The year 2023 marks the preliminary establishment of the domestic RF front-end chip landscape, with each niche track set to welcome listed companies.

Tigers and leopards go their separate ways. In the various niche tracks of domestic RF front-end, each will have its own benchmark enterprises, leading companies, and corporate camps. This pattern will last for at least ten years, due to the specialization of expertise, the preference for being the head of a chicken rather than the tail of a phoenix, and the time needed for Chinese RF company leaders to transition from entrepreneurs to true corporate managers.

After the initial establishment of the domestic RF landscape, there is not much time left for startups. Benchmarking against the leading companies in their own tracks and achieving a top three position is the only opportunity and goal for domestic RF startups.

Five tracks, five benchmarks

Successful domestic RF front-end chip companies have all made breakthroughs in a single point, eventually forming five tracks and five benchmarks.

Advertisement

In my opinion, filters represent a large market, but discrete filters are difficult to form tracks and leading companies. The end of RF front-end chips is modules, with receive filters moving towards DiFEM and LFEM, and transmit filters moving towards PAMiD.

Track one: RF switches/LNA, benchmark company: ZTE Corporation

On June 18, 2019, ZTE Corporation went public on the Growth Enterprise Market of the Shenzhen Stock Exchange. Revenue grew from 44 million in 2014 to 560 million RMB in 2018. In 2018, the revenue from discrete RF switches was 460 million, accounting for 82%; LNA revenue was 85 million, accounting for 15%. After going public, the revenue in 2019 reached 1.512 billion, a threefold increase. The revenue in 2022 is expected to be 4.167 billion.

After going public, ZTE Corporation entered all mobile phone brand customers, firmly locking in the position of the first supplier, helping performance to grow rapidly. At the same time, ZTE Corporation moved from discrete switches to DiFEM and LFEM, and performance grew rapidly again. In 2021, ZTE Corporation also launched PAMiF, officially entering the mobile PA market, and will inevitably move towards PAMiD.

In track one of the RF track, under the leadership of ZTE Corporation and the promotion of the industry, it will transform from a discrete switch/LNA track to a DiFEM and LFEM track, and ZTE Corporation will still be the benchmark and leader of this track.Track Two: Phase 2 and Phase 5N PA, Benchmark Enterprise: Weiyechuangxin

China began to popularize 3G networks on April 1, 2008, with three 3G standards: WCDMA, CDMA 2000, and TD-SCDMA.

Weiyechuangxin was established in June 2010, specifically for the 3G PA. By 2014, Weiyechuangxin decisively abandoned 3G PA and fully entered the 4G Phase 2 PA. The products began to enter the market in 2016, and by 2018, sales reached 283 million yuan. Since then, Weiyechuangxin's revenue has entered a fast lane, with 558 million in 2019, 1.786 billion in 2020, and 3.509 billion in 2021.

In 2022, mobile phone shipments plummeted significantly, with Weiyechuangxin's revenue in the first three quarters at about 1.777 billion yuan, and the full-year revenue is expected to be around 2.5 billion, a 30% decline compared to 2021. Despite this, for Phase 2 and Phase 5N PA products, Weiyechuangxin remains the top supplier for domestic mobile phone brand customers.

Weiyechuangxin is also heavily investing in PAMiD and PAMiF products. In the discrete switch area, it used to be a side business, but now it is focusing more on it by investing more manpower and resources, which may be a preliminary preparation and layout for Weiyechuangxin to enter DiFEM and LFEM.

Track Three: PAMiD and PAMiF, Benchmark Enterprise: Hui Zhi Micro

Hui Zhi Micro was established in November 2011 and is one of the earlier companies to accumulate experience in the field of PA module products, forming a series of achievements and advantages. The front-end PA/LNA modules adopt a reconfigurable architecture, with higher integration, fewer wafers, and are conducive to compatibility with larger-sized filters; they have software tuning characteristics, which are helpful for secondary adaptation after integration; and multiple generations of products have continued to accumulate, possessing the packaging and control capabilities of PAMiD.

Hui Zhi Micro is also the first domestic company to mass-produce the 5G frequency band RF front-end integrated transceiver module L-PAMiF, covering the new 5G UHB frequency bands n77/78/79, which are adopted by OPPO and Samsung mobile phones.

Track Four: WiFi FEM, Benchmark Enterprise: Kangxi Communication

The success of Kangxi Communication and Zhuosheng Micro is an unexpected event for China's RF industry.Unexpectedly, within reason. Success is not about the track, the market is not about size, only focus can lead to success. The success of Kangxi and Zhuosheng Micro is reflected in their focus.

