Written by | Liang Qiumeng
Edit | Huang Zhi
Produced | Automotive Sankei
On January 12, Black Sesame Intelligence, a maker of autonomous driving computing chips, announced that it has received investment from Boyuan Capital, a subsidiary of Bosch.
Black Sesame Intelligent became the first autonomous driving chip company invested by Bosch in China. This is also the second time that the young company has received a very high endorsement after receiving investment from Xiaomi last September.
In fact, although NVIDIA, Qualcomm, and Intel have made a large-scale layout in the field of automotive main control chips in recent years, the products represented by the Horizon Journey series and Huawei Ascend chips have also emerged.
For example, in 2020, changan automobile's main model UNI-T chose the horizon journey 2 car smart chip.
Regarding Bosch's strategic investment, Yang Yuxin, CMO of Black Sesame Intelligence, interpreted it like this:
"With the development of the entire automotive industry, even these global iron fists, it also needs to adapt to the needs of the times and the local market, and the local supply chain is actually one of their directions."
Underneath the understated tone, it seems to reveal the confidence that local companies can also lead the development of car rules chips.
Looking at the mass production and loading of main control chips in the Chinese market, the dominant force in the mass production link is still foreign companies such as NVIDIA, Qualcomm and Intel.
Mass production and installation of main control chips in the Chinese market
In April last year, Black Sesame released the Huashan No. 2 A1000 Pro autopilot chip, and the INT4 computing power of a single chip can reach 196TOPS, and it is expected to achieve mass production and listing of the model as soon as the end of 2022.
If the big computing chip of black sesame is successfully installed in 2022, the number of independent players will increase by one. The Horizon Journey 5 series of high-computing chips is also about to be mass-produced in 2023, and it seems that local companies have indeed broken the competitive barriers of vehicle-grade chips and may usher in a golden period of development.
Affected by the wave of core shortage in 2021, it has become an industry consensus to independently master the core technology in the supply chain. Under the trend of software-defined automobiles, the main control chip and computing platform have become the core of the development of automobile intelligence.
If the local core enterprises successfully rise, this will undoubtedly add a powerful tonic to the dream of overtaking in the Chinese automotive industry.
How will the rise of local companies evolve? Can they stir up this market?
In mid-January, in Beijing, Yang Yuxin and Dr. Deng Kun, vice president of black sesame intelligent application engineering, expressed their views on the development of the industry.
What is the level of local in-vehicle chips?
Continuously improving the computing power of chips is common in the competition in the era of smart phones, and now this scene is also playing out in the field of smart cars. In the past two years, some car companies have begun to use computing power as a publicity point:
NIO ET7, equipped with Nvidia DRIVE Orin chip, hash rate 1016 TOPS;
Zhiji Automobile and SAIC R Car also use Nvidia's chips, with a computing power of 500-1000+ TOPS...
To compete with the competitiveness of local core companies, we must first ask what level of computing power they are in the industry?
Compare Black Sesame's Huashan No. 2 A1000 Pro with the two large-power car specification chips that have been launched on the market:
Tesla FSD uses a 14nm process with a single-chip computing power of 72 TOPS.
NVIDIA Orin uses a 7nm process with a single chip computing power of 254 TOPS.
Black Sesame Intelligent A1000Pro adopts a 16nm process, and the int4 computing power of a single chip is as high as 196TOPS, and the INT8 computing power is 106 TOPS.
The computing power of Tesla FSD and Nvidia Orin is calculated according to the INT 8 standard, so it can be roughly understood that the location of the Black Sesame Smart A1000 Pro is between Tesla FSD and NVIDIA Orin.
However, according to Horizon founder Yu Kai, the pursuit of pure computing power breakthroughs is not sustainable and will encounter a ceiling. Computing power is important, but it does not represent the true performance of automotive smart chips. Therefore, the real level of the A1000Pro will not be known until the mass production is loaded.
