Text/"Autobot" Huang
In the consumer market, automotive chips are less concerned than other automotive parts, and the former is usually not within the scope of after-sales maintenance and replacement. However, the "lack of core" of the north and south masses has appropriately ignited public interest. Hundreds of billions of companies will also be forced to stop production because of gadgets such as chips.
After Huawei was sanctioned, the chip once again entered the public eye. Only this time, it's just a local application of the chip.
Have fallen behind
In 20 19, the global semiconductor business scale was about 500 billion dollars, and the automobile chips were 47.5 billion dollars, accounting for less than 10%. However, with the "new four modernizations" becoming a definite trend, software and chips quickly become the value-added part of automobile value. From 202 1, the annual increase is at least 20%.
The scale of China's automobile chip industry is less than 654.38+05 billion yuan, accounting for 4.5% of the global share; Cars in China account for 30% of the global business. In contrast, automobile chips lag behind the whole vehicle business in scale, technology, brand and category.
Generally speaking, China chip enterprises still have a long way to go before they enter the automobile supply chain on a large scale. But things are different here.
If roughly divided, automobile chips can be roughly divided into three categories: the first category is responsible for computing power, specifically processor and controller chips, such as central control, ADAS and automatic driving system, as well as engine, chassis and body control; The second type is responsible for power conversion, which is used for power supply and interface. IGBT is used for EV power chip. The third category is sensors, which are used in various radars, airbags, tire pressure detection and the like. Of course, there are some miscellaneous items, which are ignored here.
Globally, four European companies, NXP, Infineon, STFT and Bosch, account for 36.2% of the total. Among American enterprises, Ti Yi and Amway +ADI (the former acquired the latter, and now they are one) account for15.7%; Japanese Renesas and Toshiba account for 14.3%. The above 8 suppliers account for more than 2/3 of the total.
In fact, some China enterprises have explored ways to enter the supply chain. Driven by extensive social knowledge and huge investment, it is only a matter of time before the proportion of China enterprises' automobile chip business equals or even surpasses the whole vehicle business.
Demand traction
As mentioned above, it is difficult for China enterprises to enter the supply chain. The technical threshold is not an isolated factor, because the abnormal stability requirement (0PPM) of automotive chips makes Tier 1 suppliers who are responsible for purchasing chips reluctant to let new players practice. If it is not an established enterprise (in fact, the rise of the chip industry was after the 1980 s), there is no chance to be a spare tire.
The problem is that it is impossible to iterate technology and keep up with the development of technology without coming in. The more you can't keep up, the more you can't get in. This is an infinite loop.
How to break the game? BYD provides an idea, that is, demand traction. Aren't the automakers and Tier 1 suppliers not allowed to enter? BYD itself is a vehicle manufacturer, Tier 1 supplier. If it places an order for itself as a customer, won't the demand problem be solved?
In the composition of new energy cost, IGBT ranks second. Like other chips, IGBT is cut and etched from the wafer. BYD has an existing production capacity of 50,000 vehicles/month, and will double its production capacity next year, which can meet the demand of10.2 million new energy vehicles. BYD 10 billion yuan Changsha wafer production line has started, with an annual output of 250,000 8-inch wafers. Like batteries, BYD is striving to be a supplier of other car companies in addition to its own production and sales.
The big players of IGBT are Infineon, Fuji Electric, Anson and Mitsubishi. In 20 19, Infineon supplied 628,000 sets of IGBT modules in China, accounting for 58% of the market. BYD has194,000 units, accounting for 18% of the market.
Byd used IGBT independently developed and produced on Qin and Tang models, but it was not announced until 20 18. By the end of 20 18, BYD had 175 patents in IGBT.
Infineon has an overwhelming advantage in the application field of automobile 600- 1200V; In Japan, Mitsubishi and Fuji Electric have divided the market, and the IGBT used in Toyota's hybrid vehicles is completed internally.
In China, Starr Semiconductor (a listed company) and IGBT technology are more advanced and have developed to the sixth generation (BYD is the fifth generation). Last year, Starr supplied 1.7 million units, accounting for 1.6% of the market, far less than BYD, which shows the traction function of the main engine factory.
