According to the report data of iResearch, to? 2023? In 2008, there will be about 6,000 vehicles with intelligent driving function in the world. Ten thousand vehicles, L 1/L2? The penetration rate of the first-class autopilot function will be close to 50%, and the market penetration rate of the third-class autopilot function will also reach 7%.
From the immediate point of view, the deep charging Tesla, the global smart car, the stock price and the total market value have continuously hit a new high of more than 2000? A billion dollars.
This also shows that the whole market has high expectations for the development of smart cars.
1, the domestic smart driving market is ready to go.
We turn our perspective to China: After two years of rapid development, many joint-venture brands, independent brands and new forces in making cars are here? Adas? The mass production of function and intelligent driving system in the car shows a trend of chasing each other.
According to the statistics of Zuosi Industrial Research Institute, before this year? 4? Last month, in the domestic market? L2？ The assembly rate of the first-class autopilot system reached 10.6%.
Among them, traditional brands such as Toyota and Volvo, as well as emerging brands such as Geely, Link, Geometry, Great Wall Weipai and Qiruixingtu, are in a leading position in assembly rate.
Not to mention the new forces of building cars, Wei Lai? NIO？ The pilot, Tucki? XPilot, Weimar's job? A pilot? The ideal assisted driving system has been installed in the car and is developing continuously.
In particular, Li recently received financing from Meituan, and will invest heavily in the development and application of autonomous driving technology in the future.
L3？ The market for autonomous driving is also expanding. 20 17? Audi launches new models? A8？ Advertise the world's first mass production? L3？ Since the automatic driving system, domestic brands have also been launched? L3？ Class automatic driving function.
Including GAC New Energy? Eternal tower? LX/Aion？ Is SAIC Roewe listed? Marvel Comics? x？ And those that are going public? Marvel Comics? R, and Chang 'an, which has attracted much attention this year? U-NIT？ And BYD Han.
However, limited by domestic laws and regulations, it is not clear. What are these models? L3？ The automatic driving function is not yet open, but the market prospect is expected.
And for a higher level? L4？ Autopilot, there are two relatively heavy industry trends in the near future:
One is Volvo and? Waymo？ Hand in hand, will the subsequent Volvo brand, Polaris brand and Lectra brand models be integrated? Waymo？ Autopilot technology;
The second is that Didi autopilot finally announced its landing and began trial operation in Jiading, Shanghai.
Prior to this, a number of autonomous driving manufacturers including Baidu, Pony.ai, Wen Yuan Zhixing and AutoX have landed? Robotaxi？ Service, its scale will continue to expand.
Where is McKinsey, a global consulting management company? 20 18? In 2006, an autonomous driving research report was released, which pointed out that China is likely to become the world's largest market for autonomous vehicles in the future.
Where to? 2030? In 2008, the revenue of new car sales and travel services related to autonomous driving in China will exceed? 5000? A billion dollars.
From advanced assisted driving function to? L3？ Is the self-driving bus here again? L4？ Grade? Robotaxi, a huge incremental market for smart cars, has cultivated unlimited opportunities for various emerging suppliers to land their products.
No matter in the fields of algorithms, sensors or chips, there are bound to be a number of supply chain stars rising strongly.
2. The algorithm gives way to the chip, and the market structure needs to be broken.
The earliest players of autonomous driving mainly developed and improved at the algorithm level, especially the development of artificial intelligence technology with deep learning as the core, which accelerated the process of automobile automation.
Waymo？ The reason why it is at the forefront of the world in the field of autonomous driving is largely because of itself? Google? Algorithm accumulation for many years since the unmanned vehicle was founded.
Nearby? 5? This year, a large number of autonomous driving algorithm companies such as BAT, Didi, Ma Xiao Zhixing, Wen Yuan Zhixing and Momenta were born in China? Wait, these companies have developed their own algorithm systems.
Even if they started a little late, these companies are catching up and trying to gradually bridge the algorithm gap between themselves and foreign companies.
Nearby? 2? In this year, many domestic? OEM? An innovation center and an artificial intelligence technology center have also been established to constantly overcome the problem of autonomous driving from the algorithm level.
To be sure, at present, the strength of domestic autonomous driving algorithm is at the world leading level, and the algorithm is no longer a high wall that restricts the development of autonomous driving technology.
