and citizen cards, as well as security chips and solutions.
Additionally, Datang has established the first automotive semiconductor company in China, Datang NXP, in collaboration with NXP from the Netherlands to develop power management chips for new energy vehicles, motor MCUs, etc.
In short, Datang is a state-owned enterprise, primarily engaged in government business.
The ninth is Duntai, a design company from Taiwan, which I will not elaborate on.
The tenth is Zhongxing Microelectronics, which mainly produces image processing chips and camera chips.
After reviewing the top ten IC design companies in the country, what are your thoughts?
Among these ten companies, only five are truly promising: HiSilicon, Ziguang (including Spreadtrum and RDA), Goodix, ZTE, and Silan.
For example, Hangzhou Silan’s financial report for the first half of 2017 showed that the company achieved revenue of 1.298 billion RMB, a year-on-year increase of 22.90%, and net profit of 84.425 million RMB, a year-on-year increase of 243.77%. It has maintained a high growth momentum, and profits are gradually recovering.
The other five: HwaCom, Datang, Zhixin are primarily supported by the state. Duntai is a Taiwanese enterprise, and Zhongxing Microelectronics mainly produces surveillance and security camera image chips with average technical levels and unclear development prospects.
Beyond the top ten, there are several interesting Chinese IC design companies.
First, in the memory chip sector, approximately 75% of the integrated circuits imported by China each year are memory chips and CPUs, demonstrating the importance of memory chips.
In the memory chip sector, there are two Chinese companies trying to break into the market: Zhaoyi Innovation and Yangtze Memory.
Currently, there are three main types of memory chips globally, ranked by sales volume: DRAM, NAND Flash, and Nor Flash. Although Nor Flash ranks third, its market size is relatively small, only about 3 billion USD, while the other two are approximately at the 40 billion USD and 30 billion USD levels.
What are the differences among these three? DRAM refers to the memory in our phones, such as 1G, 2G, etc., while NAND Flash refers to the storage capacity in our phones, such as 32G, 64G, 128G, etc.
Nor Flash, although also a type of memory, has a smaller capacity, generally below 64Mb, and is used to store algorithms and codes for driver circuits. It is used in mobile phones, automotive electronics, and industrial control, as many circuit algorithms and codes are relatively complex and cannot all be integrated into the IC, requiring Nor Flash for storage.
In this small market of only 3-4 billion USD for Nor Flash, China’s Zhaoyi Innovation is one of the major players, although it is ranked fifth globally, with the first to fourth being Taiwan’s Winbond, U.S. Cypress, U.S. Micron, and Taiwan’s Nuvoton.
Zhaoyi Innovation became the only A-share listed company in the semiconductor memory industry after its IPO in August 2016.
In 2015 and 2016, Zhaoyi Innovation achieved revenues of 1.189 billion RMB and 1.489 billion RMB, with net profits of 142 million RMB and 151 million RMB, respectively.
As of this year, due to multiple price increases in Nor Flash, Zhaoyi Innovation’s quarterly performance has surged. The company’s first-quarter revenue was 452 million RMB, a year-on-year increase of 46.61%, and net profit was 69 million RMB, a year-on-year increase of 94.20%.
The reason for the price increase is mainly that the industry giants are gradually exiting the market. The third-largest, Micron, is exiting the NOR Flash business, which could be taken over by the fourth-largest, Taiwan’s Nuvoton. The second-largest, Cypress, is closing small to medium-sized NOR Flash production lines. Additionally, the demand for NOR Flash from AMOLED panels is driving the industry’s prosperity.
Moreover, Zhaoyi Innovation also produces MCU (microcontroller unit) products for automotive and IoT applications. In 2016, Zhaoyi Innovation’s MCU sales revenue was 197 million RMB, a year-on-year increase of 55.2%.
In the memory and flash storage sectors, South Korea holds an absolute advantage, while Yangtze Memory is tasked with breaking this monopoly.
In the memory sector, DRAM and NAND Flash are dominated by South Korea’s Samsung and Hynix.
