Huawei’s ban has been almost 30 days. Let’s do some simple deductions and the trend behind Huawei. We will write according to key time points. The following are some immature thinking.
1. Huawei’s 120 days and 2 years
515 ban on Huawei, Americans leave a 120-day buffer period, the US Department of Commerce stipulates that companies and institutions can submit comments before July 14, we believe that a large number of US semiconductor equipment and raw materials companies and TSMC We will actively prepare relevant opinions and evaluation reports and submit them to the US government, and 120 days later is September 12, we see whether Americans will continue to postpone.
In addition, in the past four months, China can do some things. The most important thing is to prepare a list of unreliable entities. The United States is now affected by the epidemic and riots. The economy is not very good. This is also beneficial to us. This It means that our physical list counterattack will make the United States even more uncomfortable.
In theory, it is possible for the US government to extend the buffer period and relax some sanctions based on the interests of its domestic companies, just as in the past two years of the Sino-US trade war, the scope of each time the US imposed additional tariffs has been upgraded At that time, a hearing will be held to listen to the opinions of the enterprise. While expanding the scope of additional tariffs, it will inevitably publish a list of tariff exemptions. In fact, it will reduce the scale of tariffs imposed on China to protect its own interests to the greatest extent.
And sanctioning giant technology companies such as Huawei and being one of the world’s largest buyers of semiconductor chips is also a process of exploration for the United States.
Especially in the semiconductor manufacturing and equipment and material manufacturers in the United States, they must be more worried. We will talk about this later.
Of course, for us, we can only stick to the bottom line of thinking, and we cannot assume that the United States will open up. After the 120-day buffer period expires, the real test will begin.
Pay attention to the US Department of Commerce’s announcement, which is described for 120 days, and pay attention to the bold part:
To prevent immediate adverse economicimpacts on foreign foundries utilizing U.S. semiconductor manufacturingequipment that have initiated anyproduction stepfor items based on Huawei design specifications as of May 15,2020, such foreign-produced items are not subject to these new licensing requirements so long as they are reexported, exported from abroad, ortransferred (in-country) by 120 days from the effective date .
This paragraph is very vague. Those that have already started production on or before May 15 can continue to be exported and shipped. In theory, if Huawei rushes to place an order on May 15, TSMC will still be able to Continue production.
Huawei has circulated on the Internet that it urgently placed an order for 700 million US dollars to TSMC to request production and supply (not confirmed by TSMC or Huawei). If you compare it with the original English text above, then there is Huawei competing for time on May 15 The possibility of sending an order to TSMC to place an order.
Let’s look at smartphones first. The total number of tens of millions of base stations in the world is far from comparable to the global stock of 3 billion smartphones. The gap between the two is more than a hundredfold, so Huawei Smart The mobile phone business will face huge challenges. Currently, because TSMC continues to produce, then Huawei’s mobile phone shipments are now no problem, so the key is how long can Huawei’s mobile phones continue to sell after 120 days?
It is very obvious that in these 120 days, Huawei will definitely ask TSMC to give priority to the production of the next generation of Kirin chip 1020 (the model circulating on the Internet, Huawei has not officially announced the official name), because this will be used for the next generation of flagship For an Android flagship phone, the peak period of general sales is the first three months to six months. For the P30 series mobile phones released by Huawei in March 2019, it has been on the market for 85 days, which is almost three. Monthly sales exceeded 10 million units, and sales reached 16.5 million units six months after being listed. You can experience the peak sales period,
For Huawei’s flagship aircraft, the first three months are the golden period of sales, and the second three months are the silver period of sales.
After the peak sales period, the price will become cheaper and cheaper, so Huawei must give priority to the production of next-generation chips.
So how many chips can TSMC produce?
Let’s look at it from two dimensions. First, from an order point of view, if the emergency order of 700 million US dollars is true, then the price of the mobile phone processor is generally several tens of dollars. If we calculate according to each 50 US dollars, assume 7 100 million US dollars are all mobile phone processors (actually believe that there will be a part of 5G base station chips, the number is relatively small), then it is about 14 million chips.
If we estimate in the more expensive direction, assuming a $100 one, there are also 7 million chips.
That is to say, but the urgently added 700 million US dollars can produce tens of millions of new-generation processors.
Note that this is only an addition, so before this, Huawei may have placed an order.
From the perspective of production capacity, TSMC’s actual production in 2019 is 10.1 million wafers equivalent to 12 inches, and the average monthly actual production is as high as about 840,000 wafers.
So how many mobile phone processor chips can a 12-inch wafer produce?
The surface area of the 12-inch wafer is approximately 70,659 square millimeters. According to the TechInsights measurement of the Kirin 990 5G chip, the area is 10.68×10.61=113.31 square millimeters, which integrates up to 10.3 billion transistors, which is the world’s first A mobile SoC with more than 10 billion transistors.
As a comparison, the area of the previous generation Kirin 980 is 8.25×9.16=75.57 square millimeters. Of course, the more advanced Kirin 990 chip has a larger area because the baseband chip is also integrated.
Therefore, we estimate that the area of the Kirin 1020 is also about 100 square millimeters. If this area can be used 100%, a wafer can produce 700 processor chips. Of course, this is definitely not achieved.
The actual number of chips produced is roughly calculated according to the following formula. According to the following formula, a 12-inch wafer can produce about 640 chips. Note that this is only the theoretical value calculated by the general formula. The actual production of wafers The edge loss will be higher than this, which will cause the output of the chip to be lower than 640.
In addition, the yield of the chip particles is also yielded, and then the chip will be packaged in the packaging factory, and the casing and pins are put on. This process will also involve yield loss.
