In recent years, "cross-industry chip making" has long become a hot topic in the tech industry.

Whether it is car companies, mobile phone manufacturers, internet companies, or home appliance giants, they have all successively plunged into the "chip making" track, and "self-developed chips" seem to have become an irresistible trend.

In fact, as early as 2010, when Apple released the iPhone 4 and clearly announced to the outside world that it had developed its own processor A4, "chip making" was no longer the exclusive domain of chip companies. Companies such as Google, Tesla, Microsoft, Amazon, Baidu, and Alibaba have successively entered the field across industries.

Advertisement

Especially in recent years, as user needs have continued to increase, and the semiconductor industry has experienced a "chip shortage" and "technology sanctions," "cross-industry chip making" has become a trend.

Giants cross-industry chip making, have poured in

In August last year, Kuaishou launched its first self-developed cloud-side intelligent video processing SoC chip SL200, officially announcing its entry into the B2B market; at the same time, its old rival ByteDance is also striding forward in the field of chips, revealing the latest progress in self-developed chips divided into three major categories: server chips, AI chips, and video cloud chips, and it was also exposed to use three times the industry's salary to dig chip talents.

In the mobile phone market, mobile phone manufacturers are also very lively in the chip-making track. The four major domestic mobile phone brands "Huawei, Xiaomi, OPPO, and vivo" gather together. Xiaomi launched the ISP chip Surge C1 and the charging chip Surge P1; vivo announced its self-developed ISP chip V1/V2; OPPO launched its first self-developed imaging-specific NPU chip MariSilicon X, and it was revealed that OPPO has more chips on the way.

Mobile phone manufacturers entering the chip research and development field are basically based on self-development, and the focus is mostly on improving imaging capabilities. Among them, although OPPO's MariSilicon X is an NPU chip, it actually also integrates ISP, which is an NPU with ISP + AI accelerator.

Compared with the all-in-one mobile phone SoC chip, the ISP that only focuses on imaging is relatively not so technically complex. More importantly, in the current era where imaging has become the core competitiveness of mobile phones, how to effectively improve imaging performance has become the main direction of "involution" for mobile phone manufacturers.On the other side, it is no longer a fresh topic that car companies are making chips.

After experiencing the shortage of chips and the impact of giants developing their own chips and the transformation of automotive architecture on the new demand for semiconductors, car companies have begun to feel the necessity to design chips or participate in chip design.

As a local representative of the new force in car manufacturing, NIO, XPeng, and Li Auto have taken up the banner of "self-developed chips".

As early as 2020, it was reported that NIO and XPeng, two new forces, were building their own autonomous driving (AD) chip development teams. By last November, the R&D team of NIO's AD chip had reached a scale of 500 people, and the research and development was progressing smoothly. There are also reports that NIO is developing AD chips and LiDAR chips at the same time; XPeng has chosen Tesla's FSD chip as its benchmark product.

In May last year, Ideal Intelligent Motion, wholly owned by Li Auto, was officially established, which seems to declare Li Auto's determination to enter the chip industry. At the beginning of this year, Li Auto seems to be researching autonomous driving chips, and its official website and third-party recruitment platforms are recruiting several positions such as SoC system architects and senior engineers for autonomous driving MCU development.

At the same time, all the new forces are also trying their best to attract industry talents to promote chip research and development. The industry estimates that the autonomous driving chips of NIO, XPeng, and Li Auto may be on the car as early as around 2024.

From competing for talents to competing for investment, the chip competition of the new forces in car manufacturing has entered a white-hot stage, but there is still a distance from the light at the end of the tunnel. Taking Tesla as a reference, "NIO, XPeng, and Li Auto" still have a lot of homework to make up.

In addition to NIO, XPeng, and Li Auto, Chinese car companies such as Geely, Leapmotor, and BYD are also developing autonomous driving chips and power semiconductor devices.

Considering the long R&D and mass production cycle of chips, high technical threshold, and the need for continuous and stable business rhythm, car companies making chips is a vertical expansion of business, and the more important main business of car companies is to make and sell cars. These attempts and efforts depend on how many resources the car companies have, patience in management and execution, and how many cars they can sell and how long they can survive.