In 2017, South Korea made significant profits from the rising prices of memory, with Samsung Electronics’ net profit reaching 6.78 billion USD in the first quarter of 2017, a staggering growth of 46%, close to 50%. By the second quarter of 2017, Samsung’s revenue surged by 19.8%, with net profit increasing by 89% to 9.9 billion USD, breaking its own record for the highest quarterly net profit and surpassing Apple for the first time.
SK Hynix’s revenue reached 38.4 billion RMB in the first quarter of 2017, a year-on-year increase of 72%, with net profit of 11.6 billion RMB, a year-on-year increase of 324%.
The flash memory incident involving Huawei in the first half of the year reflects China’s absolute weakness in this area.
DRAM is the largest memory sector, and currently, global DRAM prices are skyrocketing, with South Korean companies making significant profits.
In the second quarter of 2017, the combined sales of DRAM by the three industry giants, Samsung, SK Hynix, and Micron, reached 4.43 billion USD, a 30.1% increase from the previous quarter.
Among them, Samsung’s sales reached 1.98 billion USD, a 36.5% increase from the previous quarter, while SK Hynix’s sales reached 1.37 billion USD, a 28.2% increase. The combined market share of the two companies accounted for 75.9% of the global market, with Micron’s sales at 1.06 billion USD, a 22.0% increase.
The following chart shows the global market share in the first quarter of 2017, with Samsung + SK Hynix accounting for 73.5%, and Micron accounting for 21%. The three giants combined hold 94.5%, while the three Taiwanese companies, Nanya, Winbond, and Powerchip, account for 4.6%. Other global companies account for 0.9%.
Outside the top six, there is a small Chinese company, ISSI, which is a design company under Beijing Silicon Holdings, ranking eighth in the global DRAM market. However, this eighth place is almost negligible.
Zhaoyi Innovation once attempted to acquire ISSI to enter the DRAM sector, as the NOR Flash market is too small. However, according to news from August 2017, the acquisition was unsuccessful for two reasons: first, the price could not be agreed upon, and second, the Taiwanese were obstructing it. The fourth-largest DRAM manufacturer, Nanya, is the foundry for ISSI. If ISSI were acquired by Zhaoyi Innovation, Nanya would stop providing services.
However, in the DRAM sector, ISSI is too small. Even if Zhaoyi Innovation successfully acquired ISSI, it would not be able to compete with the Korean manufacturers and Micron.
In the NAND Flash market, data from DRAMeXchange shows that Samsung, Toshiba, SanDisk, Hynix, Micron, and Intel virtually monopolize the global market. Especially Samsung and Hynix together account for nearly half of the market share.
It is clear that China has almost no presence in the DRAM and NAND Flash sectors.
To break through in this field, we still need to rely on the national team. Previously, who said state-owned enterprises are useless and only know how to engage in monopolies? If we could establish a monopoly in the memory sector, that would be very beneficial.
In July 2015, China’s Unisoc Group made a 23 billion USD acquisition offer to U.S. Micron Technology, the third-largest DRAM manufacturer globally. The offer was rejected due to the obstruction of the U.S. government.
Why was the acquisition of Micron pursued? Because Korea is too strong; whether in the DRAM or NAND Flash sector, Micron is in a weak position against the Korean giants, especially in the DRAM sector, where Micron’s market share has been declining due to competition from Korea, dropping from 28% in 2016 to about 18% in the first half of 2017.
In terms of operating profit margins, Samsung’s DRAM operating profit margin is as high as 40%, while Micron’s is only 20%. As long as Samsung lowers its prices, Micron’s survival is at stake. It can be said that Samsung completely controls Micron’s fate.
In fact, the collaboration between Unisoc and Micron for research and development would create a win-win situation, with one side providing funding and the other providing technology. However, the U.S. government’s strategy to suppress China’s technological progress makes such cooperation extremely difficult. Although Unisoc has not given up on cooperating with Micron, it has been forced to pursue independent research and development.
Currently, the national team for independent research and development in the memory sector is Yangtze Memory, which is controlled by Unisoc Group.