Kangxi Communication has been established since 2014 and has only done one thing, focusing on the development of WiFi FEM. After more than four years of entrepreneurship, it achieved a sales volume of 28.57 million yuan in 2019. 2020 was the first year of WiFi6, Kangxi seized the opportunity of WiFi6 FEM and achieved a sales volume of 81.11 million yuan. After proving itself with technology and products, Kangxi achieved a sales volume of 342 million in 2021, a growth of 321% compared to 2021; especially in 2022, when the semiconductor industry was destocking and most RF front-end chip companies' performance declined, it achieved a sales volume of 203 million in the first half of the year, and it is expected to continue to grow throughout the year. In addition, with the rapid rise of the new technical standard - WiFi7, Kangxi has gained the recognition of global mainstream manufacturers in WiFi7 FEM through technological breakthroughs, and the future is promising.

Compared with domestic mobile phone PA companies, Kangxi and Zhuosheng Micro both have a common highlight - high gross profit margin. In 2022, Kangxi's gross profit margin was 30%, and Zhuosheng Micro's gross profit margin was 52%.

Track Five: Base Station PA, Benchmark Enterprise: A Listed Company

Base Station PA, there are already two listed companies and one quasi-listed company in China. Both listed companies are IDM companies, and the quasi-listed company is a fabless design company.

The market for base station PAs is quite volatile, with the domestic market size being around 5 billion yuan in 2021 and 2022, and it will drop to 3 billion yuan in 2023.

In addition to domestic suppliers, there are Japan's Sumitomo for macro base station PAs and the United States' Cree; for micro base station PAs, there are Skyworks and Qorvo.

Technology first, cost is king.

A chip design company is a company based on technology and oriented by the market. Therefore, technology comes first, and cost is the king.

Compared with domestic peers, if a company has neither technical features nor technological leadership, such a company is a worthless chip design company. The result can only be low-price competition and loss-making sales.The domestic RF (Radio Frequency) track is already in a state of intense market competition. While pursuing RF technology, it is also necessary to take into account the relative advantage in cost. Technology is determined by the company's R&D team; whereas the cost is jointly determined by chip design R&D personnel and suppliers, mainly by the suppliers. Generally, the size difference of the chip DIE area among different companies does not exceed 15%, but the price difference given by the same wafer factory can exceed 30%, and even double. Therefore, the chip supply chain determines the fate of chip design companies.

Choice is the key to success or failure. Once a chip design company chooses the wrong supply chain, it will be passive step by step and suffer greatly. The strategic layout of the supply chain is the top priority for chip design companies.

At present, which chip suppliers will determine the fate of domestic RF front-end chips? There are mainly five types of suppliers:

1. Gallium Arsenide (GaAs) wafer factories (PA)

The GaAs wafer foundries selected for domestic RF front-end chips include: Win Semiconductors (WIN), AWSC (Advanced Wireless Semiconductor Co.), San'an Integrated, Lianhe Microelectronics, Fujian United Microelectronics, and Changzhou Chengxin.

Gallium Arsenide is the most critical wafer process for RF front-end chips. The performance, stability, and price of the wafer process are the three most important indicators for RF front-end chip companies, followed by production capacity.

In the next three years, the production capacity of GaAs wafers will be sufficient, which will lead to more obvious positioning of chip design companies.

2. SOI wafer factories (Switch & LNA)

SOI wafers have always been a bottleneck for domestic RF companies. The main SOI wafer suppliers are GF (GlobalFoundries), UMC (United Microelectronics Corporation), and TowerJazz. These three are international giants, and it is difficult for small startup RF companies to get good prices and support. If the price is too bad, these RF startup companies will not be able to get started.

After several years of development, domestic SOI wafer factories have made significant progress. In another two years, SOI wafers will no longer be a bottleneck. Domestic SOI wafer factories now include Ningbo SMIC (Semiconductor Manufacturing International Corporation), Huatong Xianhe, and Wuhan Xinxin.III. Substrate Factories

Radio frequency (RF) substrates have been fully domesticated. The main substrate factories include: Yueya Semiconductor, Shennan Circuits (SCC), Qiding Technology (ZDT), and Xingsen Express.

Yueya Semiconductor's products hold a top three position in the global market share for mobile RF chip packaging substrates. Yueya is the first company in the world to use the "copper pillar build-up method" to achieve mass production of "coreless" packaging substrates. The main products are 3G/4G/5G wireless RF power amplifier front-end modules. Comparative verification has shown that Yueya's Coreless performance is superior when it comes to PA substrates.

Shennan Circuits (SCC) currently has factories in Shenzhen and Wuxi. The process capabilities of Shennan Circuits are comparable to those of ZDT.