Therefore, it is possible for us to produce about 500 chips in the end. Note that the final actual output, especially in the early stage, is likely to be less than this number, then we will calculate it according to 400 chips, which is more conservative.
Then TSMC only needs to produce 25,000 12-inch wafers to meet the needs of 10 million Kirin processor chips.
Note that Huawei’s Kirin 1020 is expected to use 5nm capacity, so TSMC only needs to guarantee 25,000 12-inch wafers within 120 days, and about 6,000 wafers per month.
Correspondingly, if it is 20 million chips, it will need to produce 50,000 wafers.
Although 5nm is a new process for TSMC to be mass-produced in Q2 in 2020, it is clear that we can have great confidence in TSMC. After all, TSMC has experience in large-scale production, and its total actual production in all processes in 2019 The scale of 840,000 pieces per month is proof.
And from the perspective of TSMC’s 7nm historical expansion rate, it is very fast. TSMC has produced more than 10 million 5nm Kirin chips for Huawei within 120 days. The possibility is very high, which can basically support Partial sales of the first generation flagship.
For the previous generation of chips represented by Huawei Kirin 990, the importance is relatively low, because mobile phones equipped with this chip are destined to continue to reduce prices, and TSMC’s 7nm production capacity is originally relatively large. According to TSMC’s financial report, In the first quarter of 2020, 7nm process revenue accounted for as much as 35%, so it is not a problem for TSMC to use the 7nm process to stockpile Huawei mobile phones that have been on the market continuously in the past 120 days.
According to the above estimate, after the 120-day buffer period ends, Huawei’s next-generation smartphone equipped with Kirin 1020 can also sell 10 million units. At the same time, Huawei’s previously released mobile phones can continue to store goods through 7nm TSMC capacity. Sales.
Huawei will be able to sell 60 million mobile phones per quarter in 2019, and the ban will take effect in September 2020. Huawei will consume about 60 million mobile phone processors in a month, which will be the first and second in 2021. This quarter, the huge challenge of Huawei’s smartphones came, and the inventory of mobile phones was gradually exhausted.Its terminal business revenue will face a decline.
2021 is also the first major threshold for Huawei to survive. As for how to deal with it, I will talk about it later.
Let’s look at the communication network again. In the previous article, we analyzed that 5G network is the core for Huawei. Huawei must guarantee 5G network construction, and the number of chips required for 5G base station construction is not much. For example, according to the Ministry of Industry and Information Technology, China has built 130,000 base stations by the end of 2019, and the three major operators and China Tower have invested 57.3 billion yuan in 5G this year. It is expected that the number of 5G base stations will exceed 550,000 by the end of 2020. Execution capacity is likely to exceed 600,000 units, that is, China is expected to build 420,000-470,000 5G base stations in 2020.
China is the core region of global 5G construction. The global 5G base station construction volume will be at the level of 1 million by the end of 2020. It can be seen that the demand for chips is not large, and it is completely incomparable with Huawei smartphones. Huawei 2019 The P30 series flagship mobile phones released in March have sold up to 16.5 million units in 6 months.
Huawei’s base station chips are normally stocked for +120 days to continue production, which can guarantee long-term supply to the base station. According to the annual shipment of hundreds of thousands of 5G base stations, it is feasible to hoard chips for 2 years. .
According to a Nikkei Chinese website report on May 28, it learned from its own channels that Huawei had hoarded US chips for 1.5-2 years to supply its base stations and servers.
Note that Japan Economic News is also a well-known gossip communicator, and the information it does not disclose from many sources is not always accurate.
But I think that judging from the logic and possibility, hoarding chips for 5G base stations for 1-2 years is what Huawei can do. Especially with 120 days of continuous output, I tend to think that hoarding must be possible For more than one year of use, the stocking problem of 2 years is not a big problem.
Because the CPU of the base station does not yet use 5nm, such a large device does not need to be as compact as a mobile phone, which means that the production capacity will not be limited by the 5nm production capacity.
Then this pairHuawei is good news, because 5G construction starts in 2019. According to the speech of Miao Wei, the Minister of Industry and Information Technology on November 21, 2019, 113,000 5G base stations were opened nationwide at the time, and it is expected to be reached by the end of the year. 130,000. Therefore, China alone has built 130,000 base stations in 2019.
So if 2019 is the prelude to 5G, 2020 is the first year of the peak period of global 5G base station construction. If Huawei’s chip inventory can last for 2 years, it is not difficult to do in terms of quantity. Then, the supply of these three construction peak years from 2020 to 2022 can be guaranteed, which means that despite the US sanctions, Huawei is destined to lose some customers’ orders, but it is still the main player in the global 5G field.
This is very important, because if Huawei cannot seize the position during the peak period of 5G network construction, then it will be very troublesome in the future. Can’t customers always replace the network that has just been invested heavily with Huawei?
So Huawei keeps the communication network construction from 2020 to 2022. The problem is not too big. For communication equipment, the key node is in 2022.
So what to do after 2022, we will talk about it later in this article.
However, by the middle or end of 2022, Huawei’s communications equipment business will also face the challenge of exhaustion of chip stocks, then 2022 is the second threshold Huawei faces.
2. How long does Huawei need to persist?
If everything has not changed, the United States has not postponed, nor does it consider the factors of Sino-US gaming.
As mentioned above, Huawei still has 2 years. By the middle or end of 2022, its core business 5G communication equipment will also face the risk of being unable to ship. Of course, if we are more optimistic, Huawei will start from now on Start to actively take business contraction, abandon some marginal customers, concentrate on supplying core and important customers, or Huawei actually hoards more chips than estimated, then Huawei may even continue to 2023, of course, we still follow 2 Annual calculation, that is, calculation by 2022.