On July 26, 2016, Yangtze Memory was established, with Unisoc Group holding 51% of the shares, while 25% is held by the National Integrated Circuit Industry Investment Fund, and 24% by local government funds. Wuhan Xinxin Company is a wholly-owned subsidiary of Yangtze Memory.
Wuhan Xinxin Company was established in 2006 by the Hubei provincial and Wuhan municipal governments to enter the integrated circuit manufacturing field as a research and development and manufacturing integrated company. In fact, Wuhan Xinxin began collaborating with U.S. Cypress Semiconductor in 2014 to jointly develop NAND Flash technology, but the industry is not optimistic about this, as NAND Flash is dominated by Samsung, Hynix, Toshiba, and Micron. It is not only that Wuhan Xinxin lacks research and development strength, but Cypress also does not have enough strength.
Therefore, through collaboration with Cypress, Wuhan Xinxin has gained some independent research and development capabilities, but they are very weak, representing the initial spark for China’s independent research and development of memory.
After Yangtze Memory was established, due to a large influx of funds, it began to heavily recruit talent from Taiwan and South Korea. If you pay attention to recent technology news from Taiwan and South Korea, you will find that their media often complain about semiconductor talent being lured away by high salaries from China.
How many heavyweight talents has Unisoc recruited from Taiwan?
Unisoc Group’s global executive vice president, Gao Qiquan, was personally recruited by Unisoc Group’s chairman Zhao Weiguo from Taiwan in October 2015. Gao Qiquan is known as the “DRAM godfather” of Taiwan. He is the founder of Winbond, the world’s largest NOR Flash company mentioned earlier, and the general manager of Nanya Technology, Taiwan’s first and the world’s fourth-largest DRAM company. He is also the chairman of the joint venture with Infineon, Huaya Technology (which has now been acquired by U.S. Micron). He is a heavyweight figure in Taiwan’s semiconductor industry.
Many Taiwanese media compare Gao Qiquan to Liang Mengsong, who jumped from TSMC to Samsung, where Samsung’s foundry business has made significant progress, posing a major threat to TSMC and even surpassing TSMC in process technology, leading to the priority production of 14nm chips and even taking a share of TSMC’s Apple chip orders.
After announcing his retirement in 2015, Gao Qiquan joined Unisoc as global executive vice president. Taiwanese media view him as a traitor, similar to Liang Mengsong.
The fact is, it is unknown whether Gao Qiquan has any great Chinese sentiments, but based on his statements during interviews, he seems to express dissatisfaction with the limited development space in Taiwan, which makes it impossible to compete against his lifelong enemy, Samsung.
In an interview with the media, Gao Qiquan stated:
“In 1985, Intel officially exited the DRAM sector. After Japan and Korea invested heavily, the latecomers surpassed them, and Korea ultimately secured the top position. Taiwan established the Taiwan Innovative Memory (TMC/TIMC) program in 2008 to concentrate resources on developing the DRAM industry, but ultimately failed due to funding issues. The original funding requirement of 50 billion NT dollars was only met with 8 billion from the Taiwanese government, far short of expectations.”
He emphasized, “Now, the only opportunity to become a global counterbalance to Samsung is mainland China. Unisoc will turn this dream into concrete practice!” There must be someone to play the role of balancing Samsung’s power globally, and currently, only mainland China has the capability and financial resources to invest, which is the reason for the establishment of Yangtze Memory and the only truly invested memory enterprise by the National Integrated Circuit Industry Investment Fund.”
From Gao Qiquan’s words, we can see his inner thoughts, which is to make a mark in the industry. Therefore, as long as we provide a platform that is large and good enough, regardless of his political stance, he can still be of use to us. The stage in Taiwan is too small.
In addition to Gao Qiquan, Unisoc Group has also recruited Yang Weiyi, the founder of Morningstar, to serve as CTO of Yangtze Memory, and has hired Sun Shih-wei, the former CEO of UMC, as global executive vice president. Both of these individuals are well-known figures in Taiwan’s semiconductor industry. Morningstar was once one of the two largest chip research and development companies in Taiwan, alongside MediaTek, but was later swallowed by MediaTek.