In addition to the algorithm, it is the sensor package of automatic driving.
For low-level autonomous driving, the combination of millimeter-wave radar and camera is enough to support, and there are many such sensors in the world. Tier？ 1? And domestic suppliers to ensure supply.
The lidar market for advanced autonomous driving has also developed from a few years ago to a hundred flowers, including DJI? Livox, Wosai Technology, Sagitar Juchuang and so on.
Various products have been widely deployed on self-driving vehicles.
At present, there is still room for the overall cost of sensors to fall, but mass production is just around the corner.
Overall, the short board from the sensor part in the research and development of autonomous driving is gradually being repaired.
In this way, the focus of the future development of autonomous driving lies in computing hardware, which is what we often call autonomous driving chips.
There are a large number of sensors on self-driving vehicles, and the improvement of the algorithm requires these sensors to continuously collect external data, traverse various extreme situations and constantly evolve.
In addition to solving the problem of large-flow data transmission, the autopilot system also needs to deal with these massive data quickly, and the powerful autopilot chip is the key.
It is generally believed in the industry that L2? The computing power required for autonomous driving is 10? L3 needs 30 computing power at most? –? 60? At most, L4 needs more computing power than 100? At most, L5 needs at least 1000 computing power? At most, sufficient computing power is only a part of the requirements for autonomous driving chips, and it must also meet the requirements of vehicle environment in terms of power consumption performance, safety and stability. The contents involved here are very extensive, including temperature, humidity, vibration, dust, electromagnetic compatibility and so on.
As we all know, the threshold of chip design is quite high. At present, only a few companies in the world can design and manufacture chips widely used in consumer electronic devices, and the development of automotive chips is much more difficult than that of consumer electronic chips.
Furthermore, the design difficulty of the sensor chip applied to autonomous driving is higher than that of the ordinary automobile instrument chip.
It is no exaggeration to say that the automotive sensor chip used in autonomous driving can be called the "Everest" in the field of chip design.
So far, only autopilot chips with large computing power can be mass-produced in the world? Mobileye, NVIDIA, Tesla.
According to? Mobileye？ The latest published data, its? EyeQ？ The chip has been sold more than? 5400? 10,000 yuan, which is super-lifted in the world? 5000? On ten thousand cars, its customers include? BBA？ Top car companies include.
What did Tesla use for a long time? EyeQ？ Series of chips to drive? Autopilot? System.
In China, Wei Lai? ES6/ES8, Ideal? First, GAC New Energy? Eternal tower? LX？ And other models have been adopted? EyeQ4？ As the core of its driver assistance system.
Mobileye？ This kind of record is basically hard to match in the market, which is why Intel wants it. 20 17? The reason why we have to spend huge sums of money to buy this company.
GPU？ Overlord NVIDIA has also gained many customers in recent years with its high-performance computing platform, including Didi, Pony.ai, Wenyuan Zhixing and AutoX? Including? Robotaxi？ Operators have adopted NVIDIA's autopilot computing platform.
Has Tucki been listed recently? P7？ It is also equipped with an autopilot controller based on Xavier and NVIDIA, jointly developed by IT and Desai Siwei? IPU？ 03。
Needless to say, Tesla realized the importance of the autopilot chip itself, and Musk also made up his mind to invest a lot of money in research and development. Fire Services Department.
To be sure, the fire department? After all, it will not be supplied to the outside world. FSD？ It will be the core competitiveness of Tesla in the future, and Tesla's own needs can fully support its chip development team.
Under such a market structure, car companies should seriously think about how to build their own self-driving chip supply chain in the future.
So far, Mobileye? The chip and its algorithm are closely coupled and sold in packages. Car companies have little autonomy. It is almost impossible to use the self-developed algorithm, so the industry is right. Mobileye？ The evaluation is too closed.
Moreover, if car companies want to use NVIDIA's chips, the threshold is also very high, and the high membership fees and joint development costs are not affordable for ordinary enterprises.
It is said that Xpeng Motors and Desai Siwei are both trying to develop? P7？ On the autopilot domain controller, pay near NVIDIA? 8? A membership fee of several dollars.
China's automobile industry, which keeps saying that it wants to be the world leader in the field of automobile intelligence and autonomous driving, is now highly respected by people in the field of autonomous driving chips.