Is it possible for domestic foundry manufacturers to come up with a more advanced de-beautifying chip production line within 2 years, that is, by 2022? Here is more advanced, do not expect 7nm or 14nm, that is, 28nm.
The answer is that no one knows, because this has never been doneThings, but with this possibility.
We analyze the existing information. In theory, if we take out every link of domestic semiconductor equipment, there are domestic manufacturers, and many have developed 14nm equipment, and even some links have reached 7nm. the following.
In the field of semiconductor production equipment, even those products with a very low localization rate are made by domestic manufacturers, and most of them have been applied in the production line.
For example, ALD is a device with a low localization rate, but from the picture below, from the official website of North China Creation, the first domestically produced ALD (atomic layer deposition equipment) was achieved in 2018, which can achieve FinFETs of 28nm-14nm, etc. Process requirements.
Even if the production line’s mass production verification period is as long as 1-2 years, it should be verified this year.
Chinese companies are doing all the fields of semiconductor production equipment, not starting from scratch in 2020, including those areas that are monopolized by American and Japanese manufacturers and are considered to be extremely difficult to localize.
The most important lithography machine, according to the time schedule of the lithography machine 02 special, Shanghai Microelectronics should have a 28nm node lithography machine at the end of this year or early next year, plus a year of production line verification. Is it possible to make a 28nm de-beautification production line within 2 years and realize preliminary trial production or even mass production?
Regarding the progress of the lithography machine, I see that there are many doubts about Shanghai Microelectronics. For example, although Shanghai Microelectronics has a 90nm 600 series IC front lithography machine, it has only been exhibited at the exhibition in recent years, or In the laboratory, there has never been actual mass production on the production line, and SMIC has not purchased it.
Until May 2018, Shanghai Microelectronics shipped a total of 100 lithography machines, and concentrated on LED and IC back-end packaging lithography machines.
In fact, things are not so complicated. Shanghai Microelectronics’ 90nm lithography machine was successfully developed in 2016. In 2017, the task of “90nm lithography machine prototype development” passed the field test of the 02 major scientific and technological special expert group. In 2018, only Completed and officially verifiedClose.
At this time in the world, TSMC has taken the lead in mass production of 7nm, and SMIC is also conducting customer import verification of the 14nm process.
Shanghai Microelectronics’ 90nm IC front-end lithography machine can be seen more as a process of technological R&D accumulation.
For SMIC, the old-fashioned lithography machine has already been depreciated, the price is already very cheap, and there are many cheap second-hand goods available in the market. Then spend 2018 to verify a 90nm domestically produced It takes another one or two years to introduce the domestic 90nm lithography machine into the production line in 2019-2020, which has little financial and technical significance and cannot be purchased on a large scale.
It is more meaningful to verify the more advanced domestic lithography machine.
The same is true for Shanghai Microelectronics. In the current order of tasks, the next-generation 28nm node lithography machine has been in research for several years compared to investing in the establishment of a 90nm lithography machine production line. This is the key investment direction. .
We can see from the well-established domestic semiconductor equipment manufacturers in China, North Huachuang, China Micro Semiconductor, etc., we must make our equipment 28nm-14nm, and the etching machine of China Micro Semiconductor even enters Advanced processes with higher prices below 7nm, or mainstream processes, so that the foundry has the financial incentive to import domestically.
On the other hand, semiconductor manufacturing equipment manufacturers can truly achieve meaningful product sales.
Huawei is on the physical list in 2019, and the domestic people are supporting Huawei. I also got two P30 pros, one of which is for family members, and you can see the problem from it. Even if we support Huawei, the new one is new. For Huawei’s flagship machine, if Huawei still sells the P6 equipped with the K3V2 chip, the motivation of the people who eat melons to support it will also be greatly weakened.
Why should I mention P6? At that time, I highly recommended my wife to buy one. As a result, she actually left a very bad impression of Huawei mobile phones in actual use, which turned into a loyal Apple fan, and I lost A lot of money.
We must understand the progress of technology with market-oriented thinking.
A new generation of products has been made, even if it is a technological advancement for the manufacturer itself and even for China, it does not mean that it has enough competitiveness to survive in the market, that is, it cannot compete with buyers in the competition. Products that bring more benefits are not real market-oriented products.
We will use domestic as ionAs an example, Shuo Ke Zhong Ke Xin of the implanter, according to the report of Beijing Business Daily on November 27, 2019, the ion implanter independently developed by Shuo Ke Zhong Ke Xin is entering the major domestic integrated circuit manufacturers, among which Zhong Shu The fluid ion implanter CI P900 series has already passed the verification of SMIC’s industrialization, laying a solid foundation for the realization of high-volume applications.
In 2019, the company sold 11 ion implanters.
“After years of technology accumulation, we have realized the import substitution capability and independent controllability of ion implanter, the key equipment for integrated circuit manufacturing in the process field above 28nm, with an annual output of 30 ion implanters. Ability.”
Shuoke Zhongkexin General Manager Shu Yongdong introduced that the medium beam ion implanter developed by R&D has mass produced more than 4 million 12-inch wafers in 65-28nm process, which has reached the level of the same type of equipment abroad, and the products have entered the market in batches. .
The large beam ion implanter process covers 28nm, and the 65-28nm process produces more than 200,000 wafers. The high-energy machine is expected to enter the client verification by the end of 2020. The customized ion implanter is mainly aimed at the market of 6 inches and below. The equipment is widely used and can be customized according to the specific needs of customers.
It can also be seen from the above that the equipment of China Electronics Co., Ltd.’s Zhongkexin’s industrial verification at SMIC can also achieve 28nm nodes, which also indicates that more advanced node devices still need to be verified in the market. It makes sense meaningfully.