UMC is the second-largest integrated circuit manufacturer in Taiwan, after TSMC, with sales of 4.45 billion USD in 2016. In contrast, SMIC’s sales revenue in 2016 was only a little over 2.9 billion USD.
Currently, Yangtze Memory is conducting research and development for both NAND Flash and DRAM. According to Gao Qiquan’s interview in April 2017, “Yangtze Memory’s R&D team includes Americans, Japanese, and South Koreans, not just Taiwanese engineers. The internal communication is required to be in English due to the presence of many foreign executives, and we are building the company with international standards.”
As of now, Wuhan Xinxin has about 1,200 employees, and Yangtze Memory has recruited nearly 700 people since its establishment, totaling around 1,900 people.
Currently, Yangtze Memory has about 500 R&D personnel, including about 50 from Taiwan.
According to news from September 2017, Yangtze Memory plans to establish an R&D center in the U.S. and take advantage of the opportunity presented by Toshiba’s sale to recruit Japanese memory R&D talent. Many young engineers in Japan are dissatisfied with the hierarchical culture in Japanese workplaces, which limits their salaries and career advancement opportunities. Given that Toshiba is in turmoil, many will likely join under the high salary temptations from China.
Currently, Yangtze Memory is focusing on the development of 3D NAND Flash while also advancing the development of 20/18nm DRAM.
NAND Flash will undoubtedly be the first product to be mass-produced by Yangtze Memory. After the transition from traditional 2D to 3D NAND technology, almost all semiconductor equipment will need to be replaced, making this investment timely. Every memory camp starts from the same point.
On the other hand, the DRAM technology requires only a 20% increase in semiconductor equipment with each new generation of process technology. This means that most of the equipment from established semiconductor giants has already depreciated, and new entrants in DRAM technology will face high costs and poor competitiveness. From this perspective, China must prepare for a long-term investment in the memory sector.
How is Yangtze Memory’s R&D progress currently? According to Yangtze Memory CEO Yang Shining, their 32-layer 3D NAND chip has successfully achieved complete technical verification of process devices and circuit designs, and its electrical characteristics and other indicators have met expectations. Sample chips will be provided by the end of 2017, with plans to continue developing to 64 layers 3D NAND. Optimistically, mass production is expected in 2019.
Currently, Samsung has already achieved mass production of 64-layer NAND in 2017, indicating that the technology gap between China and South Korea is two years. Two years may seem short, but in the fiercely competitive and rapidly changing market, it represents a significant gap. Countering South Korea will be a long-term process.
However, we should not view South Korean semiconductors as invincible. In fact, the strong performance of South Korea in the semiconductor field is mainly in NAND FLASH and DRAM memory sectors, with companies led by Samsung having started research and development in memory as early as the 1990s, twenty years ahead of us. To surpass them, we will inevitably have to pay the price of time. It won’t be three to five years; achieving this in ten years would be quite remarkable.
Additionally, while South Korea is strong in memory, in other areas outside of memory, apart from Samsung’s self-produced processors and baseband chips for smartphones, South Korea’s chip design industry has faced severe decline under the impact of China.
For example, as of the first half of 2017, according to the South Korean media Business Korea, among the top 15 IC design companies listed in Korea, 10 experienced a decline in operating profits, and 50% of the companies reported losses, with only two companies showing growth in revenue and profit.
The reason is the impact of China on sectors like driver chips for liquid crystal panels, touch chips for smartphones and tablets, and mid-to-low-end image sensors.
In my previous article, I wrote about the irreversible trend of global display panels shifting to OLED. A Chinese company producing OLED driver chips is experiencing rapid growth: Zhongying Electronics. In the first half of 2017, Zhongying Electronics achieved revenue of 312 million RMB, a year-on-year increase of 30.65%, and net profit of 62.08 million RMB, a year-on-year increase of 61.82%. Earnings per share were 0.30 RMB.
Therefore, currently, only China is making massive investments in memory research and development and manufacturing, and the only possible disruptor in the future for the South Korean memory industry will be China, specifically Yangtze Memory.