The "domestic autopilot chip" needs to accelerate to the stage and become the real protagonist.
Fortunately, there are many enterprising people in domestic enterprises.
3. Breakthrough of domestic autopilot chip.
In recent years, many excellent companies have emerged in China, and they have been working hard in the field of autonomous driving chips. They all hope to get a slice of the huge smart car market in China.
Among the giants, Huawei has confirmed that it will rely on its Hisilicon to enter the field of intelligent driving computing chips and platforms. AI？ Chip and? MDC？ It will shoulder the heavy responsibility of the rise of domestic intelligent driving chips.
Among many innovative companies, black sesame intelligent technology is also a typical representative of self-driving "domestic core".
Black sesame seeds one after another? 20 19? Year? 8? The moon and? 2020? Year? 6? Two generations of self-driving chips "Huashan No.1" and "Huashan No.2" were listed in May, and they can already compete with foreign competitors in performance and energy efficiency ratio.
This year? 6? Huashan II, the latest product of Yuehei Sesame? A 1000？ For example, the chip, has it reached the computing power? 40? -? 70? TOPS, what is the corresponding power consumption? 8? W, the energy efficiency ratio exceeds? 6? TOPS/W, which is currently at the leading level in the world.
The reason why we can achieve such high computing power is inseparable from the core of black sesame self-research. IP——DynamAI？ NN engine
The engine adopts a large computing power architecture and supports multi-form and multi-precision operations. It has the advantages of adaptive quantization, structured cropping and sparse acceleration, and is also equipped with automatic development tools.
Perception technology is the core advantage of black sesame, and A 1000? It is a very typical sensor chip.
This chip is built-in? 8? Star? CPU？ Core, including? DSP？ Digital signal processing and hardware accelerator, supporting the access of mainstream sensors in the market, including lidar, millimeter wave radar and 4K? Camera, GPS? Wait a minute.
It is worth mentioning that A 1000? Because of the integration of self-developed black sesame seeds? NeuralIQ？ ISP？ Pipeline, how much can you support at most? 12? Road high-definition camera access.
This is very helpful for the automatic driving system that relies heavily on visual perception.
At the level of chip architecture, black sesame has also created multiple layers of heterogeneity? TOA？ Architecture.
This architecture organically combines light control technology, image sensing technology, image and video compression and coding technology, computer vision processing technology and deep learning technology.
TOA？ The architecture is extensible and supports multi-chip cascade expansion, which is why it is based on Huashan II chip. Black sesame? Fashion? Are computing platforms compatible? L2？ Autopilot to? L4？ The reason for autonomous driving.
At present, for the low level? In the ADAS scenario, can customers base on chips with low computing power? A 1000L？ Building computing power? 16? What is the maximum power consumption? 5W？ Computing platform based on.
And for the top? L4？ Autopilot, can customers put it? 4? Shielding A 1000？ The chips are connected in parallel to achieve up to? 280? Most? A computing platform for computing power.
Scalable mode is also the strategy adopted by most mainstream autopilot chip manufacturers at present, NVIDIA's? Driving? AGX？ Pegasus, ZF? Dear? AI？ All adopt this model.
Is there any plan for black sesame after Huashan II? 202 1? Sometime in 2008, Huashan III was launched. What are the main objectives? L4/L5？ Class-8 autopilot platform, its computing power will surpass? 200? Tops, while using more advanced? 7nm？ Technology. 200? Most? Will the level of computing power be equal to that of NVIDIA? Olin? Chip.
And at the level of soft power, compared? Mobileye？ Under NVIDIA's closed and high entry threshold, Black Sesame's system is more open and cost-effective.
This is reflected in the fact that Huashan chip of Black Sesame has a perfect tool chain and an open software and hardware platform, which supports car companies to carry out independent and controllable innovation.
This may make "bitter?" Mobileye？ A long time? OEM? The children saw the dawn.
On the whole, Black Sesame has developed the sensing algorithm and core needed for autonomous driving. IP, chip system architecture and tool chain have formed a mature product system.
The key behind these products is of course the team.
Autopilot chip products cover two important industries: chip industry and automobile industry.