We return to the research and development progress of the 28nm node immersion lithography machine. According to the progress requirements of the 13th Five-Year Plan, China will be responsible for integration by Shanghai Microelectronics at the end of 2020, and a 28nm lithography machine needs to be developed. It can be confirmed from the massive amount of public information that it takes 1-2 years from R&D to real mass production. We can calculate it according to 2 years, which is expected to achieve 28nm lithography by 2022 Mass production.
Of course, the lithography machine, even ASML produces hundreds of units a year, and Shanghai Microelectronics produces dozens or dozens a year, which can already be called mass production.
It is generally believed that Shanghai Microelectronics is fighting alone. It is not that Shanghai Microelectronics is actually an integrator, more responsible for the research and development of switchboard integration.
For example, in the national planning of my country’s lithography machines, the Chinese Academy of Sciences gave Shanghai Microelectronics strong R&D support and was responsible for the research and development of multiple core subsystems.
middleThe Academy of Sciences has been involved in many semiconductor localization tasks before. China is already mass-producing NAND FLASH’s Yangtze River Storage, which was jointly developed by the Institute of Microelectronics of the Chinese Academy of Sciences and Yangtze River Storage. At the same time, he is a doctoral tutor of the Chinese Academy of Sciences.
In 2003, he graduated with a Ph.D. in microelectronics and solid-state electronics from Peking University and entered the memory division of Samsung Electronics Semiconductor R&D Center in South Korea to engage in flash memory technology research and development.
In 2010, he returned to China to join the Institute of Microelectronics of the Chinese Academy of Sciences to engage in storage technology research. In August 2014, he joined Wuhan Xinxin Integrated Circuit Manufacturing Co., Ltd. and was responsible for the technology development of 3D NAND memory.
Let’s talk about the lithography machine again. Let’s take Beijing Guowang Optical Technology Co., Ltd.’s spring 2020 campus recruitment information released by Peking University’s official website as an example. The standard of 200,000-250,000/year (special excellence is discussed separately) is recruiting fresh graduates. The recruitment profile is as follows:
Guowang Optical was formally established on June 1, 2018, and its registered place is Beijing Economic and Technological Development Zone.
In July 2019, the Changchun Institute of Optoelectronics and the Shanghai Institute of Optoelectronics of the Chinese Academy of Sciences completed the capital increase of Guowang Optical’s intangible assets. At present, the registered capital of Guowang Optical is 3 billion yuan, and it has nearly 200 authorized invention patents,
Fully inherited all intellectual property rights formed by the first phase R&D task of the exposure optical system of the 02 special large-scale IC manufacturing projection lithography machine.
The first set of 90nm node ArF projection lithography exposure optical system developed by Guowang Optics’ core team has been successfully delivered to users in 2016 (Remark: Shanghai Microelectronics).
At present, the 02 special phase II core task undertaken by the team-the research and development task of the ArF immersion lithography exposure optical system for 28nm nodes is progressing smoothly;
In the second half of 2019, Guowang Optical will start the construction of R&D and production bases in the B13 block of the Beijing Economic and Technological Development Zone. The project has a planned investment of 6 billion yuan, covers an area of nearly 110 acres, and has a planned construction area of more than 130,000 square meters. The estimated construction time is 3 years. After it is fully completed, Guowang Optics will have the research and development, design and mass production supply capabilities for the exposure optical system products of very large-scale IC manufacturing projection lithography machines at 110nm node, 90nm node, 28nm and below nodes.
At the same time, we have successfully developed ultra-precision optical processing, inspection, and adjustment equipment, as well as high-performance micro-objective lenses and ultraviolet imaging probes that are being developed.Testing systems and other products will also be put into the market one after another. Such high-market-value derivative products will work together with lithography machine exposure optical system products to build a successful product foundation for the Guowang optical market.
For the time node of the above production base construction, the production base was built in 2022 and has the capacity to mass produce 28nm node and below lithography exposure optical systems, indicating that the 28nm node immersion lithography machine estimated by Guowang Optical is really The time point to achieve mass production on the production line is in 2022,
Then this and the successful development of 28nm immersion lithography machine in 2020, the test was completed in 2021 and the production line was verified, and the scale of 28nm node lithography machine production line was achieved in 2022, and the time node for achieving mass production is Agree.
Although the time node can generally be completed successfully, for Huawei, as a bottom line thinking, it cannot be assumed that China will be able to make a 28nm de-beautification line in 2022, and this time needs to be extended to 3 years Even longer.
3 years or more, this is Huawei’s survival time when HiSilicon cannot be mass-produced.
Of course, the country can take countermeasures, and there are many ways to countermeasures. We must be reasonable in our countermeasures, and we should not make those foreign companies that comply with the law and regulations feel cold. The list of unreliable companies at the first meeting Yes, the companies that performed badly in the 2019 US ban on Huawei, typical of which are Flextronics and FedEx. These two US companies, Flextronics once seized Huawei’s hundreds of millions of yuan in goods, and this is still implemented on Chinese land. Operation, FedEx actually transferred Huawei’s package to the United States, suspected of theft and betrayal of trade secrets.
The second batch is of peer-to-peer companies, namely Qualcomm, Cisco and the like. They and Huawei are in a competitive relationship. Since there are no 5G equipment manufacturers in the United States, the two companies are still involved in accordance with the principle of reciprocity. To 5G and communications, and not deeply involved in China’s industrial chain, does not involve a lot of China’s employment and exports.
There are many others, this article will not expand.
Here we will not consider the factors of national countermeasures, which may cause the United States to relax or delay sanctions on Huawei. We only consider how Huawei persists in survival.