Future Hope: Cambricon
Why mention Cambricon specifically? Artificial intelligence is the trend of development in all industries, and in this field, specialized chips will become the mainstream in the future. Almost all ICT technology companies, whether domestic like Baidu, Alibaba, Tencent, Huawei, or American companies like Google, Facebook, Amazon, and Microsoft, are moving towards artificial intelligence.
In August 2017, Cambricon Technologies, founded by Chen Yunji and Chen Tianshi from the Institute of Computing Technology of the Chinese Academy of Sciences, announced the completion of a 100 million USD Series A financing led by Alibaba, valuing Cambricon at 1 billion USD. This also marks China’s first unicorn company in the integrated circuit industry (valuation exceeding 1 billion USD), which can be considered a milestone.
Notably, the average age of Cambricon’s team members is only 25, with many key members having started working in related fields during their time in school. Founders Chen Yunji and Chen Tianshi are brothers, both of whom are academic high achievers.
Chen Yunji entered the youth class of the University of Science and Technology of China at the age of 14. Before founding Cambricon, in his final year of university, he participated in the research project for China’s first general-purpose CPU chip, Longxin No.1. In 2002, he joined the Institute of Computing Technology of the Chinese Academy of Sciences, where he followed researcher Hu Weiwu, the chairman of Longxin Company, for his master’s and doctoral studies, becoming the youngest member of the Longxin research team at that time. After graduating with a doctorate, he stayed at the institute. By the age of 25, he had become the main architect of the eight-core Longxin No.3 chip.
He was selected as one of the 35 Innovators Under 35 by MIT Technology Review.
Interestingly, the name of the artificial intelligence chip launched by Cambricon is called DianNao, which is the pinyin of the Chinese word for computer. Currently, DianNao has derived several models, including 1A, 1H, etc. The name DianNao was suggested by a French member of the Cambricon R&D team, Olivier Temam, who came from the National Institute for Research in Computer Science and Automation (Inria). He suggested that instead of using a bland English name, it would be better to use a Chinese pinyin name, which would sound “exotic” to foreigners.
The most noteworthy aspect is that the DianNao chip not only features Cambricon’s independent architecture but also its independently developed instruction set, which is intriguingly named DianNaoYu (computer language). This is the world’s first deep learning instruction set, and the paper on the DianNaoYu instruction set was accepted at the top international conference in computer architecture, ISCA 2016, ranking first among nearly 300 submissions, signifying a groundbreaking achievement. Previously, China’s independently developed Longxin chip also used the MIPS instruction set under a free license, which can be modified independently but is still derived from abroad.
According to the Institute of Computing Technology, “The instruction set is the core of the computer software and hardware ecosystem. Intel and ARM control the PC and embedded ecosystems through their instruction sets. Cambricon’s pioneering progress in deep learning processor instruction sets provides technical support for China to occupy a leading position in the intelligent industry ecosystem.”
How do Cambricon’s AI chips compare to traditional general-purpose processors? Cambricon is a dedicated AI chip. Please note the word “dedicated”; therefore, Cambricon’s chips are not a substitute for general-purpose processors.
Since the basic frameworks of CPUs and GPUs are not designed for artificial intelligence, if a general-purpose processor is used to build a neural network of synapses at the scale of the human brain, it might require a power plant to supply it with electricity. For instance, AlphaGo used 1,000 CPUs and 200 GPUs for a single game, and the electricity cost per minute was as high as 300 USD, while the network scale was only one-thousandth that of the human brain.
Cambricon’s AI chip precisely addresses this issue—it can simulate neurons and synapses in computers, intelligently processing information, and through specially designed storage structures and instruction sets, it can process 16 billion neurons and over 20 trillion synapses per second, consuming only one-tenth of the original power. In the future, it may even be possible to fit the entire AlphaGo system into a smartphone.
Other FPGA solutions, while rapidly iterating, still lag behind dedicated AI chips in terms of computational speed and energy efficiency. Therefore, apart from Cambricon, other foreign companies are also tracking Cambricon’s research results, such as Google’s TPU.
