However, in the uncertainty, the automobile industry is very sure about the future direction. There are some big news in the autopilot-
On June 23rd, Mercedes-Benz, which just announced a peaceful breakup with BMW in the field of autonomous driving, announced that it has reached a cooperation with chip supplier NVIDIA, and will use the latter's Orin chip to develop the next-generation on-board computing system, which will provide computing support for L2-L3 autonomous driving function and automatic parking function up to L4 that will be fully carried by Mercedes-Benz production models in 2024.
On June 25th, Volvo Car Group announced that Volvo will reach a strategic partnership with Waymo, an autonomous driving company owned by Google, and cooperate with L4-class autonomous driving technology on a brand-new electric vehicle platform to explore business scenarios such as autonomous driving network car.
On June 26, Amazon officially acquired Zoox, an American autonomous driving company, and Amazon paid more than $654.38+02 billion for it.
On June 27th, the manned demonstration operation of Didi Autopilot Network was officially launched in Shanghai, and the whole process was broadcast live by CCTV. Starting from this day, Didi's self-driving test car in Jiading, Shanghai will be open to the public. Didi has launched a "Future Travel" page in the application for the public to apply for a test ride on the autonomous driving network.
For a time, big companies almost started the autonomous driving arms race. There is no doubt that the relevant companies realize that the future car will be a supercomputer running on wheels. High performance computing chips are becoming more and more prominent in this arms race.
First, Mercedes-Benz has a new love, just because of it?
On June 23, four days after the cooperation with BMW in autonomous driving was suspended, Mercedes-Benz threw herself at chip supplier NVIDIA, and the two sides reached a cooperation to develop a computing platform for self-driving vehicles that Mercedes-Benz will mass-produce in 2024.
In the announcement a few days ago, the two sides also said that "in view of the cost of establishing a technical platform for * * *, as well as the current business and economic situation, now is not the right time to successfully implement cooperation." Too much money seems to be the key reason why the two sides decided to suspend technical cooperation.
However, the light speed cooperation between Mercedes-Benz and NVIDIA later pointed to factors other than money. Generally speaking, the cooperation between car companies will not affect the cooperation between car companies and suppliers, but the cooperation between Mercedes-Benz and BMW is different. Before reaching a cooperation with Mercedes-Benz, BMW has formed an autonomous driving alliance with Mobileye, the world's largest supplier of ADAS systems, and developed autonomous driving based on its EyeQ series chips.
The cooperation with BMW means that Mercedes-Benz will choose Mobileye's chip to build a key autopilot computing unit. And this may be the most important difference between the two sides. Sam, chief analyst of foreign consulting firm Guidehouse? Abuelsamid said, "What? I suspect that these two automakers can't reach a * * understanding of the platforms they use. Now, they are working with Intel/? Compared with Mobileye's products, Orin seems to be a more powerful solution. "
Judging from the public information, Sam's analysis is not unreasonable. EyeQ, the next-generation autopilot chip planned by Mobileye? 5, its computing power is 24TOPS (24 trillion operations per second), while the computing power of Orin released by NVIDIA at the end of last year is as high as 200TOPS. In addition, Mobileye has been very strong in the past cooperation with car companies (despite the promise of EyeQ? 5 will be more open), and the functional modules it provides are often the "black box" of OEMs; And what about the driver built by NVIDIA Autopilot? AGX software platform has taken an open road from the beginning, which can support automobile manufacturers to independently develop algorithms on their computing platforms.
In fact, before this, when Mercedes-Benz explored the development of autonomous driving network, it chose NVIDIA's Drive because of its high requirements for chip computing power. Pegasus car computer. The official announcement on June 23rd means that Mercedes-Benz has turned to NVIDIA for chip selection in the era of autonomous driving, and the cooperation between the two parties has been extended to the production models of Mercedes-Benz.
Waymo, which has reached an autonomous driving strategic cooperation with Volvo, relies on Google's technical strength in the AI field and uses a self-developed TPU. Although the computing power of TPU used by Waymo on the vehicle side has not been announced, according to the official disclosure of Waymo, the performance of its autopilot system has improved by 15 times after using TPU.
The position of the chip in autonomous driving can be described as "invisible champion". You can't see it from the appearance of the vehicle, but the self-driving car can drive smoothly and is definitely the number one hero.
Second, the autonomous driving competition is also a chip competition.
Whether Mercedes-Benz gave up BMW to hold hands with NVIDIA, Volvo and Waymo reached a strategic alliance, or Didi's autonomous driving network started to get a car, the big news that happened last week showed that auto companies and technology companies put autonomous driving in a crucial position: In the near future, autonomous driving function is an important part of automobile product strength; In the long run, L4 autopilot may completely change the business model of the automobile industry after it is put into large-scale application.