In addition, capital cooperation has also become another path chosen by car companies. Since 2021, Chinese car companies have also started to invest in the chip industry on a large scale to layout their own automotive chip supply chain.Automotive companies are collaborating with semiconductor enterprises, with the advantage that domestic automotive-grade chips have gained more opportunities for vehicle certification. Engaging in vertical capital cooperation with industrial chain enterprises is almost one of the best choices that domestic automotive companies can make to improve their chip industry chain and supply chain at this stage.

The self-developed chip layout of home appliance giants is inseparable from their business. According to incomplete statistics, well-known home appliance brands such as Midea, Gree, Hisense, Haier, Galanz, TCL, Skyworth, Konka, and Changhong have all actively laid out in the field of chip semiconductors through setting up chip departments, establishing subsidiaries, or investing in chip startups. The layout mainly focuses on various types of home appliance chips such as MCU control chips, power management chips, connection chips, driving chips, and image processing chips.

The new generation is eager to try, while the veterans have already established a solid position.

Starting from Baidu's release of the Kunlun 1 chip in 2018, old internet giants such as Alibaba and Tencent have successively entered the chip field. Alibaba established the chip company PingTouGe Semiconductor, and Tencent entered the field in the form of "investment + self-development."

In the ups and downs of the times that followed, BAT has withdrawn many business lines, but still firmly held the chip battlefield.

Baidu's release of the first domestic cloud AI chip, Kunlun 1, opened the precedent for internet giants to make chips. At present, Baidu is fully deploying autonomous driving. Last year, when the Kunlun 2 chip was released, it was determined to be used in the fields of autonomous driving and intelligent transportation assistants.

So far, Alibaba PingTouGe has formed four major series: Xuan Tie, Yi Tian, Han Guang, and Yu Zhen, building an external ecosystem and serving cloud computing internally.

At the end of last year, Tencent publicly disclosed the progress of self-developed chip research and development for the first time - the three chips "Zi Xiao," "Cang Hai," and "Xuan Ling" are all special-purpose chips, used for AI computing, video processing, and high-performance intelligent networking.

In the matter of making chips, Huawei Hisilicon is also a veteran. In addition to the Kirin series known for smartphones, there are baseband chips represented by the Balong series, communication base station chips represented by the Tiangang series, AI chips represented by the Ascend series, and PC and server chips represented by the Kunpeng series. It was later affected by US sanctions, which everyone is more clear about.Overall, industry giants are all targeting the chip track, creating a kind of illusion that "not making chips is out of date". However, looking at the market environment and trends, although the "path to making chips" is fraught with dangers, not doing so seems more like "boiling frogs in lukewarm water", being caught off guard by crises.

Giants making chips have reached a new stage.

The giants' cross-industry chip-making cases mentioned above, such as BAT and Huawei, are already well-known; the semiconductor industry observation also introduced the chip self-research progress of local new forces in the previous article "The chip layout of 'Weilai, Xiaopeng, and Li Auto'"; the chip-making path of car companies, mobile phone manufacturers, and internet giants is also included in "Is the end of technology 'making chips'?"... No more elaboration is needed.

In addition to these, the giants' chip-making is also ushering in a new stage.

Hisense Chip Company plans to spin off and go public.

Recently, Hisense Visual released a plan to spin off its subsidiary Qingdao Hisense Xinxin Microelectronics Technology Co., Ltd. to the Science and Technology Innovation Board for listing.

It is understood that Hisense invested 500 million yuan in 2005 to establish Qingdao Hisense Xinxin Technology Co., Ltd., and in the same year, it developed China's first digital video processing chip with independent intellectual property rights, "Xinxin HiView" VPE1X, which reduced the price of similar imported chips from $13 per piece to $5.

In 2017, Hisense acquired Japan's Toshiba TV and integrated its picture quality chip design team. In June 2019, the chip department Hisense Xinxin and Shanghai Hongyou Company were integrated to establish Qingdao Hisense Xinxin Microelectronics Company. To date, its picture quality processing chip has been iterated 4 times. It can be seen that Hisense's chip path mainly revolves around optimizing display effects and has achieved multiple "first in the country", including the launch of a 4K 120HZ ultra-high-definition picture quality engine chip in 2015, and the release of a fully self-researched 8K AI picture quality chip in January 2022, and so on.

Public information shows that Xinxin Micro is a fabless chip design company focusing on display chips and AIoT intelligent control chips, committed to providing display chip solutions for various display panels and display terminals, and providing frequency conversion control and main control solutions for smart home appliances.