Note that Huawei still has 1-2 years to make preparations, and the current response can be carried out as follows:
First point: EDA localization
Take Huawei as the core, and develop EDA softwareHuawei has taken the lead in the division of labor and cooperation of domestic EDA companies to realize the localization of EDA. Huawei has 2-3 years to go and has great hopes.
EDA Huawei already has many years of experience in use and also has a foundation for pre-research.
Second point: Continue to purchase non-US chips for shipment.
The United States currently does not prohibit Huawei from purchasing external non-US chips. Huawei can expand the use of non-US chips. Of course, people who eat melons may ask, you may be useless, because the United States may order at any time. Huawei cannot buy any chips made with US production equipment and materials.
This kind of bottom line thinking is right, but for Huawei, this work has been done before, rather than just rushing to design products that use non-US chips.
That is, Huawei has been doing multi-vendor systemization work, including Huawei’s mobile phones, not only HiSilicon is a supplier, which is now bringing benefits to Huawei.
The picture below is the 5G version of the Honor Play 4 first sold by Huawei Honor on June 12, using the MediaTek Tianji 800 chip, priced at 1999 yuan.
At least for now, this phone will not be affected by the 2021 limit of the Huawei Hisilicon flagship.
Comparing the third quarter of 2019 with the third quarter of 2018, the proportion of Huawei mobile phones equipped with MediaTek chips has increased from 7% to 16.7%, more than doubled. This year, this situation is expected to continue to increase.
We have reason to believe that Huawei is now gradually expanding the proportion of mobile phones equipped with MediaTek chips. If this ratio rises to more than 30%, then Huawei can still retain a considerable portion of mobile phone sales by next year, and will not be wiped out by the army.
In addition to Taiwan’s MediaTek, non-US chip manufacturers also have STMicroelectronics in Europe, Ziguang Zhanrui in mainland China, etc. On May 1, the Science and Technology Board Daily reported that many of them Industry sources confirmed.
HuaweiAt present, it is working with STMicroelectronics to develop chips, and it is unknown what products are used in it.
Note that this road may be blocked by the United States at any time, but at least it is now feasible and has continued to bring revenue and profits to Huawei.
Third point: Push TSMC to make a beautification choice
Huawei will inevitably promote the de-beautification of suppliers (foundries, production equipment, raw materials) in Taiwan, South Korea, Japan, and Europe. In addition, China can also drive upstream de-beautification with super-large-scale tariff weapons.
The progress of my country’s domestic equipment and raw material manufacturers is certainly gratifying, but in addition to domestic equipment and materials, it is necessary to fully mobilize third-party forces outside of China and the United States. In the global semiconductor production equipment and raw material manufacturers, except the United States,
Japan has the world’s second largest share in production equipment and semiconductor raw materials has the world’s largest share. Europe and South Korea also have semiconductor production equipment companies.
Huawei can rely on the huge size of Huawei and the strong appeal of the Huawei brand. In addition to SMIC, it also needs to promote TSMC. Samsung also mobilizes its upstream Japanese, European, and Korean semiconductor production equipment manufacturers. Do de-beautification research and development, and establish de-beautification production lines.
Upstream foreign production equipment manufacturers have mostly infiltrated the industry for more than ten years or even decades. The source of parts has long been diversified, and it is possible to separately develop monopoly American parts.
The risk of this action is, whether it is Taiwan’s TSMC, Samsung in South Korea, or ASML in the Netherlands, do they have the courage to dare to beautify, after all, American hegemony, and Taiwan, South Korea, the Netherlands and the United States The political relationship is there, this is a question.
However, if they don’t do this now, it means that 100% of Huawei’s big customer is completely transferred to SMIC. Once SMIC has made a de-beautification line, they can immediately get Huawei’s order. Even if Huawei’s sales of chips are cut due to chip outages, its huge intangible assets of brand value and appeal will still exist. Once it is supported by a de-beautifying production line, it will soon recover. You can refer to Huawei’s smartphones after 2010. The market’s sales development rate, Huawei’s order value is very high, accounting for 14% of TSMC’s revenue in 2019, and TSMC’s revenue was 35.774 billion US dollars in the year, which means that almost 5 billion US dollars of orders came from Huawei.
And SMIC’s full-year revenue in 2019 is only $3.116 billionObviously, even if SMIC is progressing on the domestic equipment, for example, the lithography machine with only 28nm node at the beginning,
It is impossible to take such a large amount of Huawei orders, but it has undoubtedly obtained a stable global large customer, and its development will enter the fast lane.
This is undoubtedly a pressure on TSMC, especially on TSMC, not to beautify the production line means completely and permanently abandoning Huawei. It will be a painful choice, and even a strategy related to the future development of TSMC for 10-20 years. Choice.
Huawei’s corporate capabilities have been fully demonstrated in the global 5G and smartphone markets. Huawei has achieved an absolute lead in the 5G field, beating companies in Europe, North America, Japan and South Korea, Nortel Networks and Motorola in North America All communication equipment companies went bankrupt or withdrew from the market, while Ericsson and Nokia in Europe gradually fell behind Huawei, and in the smartphone market, if it was not suppressed by the United States, 2020 should be the year in which Huawei’s smartphone sales reached the top in the world. The strong competitiveness reflected in the evidence.
So this means that once Huawei recovers, with its global attack at the system brand level, HiSilicon’s share will inevitably increase, which means that it will form a squeeze on TSMC’s customer sales.
Not only that, if TSMC does nothing on the de-beautification line and is regarded as abandoning Huawei completely, then for other mainland Chinese chip companies, TSMC will also be regarded as an unreliable business partner. SMIC is getting bigger and bigger with the support of Huawei’s rights, so the chip manufacturers in the mainland will inevitably be more inclined to OEM with local manufacturers.