Cambricon’s AI chips can have extensive applications in two major industries: cloud and end devices. In the context of the booming cloud computing industry, cloud servers face increasing big data computing pressures, making AI chips gradually indispensable. Additionally, the increasing intelligence and computational requirements of end devices also necessitate the use of AI chips.
On September 2, 2017, Huawei launched the Kirin 970 processor at the IFA exhibition in Germany, which was first used in the Huawei Mate 10 smartphone. Although not publicly announced, the AI chip behind it comes from Cambricon. The AI chip integrated into the Kirin 970 is referred to as NPU (Neural Processing Unit).
Riding on Huawei’s coattails, Cambricon’s sales will quickly increase. It is currently speculated that Huawei and Cambricon have adopted an IP licensing model, with Cambricon being the world’s first unicorn company in AI chips, as well as the first company to achieve large-scale mass production of AI chips, while its chip architecture and instruction set are entirely independent, indicating a bright future ahead.
I am now looking forward to when Cambricon will go public. Artificial intelligence will be a super industry, and theoretically, all hardware will need to implement deep learning functions.
In the semiconductor industry, the U.S. is super strong, while South Korea has a significant advantage in memory.
Europe has three top semiconductor companies: NXP, Infineon, and STMicroelectronics. Among them, NXP is in talks for acquisition with Qualcomm. If Qualcomm successfully acquires NXP, only two will remain in Europe.
Japan has three top semiconductor companies: Renesas Electronics, Sony’s CMOS image chips, and Toshiba Semiconductor. Among these, Sony’s CMOS image processing chips are one of the few industries in Japan that have achieved a significant turnaround in recent years, greatly increasing its market share and obtaining substantial profits. Toshiba is seeking to sell itself, and once Toshiba is sold, Japan will also be left with only two semiconductor companies.
We can see that the U.S. is super strong, while China and South Korea are on the rise, and Europe and Japan are experiencing slight declines.
Additionally, another area of significant decline is Taiwan. The Taiwan Semiconductor Industry Association estimates that the overall output value of Taiwan’s semiconductor industry will grow slightly by 1% in 2017, while the global semiconductor market size is expected to grow by 9.8% compared to 2016, indicating a declining market share for Taiwan’s “lifeline” semiconductor industry.
In terms of the entire semiconductor industry’s output value, in the first half of 2017, Taiwan’s output was approximately 1.144 trillion NT dollars, while mainland China was about 990 billion NT dollars. Taiwan’s semiconductor industry narrowly won, but mainland China’s development speed in the entire semiconductor industry clearly exceeds that of Taiwan.
According to statistics from the China Semiconductor Industry Association, from January to June 2017, the sales revenue of China’s integrated circuit industry was 220.13 billion RMB, a year-on-year increase of 19.1%. In contrast, taking the second quarter of 2017 as an example, Taiwan experienced a decline of 4.8%.
In the semiconductor sectors of design, manufacturing, and packaging, mainland China has already surpassed Taiwan in design and packaging. Taiwan’s IC manufacturing remains its stronghold, with the output value of Taiwan’s IC manufacturing in the first half of 2017 being 626.8 billion NT dollars, while mainland China’s output value was 257 billion NT dollars, with Taiwan maintaining its advantage. In IC packaging and testing, the output value of Taiwan’s IC packaging and testing in the first half of 2017 was 226.8 billion NT dollars, while mainland China’s output value was 360 billion NT dollars, indicating that this segment has also surpassed Taiwan.
From previously looking up to now surpassing Taiwan, as well as on the path of catching up with Japan, South Korea, and Europe, the progress of China’s semiconductor industry is evident. In the first half of 2017, China’s chip design industry grew by 21.1% year-on-year, with sales reaching 83.01 billion RMB, continuing to maintain the fastest growth rate in the world.
We should have confidence in China’s chip development. In the future, HiSilicon, Ziguang, and Cambricon will become the main forces of China’s chip industry. Although Cambricon currently has low sales and has not yet begun to profit, this company is destined to rise as a major player.
What Cambricon needs to solve is the issue of commercial experience. The founders are from a technical background in the Chinese Academy of Sciences, and in the competitive commercial market, they need to accumulate business experience and support, which is a challenge for Cambricon.