Hisense Visual said that through this spin-off, Xinxin Micro will achieve independent listing, enhance capital strength through listing financing, and improve the company's continuous profitability and core competitiveness.On this basis, Hisense Visual Technology also stated in its annual report that the company will continue to expand its semiconductor layout, optimize the display industry through vertical integration, and accelerate the horizontal expansion efficiency of chips.

BYD Semiconductor: Unwavering Commitment to the Listing Plan

As early as 2013, Elon Musk proposed the development of autonomous driving chips. Due to the lack of technology and talent reserves, Tesla could only cooperate with Mobileye in the early stage, and the products developed did not meet expectations, only reaching Level 2. From 2015, when Tesla reorganized its team to layout autonomous driving chips, to 2019, when Tesla officially released its first self-developed AP chip, Autopilot HW3.0, Tesla experienced a five-year journey.

Similarly, BYD Semiconductor, a leading enterprise in the domestic automotive chip field, has also gone through a long road of technology cultivation and vehicle verification.

In terms of market and capital, compared to the continuous rise in the sales of its electric vehicles, the IPO journey of BYD Semiconductor, a subsidiary of BYD, has been full of twists and turns. After several "suspensions," it rang the bell of termination again in November last year.

However, at BYD's performance meeting on March 29, Chairman Wang Chuanfu once again stated, "BYD Semiconductor's listing plan remains unchanged, but there are some adjustments in the process."

In fact, raising funds through listing to invest in power semiconductors was originally one of BYD's plans. According to the previous plan of BYD Semiconductor, the company's listing will raise 2.686 billion yuan, of which 312 million yuan will be used for the industrialization and upgrading of new-type power semiconductor chips, 2.074 billion yuan for the research and development and industrialization of power semiconductors and intelligent control devices, and 300 million yuan for supplementary working capital.

In addition, facing the continuous growth of the new energy automotive industry, the newly added wafer capacity is still far from meeting downstream demand. In order to improve production capacity supply and independent control capabilities as soon as possible, BYD Semiconductor plans to seize the time window and carry out large-scale wafer capacity investment and construction. To expand wafer capacity, BYD Semiconductor invested about 4.9 billion yuan during the listing review period to implement the Jinan power semiconductor capacity construction project. The project has now been completed and put into production, with smooth capacity ramp-up. It is expected to reach full production status in March 2023, with a capacity of 30,000 wafers per month, which is expected to have a significant impact on the future asset and business structure of BYD Semiconductor.

Furthermore, at the time of the termination of the listing, BYD Semiconductor also took over the Chengdu Unisplendour project, which was also interpreted by the industry as a step to expand wafer capacity. By optimizing the company's semiconductor business layout, expanding capacity to meet domestic demand first, and then supplying the entire industry chain, reducing the proportion of related transactions with the parent company, and ultimately achieving the purpose of listing.As a semiconductor supplier, BYD Semiconductor has now achieved mass production of products such as IGBT, SiC devices, IPM, MCU, CMOS image sensors, electromagnetic sensors, LED light sources, and displays. Among them, BYD Semiconductor is best known for its IGBT automotive power modules, which began self-research as early as 2005 and currently has the operational capability of a full industry chain IDM model.

Currently, in the field of IGBT modules, BYD Semiconductor has also taken a leading position in the domestic market. Data shows that in 2018, BYD Semiconductor released the benchmarking IGBT 4.0 technology in the automotive-grade field; in 2021, the IGBT 5.0 technology based on high-density Trench FS was mass-produced. At present, in the field of IGBT modules, BYD Semiconductor has also taken a leading position in the domestic market.

According to the prospectus, in the field of IGBT, BYD Semiconductor ranked second globally and first among domestic manufacturers in the new energy passenger car motor drive manufacturers for two consecutive years in 2019 and 2020, with a market share of 19%, second only to Infineon. In addition, according to the research report of Caitong Securities, in the first three quarters of 2022, BYD Semiconductor's power module installation volume market share reached 21.1%, close to the market share of 25.7% of the leader in the track, Infineon.

Should we self-develop or purchase chips, is it still "better to buy than to make"?

The debate on whether to purchase core chips or self-research in China has a long history.

The chip industry thinking of "better to buy than to make" has led to the fact that the gap in the domestic semiconductor industry has gradually widened. Especially with the occurrence of the ZTE and Huawei incidents, the discussion among domestic manufacturers on whether to self-develop or purchase chips has gradually shifted. Not only are local chip companies increasing their self-research efforts and actively improving competitiveness, but terminal giants have also begun to exert themselves upstream in the industrial chain, and "cross-border chip making" has quietly risen.