At the same time, the more powerful Chinese government can also use national forces to drive beautification. It can be stipulated that for all chips used in products that are ultimately sold in the Chinese market, its foundry needs to gradually reduce the production line of American equipment. And technical content, a debeautification schedule should be given, otherwise tariffs will be levied according to the actual debeautification progress of each manufacturer. For example, Samsung and TSMC are required to give reports on the US content of their production lines. For example, if Samsung is 15% and TSMC is 10%, then TSMC must be reduced to 8% within one year and 6% within two years, otherwise 30%-100% tariffs are levied, and the de-beautification tariffs for Samsung can be more relaxed. By loosening and tightening the two, if TSMC does not de-beautify, it will greatly increase Samsung’s competitive advantage and force TSMC to take the initiative. De-beautification, at the same time, greatly hit the US semiconductor production equipment manufacturers.
This is to use China’s huge market advantage to force the foundry to remove US equipment. The specific operation method can be further discussed. For example, if TSMC monopolizes the most advanced 5nm process, will the tariff increase only hit the bigLu himself?
I think there are two points:
The first point is that TSMC has a large number of 7nm, 12nm, 16nm, 22nm, 28nm and above production lines. These production lines have competitors and can take the lead in driving these processes through market + tariffs. Without beautification, the cost of relative competitors will rise sharply.
The second point is that for the most advanced process, TSMC may not be subject to “de-beautification” tariffs, but if its biggest competitor Samsung, even if its process is close to TSMC, if there is a competitor, then mainland China will Immediately start the “de-beautification” tariffs to drive its de-beautification.
As a giant company that accounts for 50% of the global OEM market, TSMC has a diverse source of suppliers and has great potential for beautification. The following figure is a list of TSMC’s suppliers. There are five silicon wafers. There are seven suppliers, such as polishing liquid polishing pads, etc. For TSMC, there is no question of whether to beautify it, only the willingness or daring to make the determination to promote the beautification.
Even if the de-beautification is not completely reduced to 0%, even if it is one point down, it is a blow to American semiconductor production equipment and raw material companies.
TSMC is just an example, as are UMC and other companies.
If TSMC can be promoted to participate in the de-beautification process, it will not only be a blow to US semiconductor production equipment, but also benefit domestic semiconductor equipment entering TSMC.
Of course, it needs to be emphasized that the progress of SMIC and its upstream domestic equipment manufacturers is the core. This is to strengthen itself. Using China’s huge market, it is forcing upstream manufacturers to beautify and attack American semiconductors. This is to attack the enemy, regardless of Whether it can be achieved is also worth trying.
Fourth point: use technology to counter the United States
The United States has chosen its own technical commanding heights to fight against Huawei, so Huawei can also choose to use its own technical commanding points to fight back against the United States. Among the 5G patents, Huawei holds the leading position in the world and holds the most necessary patents in the world. It is also the global 5G The most important setter of standards.
As long as any company in the United States uses 5G-related technologies, it will inevitably be unable to bypass Huawei’s patents.
Because the use of patents requires Huawei’s license, and the other party does not have enough patents for exchange, then a high patent fee is charged from the US company to sell the license, and even not to license the use of patents to some US companies is an option. Of course, the weaponization of patents is usually not a good choice. There are a lot of industry slapsticks, and there are many patent hooligans who make money through litigation, but they are generally small companies.
However, when it comes to survival, the situation is different, and Huawei does not need to ask for prices, asking for reasonable patent fees from American companies, whether legally or morally.
Of course, there is also a method in the United States, which is to claim that the price of the patent is too high to pay, or to delay payment. Reuters reported on June 12, 2019 that Huawei has asked the US operator Verizon to pay for more than 230 patents. , Totaling more than 1 billion US dollars, the amount is equivalent to half of Verizon’s net profit in the fourth quarter of last year.
In fact, this money has not been paid to Huawei so far. Another problem is that the United States can also use rogue financial sanctions on Huawei to prohibit Huawei from using US dollars, which prevents patent fees from reaching the Huawei account from the US
But at least, the first can make the United States and American companies morally disadvantageous, and in the public opinion field accuses China of being passive when it accuses China of stealing American intellectual property.
The second Chinese government can also impose fines on American companies that refuse to pay patent fees and plagiarize the use of Chinese technology. It can completely imitate the way the United States treats foreign companies, and punish such companies that are dishonest and do not respect intellectual property rights. Penalty.
US Emperor fined ZTE more than two billion US dollars. This amount can be used for reference.
The third US company has a high degree of globalization. Even if there is no business in China, there will always be business in other countries. Huawei can recover patent fees by prosecuting in third countries, and can use euros or other in third countries. Currency payment.
The fifth point: multi-service development
I have analyzed this in my previous article many times. I can see the previous article.
Many Huawei businesses do not need chips, and the previous patent fees are more typical, for example, the antennas in the base station;
Or you need a chip but you don’t need to be technologically advanced. You can purchase domestic chips, such as routers, security products, TVs, etc. Now you can ship with HiSilicon chips. I believe that you can find other domestic chip manufacturers within two years. Designing alternatives is not difficult.
Huawei has caused great threats to Haikang and Dahua in the field of security.
You can also engage in third-party integration. Huawei provides product design and key components other than chips. Integrators purchase chips on their own. For example, Huawei is a server manufacturer. It is entirely possible for third parties to make servers. Huawei can provide system design.
For another example, Huawei is now entering the auto business and engaging in auto parts.
Onboard operating systems used by cars are also products.
Sixth point: Re-design the system and small-scale IDM
In front of this article, we have said that the global demand for base stations is not large, and all base stations in the world add up to tens of millions of levels. For Huawei, the number of base stations shipped is hundreds of thousands to one million in a year. .
