Why are 5G chips so expensive?

Recently, MediaTek officially announced that it would subscribe to a total of 19,098,449 common shares issued by local PA leader Visionox at a price of US$2 per share (a total of US$40 million). This is another in-depth expansion of MediaTek in the RF PA field after the acquisition of Airoha.

The move from baseband chips to the RF front-end industry may seem insignificant to others, but it has caused concerns among industry insiders: Will this emerging market be swallowed up by major chip manufacturers one after another?

It seems that following the battle for baseband chips, the war in the RF front-end market has also been quietly provoked.

From edge to core, RF front-end modules may become a shortcoming in the development of 5G industry

Speaking of the RF front-end module, it is not important because it is a peripheral module of the main chip - the baseband chip, and does not undertake important signal processing and computing functions.

From the existing product form, it is mainly composed of a series of components such as RF antenna, antenna switch, filter, duplexer, etc., which is essentially a modem responsible for sending and receiving device signals. Because it is strongly related to signal transmission, changes in signal frequency and transmission channel will have a significant impact on it.

As we all know, compared with 4G, the most obvious difference of 5G is that it not only supports low frequency bands below 6GHz, but can also extend to the millimeter wave band of 26.5-300GHz. This change not only solves the problem of extremely scarce bandwidth resources, but also temporarily reduces the worries of industrial development. Therefore, it can be said that the increase in millimeter wave signal bands is the highlight of 5G. And because of the increase in the high-frequency band signal transmission channel, the status of the RF front-end module in the product has also changed.

Adding millimeter wave frequency band to the original band means that for hardware, it is necessary to integrate more modules for processing high-frequency signals on a chip of the same size. Therefore, the design focus of the RF front-end module falls on millimeter wave signal processing technology, which originally belonged to the field of military high-performance chips.

In fact, the changes brought about by the addition of millimeter wave signals are significant and complex. The most intuitive feeling on the RF front-end module is the increase in the number of components. It is estimated that in order to add new frequency band communication functions, from 4G to 5G, the number of core device filters required for each mobile phone will increase from 40 to 50 on average. At the same time, in order to achieve the change from dual channels to 4 channels, the number of power amplifiers will also double accordingly; in addition, the number of 5G terminal switches will increase to two to three times the original number, and the number of antennas will also increase exponentially.

Due to the increase in quantity, after the existing process, it is bound to increase the cost of chips related to 5G signal processing. Previously, HIS, an internationally renowned research institution, released a report to interpret this point after disassembling several 5G mobile phones, in which it specifically analyzed the trend of RF front-end design. IHS said that due to the pursuit of full network access and LTE network speed in recent years, a terminal often needs to support multiple frequency bands. The increase in frequency bands directly leads to the increase in the complexity of RF front-end design, and hundreds of components must be accommodated in a small space. Now with the advent of gigabit networks, the integration of technologies such as multi-carrier, high-order modulation, and 4x4 MIMO has increased the complexity of front-end design. By disassembling the Samsung S8, IHS confirmed its idea. It saw that it adopted the most complex front-end design at the time, and the increase in complexity meant a direct increase in cost, which also made the front-end module occupy an increasingly high proportion in the BOM cost of mobile phones, and its importance has therefore increased.

JPMorgan Chase has released a report showing that the cost of 5G mobile phone chips will be 1.85 times more expensive than similar 4G mobile phone products. Specifically, the internal chip of a 4G mobile phone is expected to cost about US$59 (about RMB 397). Based on this price, the chip cost of a 5G mobile phone is about RMB 1,091. What is incredible is that the chip cost alone has exceeded many low-end phones on the market, and the contribution of the front-end module is naturally indispensable.

High costs put pressure on manufacturers, and structural changes in the industry are urgent

According to Moore's Law, the cost of memory and baseband chips inside mobile phones can be reduced through iteration of processes and technologies, but this does not apply to the RF front-end module industry. This is why we can still see the eye-catching device design on the circuit board with the naked eye.

The cost of chip manufacturing cannot be reduced through process, and the radiation affects the complexity of the entire circuit board design, which brings resistance to the development of 5G terminal equipment, which makes RF front-end module manufacturers face industry pressure. As several industry insiders unanimously agree, the high integration of communication modules is the general trend, and the existing process cannot support the development of the industry, so they are also facing the crisis of competing with the main chip manufacturers for the market.

This has already been reflected in the current changes in the situation.

Previously, after more than a decade of development in the 4G era, the entire terminal RF front-end market has actually formed an oligopolistic competition pattern dominated by three companies: Qorvo, Broadcom (Avago). Statistics show that in the terminal RF front-end market, the three companies together account for more than 90% of the market share, and their gross profit margins are all above 40%.