For the above-mentioned giants to cross-border chip making, making chips requires specific scenarios to support, not for the sake of making, but based on internal needs such as short videos, image quality, cloud computing, or autonomous driving. Of course, no matter which scenarios, most of the giants make chips for two purposes: to serve their main business internally, "cost-effectiveness" is the driving force for making chips, while achieving their own differentiated competitiveness in the same track; to explore market space externally and find new growth points.

To meet its own business needs and enhance competitiveness, the former is the most obvious trend among industry manufacturers.

From the perspective of international manufacturers, as early as 2013, Google began to develop TPU chips for AI scenarios to solve the company's increasingly huge computing needs and cost issues. Amazon also launched the Nitro1 chip in 2013, also serving its own business. To this day, Amazon has three product lines of network chips, server chips, and artificial intelligence machine learning chips, and has turned out 9 cores in 8 years. Amazon uses self-developed chips to process the calculations of the Alexa voice assistant, successfully replacing Nvidia's chips and reducing costs by 30%. Google also released its self-developed video processing chip Argos VCU, replacing tens of millions of Intel CPUs, saving Google 2 billion yuan in capital expenditure at one go.Tech giants such as Amazon, Microsoft, and Google control the majority of the world's server computing power. The demand for this immense computing power is enough to motivate them to develop their own server chips. Therefore, cost-effectiveness is the primary driving force for large companies to create chips.

This trend of chip-making has spread to China, with major internet companies joining the conversation one after another.

For example, Baidu has two main chips, "Kunlun" and "Honghu". Kunlun, as an AI chip, can not only meet the cloud-side needs of commonly used deep learning algorithms but also adapt to the computing requirements of specific terminal scenarios such as natural language processing, large-scale voice recognition, autonomous driving, and large-scale recommendations. Honghu is a far-field voice interaction chip, mainly used in scenarios such as in-vehicle voice interaction and smart home, all of which are in synergy with Baidu's business.

Similarly, the chips of Alibaba's Pingtou Brother also serve the construction of Alibaba Cloud's ecosystem to a large extent. In addition, based on the Shenlong architecture, it has launched its own cloud server, Shenlong server, and designed its own intelligent network card chip, X-Dragon. In short, by continuously enriching the ecosystem, it enhances its integrated collaboration capabilities of software and hardware.

Tencent's "Canghai," "Zixiao," and "Xuanling" are all aimed at internal needs, including video processing acceleration, AI chips, and intelligent network card chips.

From the chip-making plans of Kuaishou and ByteDance, we can also see that as the business of internet companies expands rapidly, the hardware requirements for the upper-level video business need to be based on actual needs from the definition stage of the chip, to make ASICs that conform to the characteristics of their own business. This is also what the person in charge of ByteDance mentioned, "The company cannot find a supplier that can meet its requirements."

If existing chips on the market are used, it will inevitably have to choose from the mature chips of established giants, which will undoubtedly increase the investment in hardware and the risk of subsequent business stability. Moreover, general-purpose chips on the market are often not as good as special-purpose chips in optimizing the video experience. In this case, self-developed chips have become a necessary option.

On the other hand, general products provided by chip giants such as Qualcomm, AMD, and Intel are increasingly difficult to meet the actual needs of internet technology manufacturers. It also leads to the cost of purchasing chips by many technology companies rising continuously. In the past, due to weak discourse power, they could only go along with it; but now, with the emergence of self-developed chips from various companies, they have some confidence in the "arrogance" of the giants. Under various factors, self-developed chips are gradually becoming a necessary option.

It is not difficult to understand why home appliance giants are developing their own chips. On the one hand, under the national environment's requirement for chip independence and controllability, they do not want their lifeline to be in the hands of foreign capital, so they are determined to develop their own chips and enhance their core competitiveness; on the other hand, self-developed chips are more valuable for terminal companies in enhancing product differentiation, better hardware and software adaptation, and performance optimization.