Black Sesame Smart CMO Yang Yuxin
Although the computing power does not determine the final performance, in Yang Yuxin's view, local companies have their unique competitive advantages.
Compared with NVIDIA, Black Sesame Intelligence is closer to customers, and the timeliness of Black Sesame Intelligence will be higher in receiving customer needs and two-way feedback in the development process.
Compared with ADAS market leader Mobileye, because Mobileye has been using a sensor + chip + algorithm binding integrated solution, in the L2 level of autonomous driving, it can quickly increase market share, but in the long run, the "black box" method has been unable to meet the needs of manufacturers who need to quickly iterate upgrade products.
The black sesame intelligent choice is open, not to do the "black box". Its software and hardware solutions can be decoupled, and the underlying software is completely provided to customers in the form of open source, and customized development is carried out by customers to achieve overall software integration.
Yang Yuxin believes that black sesame as a local company can develop so fast, A1000Pro can greatly improve the computing power performance, inseparable from their adherence to the principle of self-developed core IP (intellectual property, the abbreviation of Intellectual Property).
There are two core IPs of Black Sesame. One is the vehicle-level image processor NeuralIQ ISP, in simple terms, the human eye in the light and dark gap between the two scenes may cause temporary "blindness", in the entrance and exit tunnel, oncoming lanes of high beams and other scenes, the human eye perception performance is not very good. The same is true for cameras, so when processed with isp modules, the entire system can be "seen more clearly".
The other is the neural network module NPU, which is mainly used to identify the data after the sensor perceives and analyze vehicles, lane lines, and other traffic participants.
"Most of the companies with the leading self-driving technology will develop their own IP, such as NVIDIA, Tesla, Qualcomm, Mobileye and Huawei, all of which are self-developed IP." Yang Yuxin said.
"The window is only 3 to 5 years, and it can't catch up with no drama"
"We judge that in about 3 to 5 years, that is, in 2025, this window will slowly close." Because the automotive industry itself has relatively conservative characteristics, after the supplier system of new technologies is established, they will not give so many opportunities to new manufacturers. ”
Yang Yuxin described the urgency of chip companies as such.
Black Sesame is currently establishing cooperative relations with major car companies, including China FAW, Bosch, Weilai, SAIC, BYD, Dongfeng, Zhongke Chuangda, Asia Pacific, Baolong, Soto Ryan, Niujie Technology, Lianyou, etc., which have carried out a series of commercial cooperation on L2/3 ADAS and autonomous driving perception system solutions.
Black Sesame Intelligent Development History
In the mass production process, the Huashan No. 2 A1000 chip will be applied to the Hongqi flagship SUV model as a smart driving brain.
As for which model will be preferred for the A1000 Pro, which will be mass-produced at the end of 2022, Yang Yuxin did not disclose.
Yang Yuxin said: "As long as you complete the mass production process with a leading car company, this will greatly shorten the time to gain the trust of other car manufacturers." So our goal is to get the first few in the hand first, and after getting it done, it may be 3 to 100 copies, which will progress faster than the previous ones. ”
At present, the number of high-computing chips on the market is not much, but it has also reached a small climax on the Weilai ET7, Ideal X01 (code name), SAIC Zhiji L7, and Jihu Alpha S HI version models, which have chips from NVIDIA Orin and Huawei MDC810.
In addition to the above two products, Qualcomm's Snapdragon Ride platform plans to achieve mass production in 2022, and is expected to cooperate with Great Wall's models.
Everyone is competing for the opportunity to mass-produce L3 and L4 autonomous driving solutions. Therefore, for Black Sesame Intelligence, mass production in 2022 will be a key symbol of whether it can successfully occupy the position.
"Car companies have high costs of self-research, which is not necessary"
In the era of intelligent cars, the term "full-stack self-research" from the Internet industry has become a slogan shouted more and more by car companies. Some people are worried that assuming that self-developed chips also become a part of the full-stack self-research of car companies, this will pose a threat to chip manufacturers such as Black Sesame Intelligence.