In fact, this is equivalent to the shipment of military products. China’s military products all require chips. They were blocked by the United States more than 20 years ago, and this did not hinder China’s military product system. More and more advanced, reaching world-class level, this is the credit for system design.
For a miniaturized system such as a mobile phone, chips with advanced manufacturing processes are indispensable. It is difficult for Huawei to get rid of its dependence on the United States. The demand for advanced process chips has dropped significantly.
The chip can also be called a large-scale integrated circuit, which is essentially a collection of circuits, which can reduce the dependence on advanced chips by redesigning the system.
When the product does not require advanced process chips, Huawei has a lot of choices. Note that because base station chips are only equivalent to the demand for military products, a small-scale, low-process production line can meet the requirements.
The first is to provide R&D resources and fully support SMIC and itsYou’s equipment company will mass-produce HiSilicon chips in its beautified production line for 28nm lithography machines within 2-3 years.
Secondly, because SMIC needs to build a fully de-beautified production line, SMIC is currently unable to supply Huawei, and for Huawei itself, it has become an option to share the risks of manufacturing chips on a small scale.
Note that the current US sanctions only mean that HiSilicon uses EDA software designed by the US and chips produced by US equipment and raw materials in the foundry. The foundry cannot be shipped to Huawei.
However, Huawei’s own chip production and supply for its own needs, at least not until now.
(i) Items, suchas semiconductor designs, when produced by Huawei and its affiliates onthe Entity List(eg, HiSilicon), that are the direct product of certainU.S. Commerce Control List (CCL) software and technology; and< /p>
(ii) Items, such as chipsets, when produced from the design specifications of Huaweior an affiliate on the Entity List(eg, HiSilicon), that are the directproduct of certain CCL semiconductor manufacturing equipment located outside the United States.
Such foreign-produced items will only require a license when there is knowledge that they are destined for reexport, export from abroad, or transfer (in-country) to Huawei or any of its affiliateson the Entity List.
So Huawei bought first-hand or second-hand semiconductor production equipment, built its own small-scale chip production line, and learned Samsung’s move toward IDM.
Note that the base station chip is only used for military products, and the scale does not need to be too large. See TSMC’s Nanjing 16nm factory, which produces 20,000 wafers per month and invests 3 billion US dollars, about 20 billion yuan. TSMC’s Nanjing plant produces 240,000 wafers per year, and each wafer makes hundreds of chips, which is tens of millions per year.
For Huawei’s base stations, it does not need a production line of this size, and it does not need a chip of 16nm process. Therefore, the investment is not so large. Huawei only needs a small investment in a small-scale production line.
In addition, although it is difficult to find talents for chip production below 28nm, it is relatively easier to find talents with 28nm and lower level processes.
In addition, the production line of the old-fashioned process is relatively technically easier and the price is cheaper. The picture below is from the speech PPT of the national 02 special expert Ye Tianchun. This speech was in September 2018. It can be clear that at the time, The localization rate of the 65/55nm production line can be achieved by 31%, and the localization rate of the 40nm production line can be achieved by 17%. Now that one or two years have passed, I believe that this ratio has increased again.
From the proportion of the localization rate, it can be seen that the equipment of low-level processes will have a relatively low technical content and are easier to buy.
The lithography machine is still a bottleneck, but Huawei can choose the 28nm coming out of Shanghai Microelectronics next year, or it can choose to buy a second-hand lithography machine.
Taking the refurbishment of second-hand lithography machines as an example, Anxin Semiconductor in Fujian, China is doing lithography machine renovations. In March 2020, it delivered a lithography machine worth nearly 10 million yuan to Hikvision. At the same time, in April, another lithography machine was shipped to a domestic research institute for CMOS research. Zhang Qi, general manager of Fujian Anxin Semiconductor, said that at present, we have not been able to develop independently, but through transformation and upgrading, we have achieved 60% to 70% localization.
ChinaThere are many kinds of lithography machines on the mainland, and second-hand is very easy to buy. After building this small-scale production line, Huawei can meet its key needs if SMIC still cannot supply in the future, and in order to prevent the United States Further increase the ban to prohibit Huawei from using US-based equipment to produce chips. Huawei can also try to cooperate with upstream equipment companies (mainland, Japan, South Korea, etc.) at the beginning of the establishment of small-scale production lines. For production or de-beautification, the lithography machine can buy the 28nm of Shanghai Microelectronics, or it can replace the US-made parts of the second-hand lithography machine with domestic parts.
Huawei has other ways. For example, my country’s military chips have always been sanctioned, but there are still companies in China that are producing military chips to meet the needs of military companies. Why can’t these companies produce for Huawei? It is already on the US list anyway.
This is a question of Huawei’s own wishes. It is estimated that Huawei still wants to be a purely commercial company.
Finally note that even if Huawei engages in IDM, it must be a small-scale supply of its own core communication equipment chips. Large-scale production will inevitably rely on SMIC and other domestic foundries because of the amount of investment in the foundries. Too high, Huawei does not need to engage in large-scale.
Finally, I want to say three things:
1: Huawei can definitely survive, it’s just a good life.
2: As analyzed in this article, Huawei has a large number of means (increasing outsourcing chips, promoting TSMC de-beautification, participating in SMIC and domestic equipment de-beautification, developing business that does not rely on chips, and charging patent fees on American companies , Self-built small-scale production lines) can be done to ensure a better survival of itself.
In addition, the national level can use the list of unreliable companies and use the power of my country’s largest market in the world to use the de-beautification tariffs to promote the de-beautification of global foundries to fight back.