But just as every communication upgrade will bring shocks to the industrial landscape, this time 5G's impact on the terminal RF module industry is systematic and comprehensive: in order to adapt to the consumer market, the costs of many originally expensive millimeter-wave devices have to be reduced as they move from high-performance scenarios to consumer electronics. At the same time, as the number of devices increases, materials and processes will undergo a certain degree of change.

In response to this opportunity for industrial transformation and in order to seize the potential growth market, companies similar to MediaTek's layout in the RF field, such as Qualcomm and Tsinghua Unigroup, which focus on the main chip field, have also begun to seize the 5G RF module market and try to open up part of the market by acquiring RF module manufacturers. For example, Qualcomm and TDK jointly established RF360, MediaTek acquired Airoha, and Tsinghua Unigroup has RDA.

Compared with low-frequency integrated circuits, the development of millimeter-wave integrated circuits has not been so ideal. Although millimeter-wave integrated circuits have also experienced the development path from discrete devices, hybrid integrated circuits to monolithic integrated circuits, due to physical limitations such as the easy interference of high-frequency signals, it is difficult to integrate simple-function millimeter-wave devices into ultra-large-scale integrated circuits. Therefore, it brings challenges to all aspects of terminal engineering such as materials, appearance size, industrial design, heat dissipation and radiation power regulatory requirements, and also brings opportunities for large manufacturers to enter at this moment, which is bound to induce a major "earthquake" in the entire industrial chain.

Historical legacy, technical difficulties of MMIC remain to be solved

Back to the problem of resistance to the development of 5G RF front-end modules, we find that the essence is that the millimeter wave module cannot be highly integrated into the mobile phone chip. It has to be said that although the technology and process development of the electronics industry has been over for a century, the problem of highly integrated millimeter wave devices has not been solved, so it has gradually developed into a sub-industry, namely MMIC, which is well known in the industry.

The benefits of achieving high integration are obvious. Thin monolithic integrated circuits have a series of advantages such as low circuit loss, low noise, wide bandwidth, large dynamic range, high power, and high added power. They can also reduce the size of electronic equipment, reduce weight, and reduce prices, which is very important for military electronic equipment and civilian electronic products.

However, this has always been a difficult point in the development of industrial technology. As early as 1986, the US Department of Defense listed MMIC as one of the military microelectronics programs. Under the leadership of DARPA, it adopted a policy of huge federal government support, mobilized the strength of universities and major companies in the industrial sector across the country, and carried out extensive and in-depth research in the field of MMIC through division of labor and cooperation. Data shows that the US federal government invested a total of 530 million yuan at that time, and with the investment of the US industrial sector, it actually exceeded 1 billion US dollars. But even so, the results were minimal.

After entering the 1990s, with the end of the Cold War, the application of MMIC in civilian fields has been growing at an annual rate of 15-20%, but the industry has not yet reached a high degree of consensus. In terms of materials alone, millimeter wave chips have various materials such as gallium arsenide (GaAs), InP (indium phosphide), gallium nitride (GaN) and silicon-based (CMOS, SiGe), and major manufacturers in the RF field are still exploring more suitable materials and processes.

From the perspective of the entire market, GaAs process has become the mainstream process for microwave and millimeter wave integrated circuits. Due to its higher electron mobility, carrier saturation drift velocity and high breakdown field strength, GaN is unanimously considered to be the material of choice for future RF devices.

However, just as the development of integrated circuits began, the integration of all devices requires the use of a unified process. Considering the integration level and market development, as well as the current background of the rise of silicon integrated circuits, although the performance of the millimeter wave band is insufficient, silicon-based technology is still the technology most likely to be commercialized in the industry.

It is worth mentioning that the automotive millimeter-wave chips developed based on CMOS technology have brought a breath of fresh air to the MMIC industry, which has given more confidence to manufacturers who are determined to integrate millimeter-wave devices based on silicon wafers in the 5G communication field.

The huge advantages of silicon technology in terms of cost and integration are attractive, and its application in digital circuits is profound and extensive. Japan, the United States, Canada and other countries are constantly conducting research in this area. Our country’s Southeast University National Key Laboratory of Millimeter Waves insists on silicon-based millimeter wave chips, and its successful development is bound to greatly reduce the cost of 5G chips and will greatly promote the development of the 5G industry.

So far, the main application area of ​​millimeter wave chips is still high-end military scenarios. Therefore, it can be said that the "many requirements" brought by 5G commercial use in terms of cost, size, etc. make its own promotion and development full of pressure.

At present, no chip manufacturer has a good solution in the field of 5G terminal RF front-end integration, so there are still opportunities for start-ups. However, the commercial development of 5G is approaching, and there are news that the cost of millimeter-wave chips has been greatly reduced (for example, Purple Mountain Laboratory claims that the millimeter-wave chips it develops can be reduced to more than 20 yuan), which indicates that the industry situation is about to change.