In addition to the three major players, there are many specialized Chinese chip suppliers that will rise, such as Zhongying Electronics, which provides display panel driver chips, and Goodix Technology, which provides fingerprint recognition chips.
Moreover, in the army of state-owned enterprises (do not underestimate state-owned enterprises; Unisoc is also a state-owned enterprise), many companies like Datang Microelectronics and HwaCom will ensure national security, such as the news from September 16, 2017, about the BeiDou navigation positioning chip released by HwaCom. The BeiDou industry will become a large industry in the future.
For instance, the Shenzhou spacecraft’s electronic control and anti-radiation chips, which were developed by Zhongke Xin Company under the China Electronics Technology Group, are also aimed at ensuring autonomy and national security. However, the scale of the aerospace industry will not compare to that of HiSilicon, Ziguang, or Cambricon.
Additionally, there is public interest in the domestically produced chips used in China’s supercomputers, such as the world’s fastest supercomputer, Sunway TaihuLight, which uses domestically produced Shenwei chips. The instruction set and architecture are extensions of the Alpha architecture from the U.S. company DEC, but this chip is only applicable in the supercomputing field and lacks competitiveness in the civilian market due to ecosystem issues and performance limitations. Therefore, it is also a niche chip.
Another example is Longxin. Dr. Hu Weiwu is a strong advocate for independent research and development, and he has cultivated the founder of Cambricon, Chen Yunji, and a group of talents in autonomous chip development. I believe this is Longxin’s greatest contribution. However, in terms of commercialization, Longxin has almost no hope. Establishing an ecosystem requires strong commercial strength and status. Currently, Longxin chips can only occupy a significant market share in specialized fields, such as anti-radiation chips used in BeiDou satellites or in dedicated fields like servers, industrial control, and embedded control. Achieving a significant market share in these areas would be considered a success for Longxin.
In 2015, Longxin’s sales first surpassed 100 million RMB, and in 2016, the sales revenue was also just over 100 million RMB. In 2016, there were 160 chip design companies in China with sales exceeding 100 million RMB, and Longxin was insignificant among them. Dr. Hu Weiwu is a technical talent, and the current treatment of Longxin, along with the operational system of similar research institutions and state-owned enterprises, makes it temporarily impossible to attract excellent high-end commercial talent. If Longxin can achieve a scale of 1 billion RMB or 2 billion RMB in the next decade, that would be quite good. Even if it reaches a scale of 10 billion RMB or 20 billion RMB by 2027, it will still not be able to enter the first tier of China’s chip industry.
The future of China’s chip industry must follow a profitable commercial path. From this perspective, only domestic chip companies led by HiSilicon, Ziguang, and Cambricon, as well as second-tier companies like HwaCom, Datang, Silan, and Goodix Technology, and possibly rising third-tier companies, represent the future of China’s chip design industry.
As for the chip design industry, looking at the development trends of the six powers—China, the U.S., Europe, South Korea, Japan, and Taiwan—China maintains a growth rate of 20%. If it continues this pace, it will quadruple in four years and reach a size four times its current size by 2025, becoming the biggest challenge to the U.S. South Korea will maintain its strong position due to high-intensity investments in memory chips. China’s Ziguang’s 64-layer NAND Flash is expected to be mass-produced in 2019, which will still have a technological gap compared to Samsung and Hynix, so in the short term, South Korea’s memory chip industry will not be undermined. The timeline for mass production of DRAM is still uncertain.
In contrast, Europe, Japan, and Taiwan are inevitably affected by the decline of their downstream consumer electronics brands, which will impact the upstream chip output value. Among these, Taiwan is particularly at risk, as the decline in the IC design industry is a major trend due to the concentration of consumer electronics brands towards mainland China, while the self-produced chip share in mainland China is continuously increasing. Taiwan lacks an automotive brand industry to compensate for these losses. With the decline of the design industry, the foremost task for Taiwan is to preserve the manufacturing industry of TSMC.
In the following articles, we will discuss the semiconductor manufacturing and packaging sectors, as well as the upstream equipment, which are China’s largest shortcomings.