3: In June 2020, there are many online voices describing the difficulty of research and development of lithography machines or other semiconductor production equipment. One is that the development time is long and the foundation is weak.
Regarding this point, I would like to say that no matter China’s semiconductor production equipment and raw material enterprises, they are not unprepared, but they have already started research and development and research in many years ago, Eleventh Five-Year Plan (2006-2010) During the twelfth five-year plan period (2011-2015), the 02 special projects are under continuous research and development.
Like 28nm lithography machine has been included in the 13th Five-Year Plan before 2015, and the responsibility division of each subsystem is clear, the technical path is clear, and the current 28nm node lithography machineThe research and development process has been carried out for at least five years.
At the same time, before the 13th Five-Year Plan, China’s 02 project has been tracking the research and development of semiconductor production equipment technology such as lithography machines, and has accumulated a considerable amount. Over the past few years, North China It is an existing fact that domestic semiconductor production equipment manufacturers represented by Zhongkexin, Shengmei and other companies continue to make breakthroughs. Then the progress of the lithography machine is slower and the breakthrough is also expected.
Another important thing is that the technical path of the lithography machine is very clear, not chaotic. You only need to go along this road to get it out. This is the same as the exploratory unmanned area. R&D is different.
One is the long industrial chain involved in lithography machines, with tens of thousands of parts
Every time I see this, I find it very interesting. I haven’t found that ASML has published a BOM, so I don’t know how this quantity came from, but in terms of manufacturing products, it is composed of thousands and tens of thousands of parts The system is normal.
According to the report of China Aerospace News on the Long March V rocket in May 2020, it mentioned:
“Long March 5 carrier rocket is not only bulky, but also new in technology, more complicated than any model of Long March rocket in our country. In the past, Long March rocket used up to tens of thousands of parts, while Long March 5 carrier The rocket uses more than a hundred thousand parts. Its design volume is more than 3.5 times that of the previous Long March rocket.”
I have seen the BOM of a large number of products myself. This algorithm is nothing more than including all the screws, gaskets, capacitors, resistors, inductors, heat insulation sheets, foam, etc. A lot of them are repeated, such as using 100 resistors, that is 100 parts.
That is, tens of thousands of hundreds of thousands of parts seem to be many, but most of them are versatile and can be made in many countries.
The real core technology is the integration of a few key components and systems.
We take Huawei’s flagship machine as an example. The Japanese research institute Fomalhaut Techno Solutions dismantled the Huawei P30 Pro. The mobile phone involved a total of 1631 components, and only 15 parts from the United States accounted for 0.9%. But the value is high, accounting for 16.3%.
On the contrary, although the number of parts in Japan is 869, accounting for 53.2%, in fact, most of them are passive components with very low prices, so the total amount accounts for 23%.
Passive components are not only available in Japan. Samsung, South Korea, Taiwan’s giants, and even developing local companies in the mainland can also serve as suppliers.
That is to say, the de-beautification of Huawei mobile phones only needs to solve a few key parts. Most parts come from multi-national suppliers. Don’t be intimidated by 1631 parts.
Long March rockets, smart phones, lithography machines are similar.
For the lithography machine, even if the output of ASML is hundreds of units in a year, China is engaged in domestic lithography machine, and it can produce dozens of units in a few years. It only needs to solve the problem of completely buying on the market. A few sub-components are not enough. As long as they are available on the market, how can it be impossible to even assemble dozens of lithography machines a year?
From the perspective of the current progress of the lithography machine industry chain enterprises, all the key subsystems are currently in the country, and certain companies are researching and developing, and basically have already made products, which means that the lower layers of these key subsystems There is no problem with material acquisition.
So don’t be intimidated by “tens of thousands of parts”, the building you live in is also tens of thousands of parts, one for each brick and one for each steel bar, the real thing worth noting is “not available The number of “key components”, and according to the market law, the proportion of such components monopolized by a single country is generally very small, because as long as the market is large, other countries will inevitably have enterprises enter the competition.
For Huawei, this road is long-term. Huawei is currently like the Red Army who started the Long March after the anti-encirclement and suppression. From 2021-2022, Huawei will usher in a difficult time, and even in 2022, China will be mass-produced. For Huawei, the 28nm de-beautified production line is just that the Red Army has just arrived in northern Shaanxi. It can only use outsourced domestically produced, MediaTek, STMicroelectronics chips, and HiSilicon chips, which are several generations behind competitors, to participate in international competition.
The next-generation domestic EUV lithography machine was originally planned to be by 2030. Now that this time point can no longer match the actual needs, it must be invested with high intensity to achieve the time node advance, but even if it can be advanced 3 years, it means SMIC won’t get EUV lithography machine mass production below 7nm until 2027Process, to advance to a more advanced 3nm, 2nm or even a higher level of the process will be after 2030,
This means that after passing the darkest moment of 2021-2022, Huawei will only solve the survival problem by 2023. It can continue to produce 28nm Hisilicon chips at SMIC, but it is lagging behind the manufacturing process level of domestic equipment. It takes a long time (after 2030, which is as long as 10 years or more) to re-use the same level of self-developed HiSilicon chips with competitors, even considering outsourcing, MediaTek, and Ziguang Zhanrui. It is also difficult to catch up with Intel, Qualcomm and Apple, and still need HiSilicon’s self-research to play high-end.
I think that for 10 years or more, it is difficult for Huawei, who is not willing to be behind for a long time, to accept. In addition to the investment of national and local funds, Huawei may also build its own small-scale production line. Involving upstream semiconductor production equipment research and development, order-to-order, person-to-person, come off the field and work with SMIC to promote the upstream semiconductor production equipment company’s technology to accelerate progress, avoid it from becoming a long-term development constraint for Huawei, and solve the bottleneck problem.