All these changes are based on a small chip. In order to gain a competitive advantage in autonomous driving ability, the companies participating in this competition either independently developed or joined forces with Lian Heng to find a high-performance autonomous driving chip. There is a very typical example in the industry: Tesla.
As the leader of intelligent electric vehicles, Tesla has had cooperation experience with two mainstream autonomous driving chip manufacturers in the market. However, due to Mobileye's strength and isolation, NVIDIA's continuous power consumption and high development cost, it failed to achieve long-term cooperation. In order to give full play to the advantages of software and hardware integration of autonomous driving, Tesla took the lead in independently developing the FSD of autonomous driving computing platform among car companies, and its computing power reached 144TOPS. The computing power support of FSD for autonomous driving mainly comes from two AI chips, and the computing power of a single chip is about 72TOPS.
So far, Tesla's FSD still maintains the automatic driving calculation record of production cars. Tesla believes that FSD is enough for its fully automatic driving (full? Self-driving) function provides support.
There is no doubt that the competition of autonomous driving is also the competition of chips. The importance of the whole automobile industry to autonomous driving and even comprehensive steering will generate huge demand for autonomous driving chips. If any company occupies a considerable share in the autonomous driving chip market, then the corresponding market value may be hundreds of billions of dollars of imagination.
At present, attracted by the huge market, there are four forces in the field of autonomous driving chips, new or old:
The first category is the old ADAS chip/autopilot chip suppliers such as Mobileye.
This kind of enterprise is the enterprise that participated in the market competition when the automobile industry began to develop advanced assisted driving system (ADAS). The competitive strategy of these enterprises for autonomous driving is to upgrade their existing products through the technology and customer resources accumulated in the ADAS market and realize a smooth transition to autonomous driving. A typical example is Mobileye's continuous iteration of EyeQ series chips.
In addition to Mobileye, Renesas, NXP, Texas Instruments, Denso and other established automotive semiconductor suppliers all have their own self-driving chip plans.
The second category is semiconductor giants who see opportunities for autonomous driving chips, from different fields.
For example, NVIDIA mentioned above, whose main business is GPU, data center, etc., belongs to consumer electronics, but NVIDIA quickly entered this market after understanding the demand for high-performance chips for autonomous driving, and has now launched Drive? PX, drive? AGX？ Xavier, drive? Olin has three generations of products and won orders from many automobile companies.
Qualcomm, whose main business is communication, dominates baseband chips and mobile phone SoC. After trying to acquire NXP and get tickets for autonomous driving competition, it launched Snapdragon at this year's CES. Take the autopilot computing platform. According to the official information of Qualcomm, this computing platform based on Qualcomm chip has a maximum computing capacity of 700TOPS, and can support L4-L5 autopilot.
Before Qualcomm, Huawei, whose main business was communication and consumer electronics, had released the autopilot computing platform MDC? 600。 This computing platform consists of eight 3 10? Integrated AI chip, the highest computing power reaches 352TOPS.
The third category is the self-driving chip start-ups born under new opportunities.
In China, the horizon is a typical example.
This month, the Chang 'an unit equipped with Horizon automotive AI chip Journey 2 was officially launched. In this way, Horizon realized the first "boarding" of domestic autopilot chips. On the other hand, Journey 2 with 4TOPS computing power is also the first automotive AI chip in China.
Later this year, Horizon will also release Journey 5, which has a computing power of 96TOPS and supports 16 HD camera signals. This chip has the computing power beyond Tesla FSD and will be oriented to advanced autonomous driving.
The last category is the self-research school of car companies represented by Tesla.
Because car companies have little experience in semiconductor manufacturing, they usually purchase chips from suppliers. Tesla, headquartered in Silicon Valley, has different genes. In order to give full play to the advantages of software and hardware integration, Tesla independently developed FSD based on the semiconductor talent resources in Silicon Valley.
At present, it is difficult to copy the model of self-developed autopilot chips of car companies, and Tesla is probably the only seedling of this path.
In China, there is no plan to develop autonomous driving chips, whether it is traditional car companies or new car-making forces. As the largest single automobile market in the world, China naturally becomes a battleground for the suppliers of autonomous driving chips.
3. Can China give birth to an autonomous driving chip giant?
With so many contestants, the blue ocean market of autonomous driving chips to be developed seems to have presented a competitive situation in the Red Sea. Over the past two years, a series of events around chips in China and the United States have made people feel sorry for the weakness of China's chip industry. At the beginning of the year, the National 1 1 jointly released "Innovative Development Strategy for Smart Cars" to "New Infrastructure", all of which focused on the research and development of in-vehicle chips. China's automobile industry hopes that more local chip companies can rise strongly.