However, in Yang Yuxin's view, there is no need for current car companies to learn Tesla, and full-stack self-development is actually something that happened at a specific historical stage.
He took Apple's mobile phone as an example, "When Apple entered the mobile phone industry, it was an outsider, and the things it envisioned were completely different from the things of the traditional industrial chain, and it could not get the support of the industrial chain." In order to succeed, Apple not only has to have good ideas, but also has strong enough technical strength, so Apple chose full-stack self-development. ”
The same is true for Tesla's choice of full-stack self-development.
"Tesla's definition of intelligent new energy vehicles can be said to have become the direction of all car companies to chase, this time the entire industry is moving." If the car company wants to do something similar, it does not need to do it at all, because if someone does it well, the car company can use faster speed and lower cost to 'get on the horse'. ”
Of course, this does not mean that car companies do not need to master new capabilities at the supply chain end.
On the contrary, Deng Kun believes that if the OEMs want to participate in the formulation of technical routes, or even take the routes into their own hands, it is certainly impossible to simply provide them by first-tier suppliers.
Deng Kun, Vice President of Black Sesame Intelligent Application Engineering
"In the future, maybe many car manufacturers will become Tier 0.5." Deng Kun said. In fact, at present, there are already car companies "paying tuition" through investment, and some car companies are doing a lot of PoC (proof of concept) and rehearsal in the early stage.
In short, car companies are no longer the inherent form of traditional manufacturers.
Correspondingly, second-tier suppliers such as Black Sesame Intelligence cannot remain static. Deng Kun said that their customers may be both traditional first-tier suppliers and car companies, so the identity of black sesame will also be divided between 0.5 and 1.5.
"Foreign giants are Nokia status, optimistic about the development of domestic brands"
Today, The Ningde era, a well-known lithium battery supplier in China, in its fledgling and little-known growth period, was selected by BMW to establish early technical cooperation, and then gradually embarked on the road of rapid development.
For chip companies, will directly conquer foreign brands make their expansion go faster?
Deng Kun disagreed.
He revealed, "We have also had in-depth communication and exchanges with foreign brands, including BMW, Daimler Benz, etc., their development progress is not very fast, they are more like the status of Nokia in the smartphone industry." ”
We can't judge whether this statement is true or not, but Black Sesame Intelligence is obviously more inclined to make good use of the domestic market policy orientation, and quickly land products and increase market share by cooperating with independent brands.
Yang Yuxin further predicted that within a few years, independent brands will surpass joint venture brands.
"The competitiveness of independent brands comes from the aggressiveness of the definition of selling points. The most typical example is that only Chinese car manufacturers dare to hand over products with a chip computing power of 1000 TOPS, although they cannot be used in the car, but for Chinese users, it is a selling point with strong attraction. Yang Yuxin said.
The chip power of mass-produced models in the domestic market is increasing year by year
The scene definition is another competitiveness of the independent brands that Yang Yuxin sees.
He analyzed, "10 years ago, Chinese mobile phone brands killed most foreign mobile phone brands in one fell swoop, relying on the tactics of the machine sea. Others out of 5 models in 1 year, may be domestic brands out of 50 models. Now the strategy of Chinese car brands is very similar to that of then, and they are all defining more segmented groups of people, such as Great Wall Motors. ”
Perhaps as Yang Yuxin said, after a few years, independent brands will successfully counterattack.
However, in this process, independent brand car companies must have stronger automotive intelligent innovation power and appeal, which urgently needs a group of powerful chip partners to help them achieve rapid transformation to automotive intelligence.
In the battle of the vehicle specification chip, black sesame intelligence has achieved leading results, the responsibility is certain, as to whether it can become the main force of the autonomy of the vehicle specification chip, can only wait for time to reveal the answer.