In terms of creating differentiated advantages, although mobile phone companies such as OPPO, vivo, and Xiaomi will not face the dilemma of being strangled, there must be a problem of chip homogeneity. To generate technical competition, self-developed chips are also a necessary option.However, there is a diversity of opinions in the industry regarding the necessity for car companies to develop their own chips. Some argue that with the rapid penetration of electric vehicles and the relative maturity of the intelligent electric vehicle supply chain, companies specializing in chip production, such as Nvidia and Qualcomm, can provide most car companies with products like autonomous driving and cockpit chips. Compared to most car companies, they have a greater scale advantage, which allows them to spread more R&D costs through huge shipments, and get a lower single-chip quote when cooperating with foundries like TSMC.

On the contrary, car companies developing high-computing power chips bear the disadvantages of the above scale advantage. Car companies need to bear extremely high R&D costs. If the chips are used entirely by themselves, spread over each car, self-developed chips may not necessarily be cheaper than purchased ones.

However, before the question of whether it is necessary to make chips and whether the cost is appropriate, leading car companies with resources and ambitions will still try to develop high-computing power chips such as intelligent cockpits and autonomous driving.

This is because chips and the intelligent experience based on chips are increasingly becoming one of the competitive strengths of car companies. Autonomous driving and intelligent cockpit chips support the software system, and data flows in the software, which is the necessary content for car companies to perceive user behavior and iterate the system. Those car companies that hope to bring extraordinary experience and brand stickiness like Apple or Tesla through continuous upgrades of software and hardware systems cannot help but try to develop their own chips and operating systems.

Therefore, car companies, including NIO, XPeng, and Li Auto, all have this ambition. Firstly, the independent and controllable high-end chips have always been driven by policies, and supporting the local semiconductor industry is a trend and inevitable; secondly, developing self-developed AD chips not only helps car companies build a full-stack self-developed capability but also enhances brand value to a certain extent.

Under the market environment, stabilizing supply chain security.

The Sino-US trade friction, which leads to difficulties in supply from overseas manufacturers, has accelerated the "domestic substitution" trend in the semiconductor industry.

As mentioned above, whether from the perspective of the company's own strategic level or the overall industry environment, "making chips" is worth trying for technology-driven large companies.

The advantage lies in that, compared to the high threshold of general-purpose chips, the R&D of specialized chip design is relatively much easier for large companies. And it can enhance its core advantages of product experience and service differentiation.Additionally, independent research and development can enable it to achieve optimal cost and process efficiency, enhance security and flexibility, and innovate faster at various levels in the long, medium, and short term, thus taking the initiative in the initiation, progress, and delivery of chips.

Expanding market space and seeking new growth points

Finally, let's discuss from the perspective of "giant enterprises making chips may explore external market space to find new growth points."

Looking at the development process of the semiconductor industry over the past few decades, it is closely related to the trends of the hardware market.

In terms of foreign companies, Intel's rise accompanied the prosperity of the PC industry, Qualcomm's development relied on the expansion of the smartphone market, and Nvidia's later rise was also due to the favorable winds of gaming, cryptocurrency, and AI.

So for the current domestic giants, the most important thing is actually to find their own application scenarios, whether it is cloud computing, AI training, or MCU, whether there is competitiveness, and whether it can fight, all need to be tested in the market.

Taking Huawei Hisilicon as an example, its first order came from Dahua's security monitoring market, and then relying on Huawei's foundation in the communication field, it developed base stations and baseband chips, followed by the well-known Kirin series, relying on Huawei's smartphones to reach countless consumers.

However, looking at the progress of the rising stars, whether it is Baidu or Alibaba, or other cross-industry giants, it is not easy to open up the external market. And spinning off from large factories and independent financing may be a necessary path for the marketization of their chip business.

On the one hand, spinning off and going public can expand financing channels, reduce asset-liability ratios, and help optimize capital structure, and introduce market-oriented incentive mechanisms; on the other hand, after spinning off from its own business, it can also be more convenient to supply to other peers and supply to customers in more fields, thereby quickly expanding market share and finding new growth points.

In conclusionFrom the current process of major manufacturers entering the chip-making industry, it can be seen that the "outlets" of the current giants in chip-making are mostly precisely aimed at their own businesses, and there are signs and cases of seeking to split and go public independently. However, looking at the chip-making itself and the competitive environment it is in, the high threshold of "making chips" determines that the major manufacturers entering the "chip-making" industry will find it difficult to avoid a protracted war in the subsequent process.

But no matter what, the development logic of "it's better to buy than to make" in the chip industry is becoming a thing of the past, left in the "creak" ruts of history.