To build such a cutting-edge product, a team needs the blessing of talents from both automotive background and chip background, and the team of Black Sesame has just formed such a benign complementarity.
Black sesame co-founder and? CEO？ Shan Zhangji (right) co-founder and? Chief operating officer? Liu Weihong (left)
Black Sesame Co-founder? Before the CEO's note, it was an image chip company? Head of OmniVision(OV) image algorithm, who owns it in the field of visual perception? 100? A number of patents.
Co-founder and? Liu Weihong, the chief operating officer, worked for General Motors and Bosch Company successively, and once served as the president of Bosch Chassis Braking Asia-Pacific region, and worked in OEM and Bosch. Tier？ 1? They all have profound work experience.
One is from the chip industry and the other is from the automobile manufacturing industry. The two complement each other and will create a world in the field of autonomous driving chips in the future.
Regarding the composition of the team, Liu Weihong once mentioned, "Our genes are chips, and there are cutting-edge talents in the team who do research and development of automotive chips and verification of automotive chips. What they want to do is different from what they have already done. At the same time, we also integrate developers who know both algorithms and computing architecture. 」
At present, Black Sesame has teams in China and the United States, and has nearly? 300? Many employees and team members are here? More than, Anba, Qualcomm, NVIDIA and other chip companies, with an average experience of more than? 15? Year? .
4. Self-driving chip mass production March
In order to achieve mass production, autonomous driving chips must pass the automobile regulations.
The vehicle regulation level of autonomous driving chip includes not only the reliability, stability and durability requirements of the chip itself, but also the functional safety of the system after integration with the vehicle system. At present, few suppliers in the market can meet these two requirements at the same time.
Is black sesame newly introduced? A 1000？ Chip, from the beginning of design, has been moving towards the goal of vehicle regulation level.
Does it fit the chip? AEC-Q 100？ Reliability and durability? Grade? 2? Standard, the chip as a whole has reached? ISO？ 26262? Functional safety? ASIL-B？ Level.
Is there still satisfaction inside the chip? ASIL-D？ Safe island level, what is the functional safety level of the whole chip system? ASIL D.C.
To this end, A 1000? What chip is used? ARM car rating? CPU？ And then what? Graphic processor.
In terms of OEM, Black Sesame also chose TSMC according to the requirements of vehicle regulations? 16nm？ Production line. The goal of all this is to achieve the design goal of this chip.
The old one? 6? In May, the R&D team of Black Sesame has tested the performance of all modules of the chip and passed the debugging completely. Next, it will conduct joint testing with customers to prepare for the final mass production.
It is reported that the first model equipped with this chip will be in 202 1? Mass production at the end of the year.
Besides, Huashan No.1 with black sesame seeds? A500？ The chip has also started mass production, aiming at domestic head car companies? L2+？ And then what? L3？ The project of automatic driving at the checkpoint is also in progress.
Compared with the traditional automobile electronic chip manufacturers, the planning of black sesame is obviously more rapid and radical, and they need to grasp the time window more agile.
This time window is also in these two years, especially this year, most smart driving models are already in chip selection, and manufacturers who can come up with this chip product will undoubtedly take the lead.
Black sesame? A 1000？ The chip went on the market this year? 6? Released in June, it has entered the rhythm of the market in the process of mass production.
On the eve of the outbreak of the global smart driving car market, the market competition of autonomous driving chips has become more intense.
What's a head? Giants like Mobileye and NVIDIA; There are also manufacturers such as Huawei Hisilicon and Qualcomm that stand out from the mobile chip market; Established automotive semiconductor manufacturers are also accelerating their layout.
Like black sesame? AI？ Chip companies will become important rising stars.
Intelligent driving cars will be a market with great growth potential.
Under such market opportunities, domestic self-driving chip manufacturers such as Black Sesame cooperate at the product level. Mobileye？ To compete with international giants like NVIDIA, the future is to attract customers, build an ecology and build their own moat.
Shan Zhangji has revealed the same vision on many occasions: personal computer? Processor enterprises represented by Time Intel; Is there a smartphone era? Arm? The mobile chip company represented by thought; And black sesame hopes to become Intel and? Arm.
The general trend of smart cars has become, "Big? Things? Start? Small.
Under the huge industrial opportunities, can today's rising stars become future industrial giants?
Let's go and have a look.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.