IV. Packaging and Testing Factories

As the integration of RF front-end modules becomes higher, the cost of packaging and testing also increases.

The main domestic RF front-end chip packaging and testing factories include: Changdi Technology, Huatian Technology, Tongfu Microelectronics, Jiasen Semiconductor, and Ningbo Yongsi.

The domestic RF front-end chip market has a scale of 30 billion, of which packaging and testing accounts for a scale of about 6 to 8 billion.

V. Filter Factories

In the PAMiD products, the cost of filters accounts for about 60% of the total cost.In the current domestic PAMiD product landscape, the main filter suppliers are Murata, Taiyu, and RF360. To be precise, without the support of filters from Murata, Taiyu, or RF360, the PAMiD products would not be able to be produced.

Each has its own barriers and respects each other

In life, the chain of contempt is ubiquitous. The RF track also has a chain of contempt.

Those who make mobile phone PAs look down on those who make RF switches, thinking that RF switches have no technical content; in addition, the unit price is too low, and sales cannot be made. Unexpectedly, any small chip, as long as it becomes the first in China and the first in the world, is a big business. Therefore, Zhuosheng Micro has been achieved.

Those who make mobile phone PAs also look down on WiFi FEM, thinking that this track is too small, and it can be done by someone casually, and it doesn't matter if it can be done or not. Seeing Kangxi today has both sales and gross margin, and it is about to go public soon, there will be some regrets. Because of the existence of the chain of contempt, Kangxi Communication has been achieved in this way.

Although it is also an RF front-end chip, different products have different technologies and thresholds. No one should look down on anyone when making chips. Different chips each have their own barriers and should respect each other. The following is a brief analysis of their respective barriers:

Barriers of mobile phone PA:

In recent years, domestic mobile phone PA technology has made rapid progress, cultivating a large number of RF R&D talents. In addition to PAMiD, other PA product technologies are very mature, and the competition has shifted from technical competition to cost competition.

For new startups, the barrier lies in the threshold of major customers and the price of the supply chain.

Barriers of discrete switches:The barrier for discrete switches lies in volume. Can we gather a substantial volume, with monthly shipments exceeding 50KKpcs, to have a chance of achieving an annual sales of over 100 million RMB?

If we can't become the primary supplier for one of the companies among Huawei, Xiaomi, OPPO, and Vivo, we won't have the opportunity to achieve sales of over 100 million RMB. Without achieving sales of over 100 million RMB, we won't be able to get good supplier prices.

Barriers for DiFEM and LFEM:

The barriers for DiFEM and LFEM lie in the resources of SAW filters and the threshold of major customers. Next, the competition in cost will be extremely fierce. ZTE has the best supply chain cost advantage and also has the support of major customers, so it will undoubtedly occupy the position of the primary supplier.

There are no fewer than 20 companies entering this track, and the price war is inevitable. Without a good cost foundation, the product will be difficult to sell in the market.

Barriers for WiFi FEM:

Once again, it is emphasized that the chip design for mobile PA and WiFi PA is different, and the technical threshold for WiFi PA is very obvious. Without time investment, it is not possible to do a good job with WiFi FEM. There is no shortcut, it is all about accumulating bit by bit over time.

Compared with mobile PA, the customer threshold for WiFi FEM is not so obvious, and the supply chain cost advantage is even less comparable to mobile PA companies, so the barrier for WiFi FEM lies in technology.

Barriers for Base Station PA:

Now, customers of base station PA are more inclined to choose IDM suppliers, and this barrier has blocked most fabless chip design companies. Unless there is technological innovation or performance breakthrough, it will be very difficult to have another opportunity.The barriers for base station PA customers are the highest among all RF front-end chips, with only three companies involved, two large and one small. Other customers can be disregarded.

Filter Barriers:

Filters represent the future and hope. Making filters is tough and challenging, but it is a necessity.

The barriers for filters are very high, allowing them to overshadow other RF front-end chips. To make filters, one must become an IDM, develop their own EDA, closely integrate design with manufacturing processes, and master packaging technology.

The biggest barrier for filters is patents.

Conclusion

In Bacon's "Of Habits," he states that cognition determines thinking, thinking determines behavior, and behavior determines outcomes.

While starting a business, one must think about how to improve their understanding of the domestic RF front-end chip industry, and how to confirm whether their understanding is correct. This can only be done by comparing subjective cognition with objective things and their laws to see if they match. Practice is the sole criterion for testing the correctness of cognition.

Faced with an emerging landscape, for founders of RF startups, it is no longer enough to just focus on technology. They must also look up and understand the industry landscape, market competition, supply chain, company team, and so on. Repositioning and finding a way out is essential to increase the success rate of entrepreneurship through improved cognition.