Globally, IGBT has developed to 7.5 generations (Mitsubishi), and there is a big gap between people and enterprises in China, but this is already a relatively successful sub-field for China enterprises to do chip business. As long as the demand side continues to exist, it is only a matter of time before it catches up with the advanced level.
Brand new track
Among automobile chips, computing chip is a "new track". With the rise of the concept of smart car, the demand for on-board computing power has surged. New demand promotes new investment, and the old revolution encounters new problems.
The eight major suppliers of automotive chips have no outstanding advantages in this field. There is no monopoly brand in this subdivision, and all manufacturers are in the testing stage. Chip companies such as Samsung, NVIDIA, Intel, Xilinx and Qualcomm have set foot in the field of on-board computing power.
Traditionally, OEMs do not directly grasp chips, except computing chips. Tesla, Weilai, Tucki and Uber (which just sold the ADAS business) have all invested in the research and development of computing chips. Of course, they are chip designers at most and must find another foundry.
Even so, it broke the dilemma that the host manufacturer only mentioned the demand and task book. The general CPU control chip can't meet the requirements of ADAS, prompting OEMs (mostly new forces) to design special chips for autonomous driving.
Although OEMs have taken risks in developing their own chips, investors like to hear such news. Moreover, OEMs don't want a large part of profits to fall into the pockets of chip suppliers in the future.
However, similar investments need to be used on a large scale to make profits. If we spend $65.438+$50 million as the design fee of special ASIC chips, the annual output will be 400,000 pieces, and the investment will be recovered within four years. At present, only Tesla is expected to reach this passing line.
The remaining opportunities belong to suppliers. Although Huawei has repeatedly reiterated that it will not build cars, it will not reassure OEMs because it has always been a provider of package solutions: cloud, 5G infrastructure, Internet of Things, in-vehicle operating system and V2X communication chips. If computing chips are added, Huawei will have enough ambition to become a brand-new Tier 1 supplier.
As early as 20 18 and 10, Huawei released MDC6000, a computing platform supporting L4 autopilot capability. The platform has eight AI chips from Huawei, including 3 10, and calculation and image processing modules. Audi China became the first important customer of Huawei's ADAS business.
Huawei claims that the computing power of MDC6000 is 352? Top, power consumption ratio 1? Tops/ watt, surpassing NVIDIA.
It needs to be clarified that supporting L4 autopilot does not mean realizing L4 autopilot. Huawei only provides a computing platform, and the software implementation is completed by customers or designated third parties. And it takes a long time to certify vehicle regulations.
Tesla FSD calculated power is 144? At most, Tucki's NVIDIA is 30? Tops, Wei lai, the ideal Mobileye? Q4 is 2.5? At most. This means that the current L2 or L2.5 autopilot host factory has a more realistic choice, instead of adopting Huawei's luxurious computing power and power consumption. Only when L4 autopilot has a realistic solution can Huawei really get involved.
Therefore, in the field of in-vehicle computing chips, Huawei is aiming at the future, not the present. The products of startups such as Horizon, Black Sesame, Cambrian, Shen Jian Technology and Feibu Technology have been applied to autonomous vehicle products. In the field of on-board computing power, China enterprises are making partial breakthroughs. Of course, Bosch and other multinational Tier 1 have complete ADAS solutions. In the long run, Huawei's "all-round" solution may be more suitable for the comprehensive requirements of future car networking, car machines, new energy and ADAS.
Traditional power
In the field of functional chips in computing chips, such as MCU (micro-control unit, especially fuel vehicle power system control) and sensor chips, traditional giants are still firmly in control, and China enterprises have similar products, but their participation is very low, only about 4.5%.
There is no sign of a breakthrough in this piece for the time being. We can only start with small orders and spare tires, hone the stability level of car regulations, and gradually accumulate capital to challenge giants. However, whether to continue to invest in MCU related to fuel vehicles is actually a dilemma for China enterprises. However, some technical reserves can still be used for a long time in the new energy era, such as the out-of-stock ESP (Body Stability Control) chip. It is too rough to think that the traditional MCU chip market is shrinking.
From a dynamic point of view, the situation is not so worrying, because companies are now more and more involved in chips. The "new four modernizations" wave is naturally friendly to new players, and China enterprises can't waste historical opportunities. (Text/"Autobot" Huang) Copyright statement This article is an exclusive original manuscript of Autobot, and the copyright belongs to Autobot.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.