Nowadays, with the intelligent transformation of the automobile industry, there is a large demand for autonomous driving chips. Can China chips catch up and cultivate a local supplier of autonomous driving chips in China? The answer is uncertain, but in June, Horizon Journey 2 chip was put on the "train" of Chang 'an unit, which has at least made a good start. It is understood that in the field of ADAS chip, the perceptual computing performance of Journey 2 chip has surpassed that of Mobileye chip, especially in view of the special road conditions in China, and it has successfully signed more than ten designated models of major domestic automobile groups.
Yu Kai, the founder of Horizon, summed up the differentiated advantages of Horizon in a media interview: "We are the only chip company in the world that can provide such a level of power consumption and computing power, and is open and empowered. NVIDIA has no products in intelligent cockpit assisted driving and multi-mode interaction, and the chip power consumption is relatively high. Our power consumption and computing power can be compared with Mobileye's positive PK, but Mobileye is not open, and we can meet the needs of independent development of car companies ",and said that we are confident to get the global market of 1/3 in the future.
In fact, when the trend of autonomous driving swept through, China chip companies such as Horizon, which took the lead in targeting the automotive AI chip market and passed the pre-assembly and mass production verification, really ushered in the best era. As the largest automobile market in the world, China, together with some typical characteristics and demands of the development of autonomous driving technology, has created rare opportunities for local autonomous driving chip enterprises.
First of all, autonomous driving technology has a strong regional character.
Due to the differences in natural conditions, traffic scenes, traffic rules and even cultural traditions around the world, it is difficult to reuse the autonomous driving technology developed by one country to other regions. This influence will be directly transmitted to the hardware level-because it is highly integrated with specific data and algorithms, it is difficult for autopilot chips to be unaffected by regional characteristics.
In this case, enterprises with strong R&D local teams and better understanding of China's data and scenes have a greater probability of developing autonomous driving chips that are more suitable for China's scenes, and the combination of algorithms and hardware is more efficient.
Secondly, when cars are regarded as electronic products by more and more people, people have higher expectations for the frequency and speed of their function iteration, and the automatic driving function is no exception.
ADAS, which used to be dominated by foreign suppliers, has never been updated since its function was put on the bus. However, when the car becomes intelligent, the car can actually upgrade its functions through continuous OTA, and even realize the leap from ADAS to semi-automatic driving and automatic driving. For example, Tesla realizes the model through upgrading? NOA function of 3 is a typical example.
Only the Tesla family, of course. For more car companies, to accomplish this task, it is necessary to maintain high-frequency and close contact with suppliers of autonomous driving chips and jointly develop them.
This change further tests the rapid response of suppliers to the needs of car companies. In other words, this requires the self-driving chip supplier to establish a large-scale on-site support team to achieve rapid feedback and support for the needs of car companies. Obviously, local teams without cultural and language barriers can be better qualified.
Finally, car companies have more functional differentiation needs in autonomous driving research and development.
When ADAS functions have been highly standardized or are in the limelight in automobile products, it is difficult to become a bright spot to attract consumers. In this regard, visionary and capable car companies have chosen to develop new and different autonomous driving functions based on the scene (such as BMW's automatic tracking and reversing), thus gaining new competitiveness.
This trend requires that the suppliers of autonomous driving chips can no longer simply adopt the past "black box" mode and directly give car companies a complete functional module, but give them the right to carry out secondary development and in-depth development. In other words, this requires auto-driving chip suppliers to change their thinking and empower auto-driving development of car companies.
Specifically, this requires chip suppliers to change their thinking, open their strategies, and empower auto companies to develop autonomous driving; In terms of product strategy, it is necessary to solve problems for car companies and reduce the difficulty and cost of developing differentiated functions based on autonomous driving chips by building a tool chain.
Judging from the above three characteristics, the arrival of the autonomous driving trend will test the service awareness and rapid development ability of autonomous driving suppliers. However, due to historical, cost, political and other factors, foreign chip suppliers rarely set up large-scale R&D and on-site support teams in China, and the past openness and development speed can hardly meet the new demand. This is a breakthrough for the rise of local suppliers of self-driving chips in China.
Finally, judging from the situation, the foreign chip giant industry is relatively mature, and it entered the automobile industry earlier, with different barriers. It is not easy for China's self-driving chip suppliers to compete with them on the same stage and finally break through.
However, if the local suppliers of self-driving chips can catch up with each other in terms of chip computing power and power consumption, and give full play to their core advantages to seize the opportunity of intelligent transformation of automobile enterprises, then the birth of China local self-driving chip giants will be a high probability event.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.