At present, the competition of intelligent driving is gradually shifting from the competition of "technology and products" to "cost and experience".
On March 3 this year, Xiaopeng Motors officially announced that all models of Xiaopeng G6 will be reduced by 20,000 yuan for a limited time, with a starting price of 189,900 yuan after the discount, and the standard NOA will enter the price range of less than 200,000 yuan;On April 1, BYD's Denza N7 launched a new model, with a starting price of 239,800 yuan According to the monitoring data of the Gaogong Intelligent Vehicle Research Institute, from January to February this year, the average delivery price of NOA standard new cars continued to fall, from 359,100 yuan last year to 337,900 yuan.
On the one hand, the user's recognition of high-end intelligent driving is gradually transforming into purchasing power, and urban NOA has become a new focus of the industry's involution; on the other hand, driving and parking share the same domain controller, and the integrated driving and parking solution that realizes the deep reuse of sensors and the sharing of computing resources has ushered in an explosion, and has become a benchmark to test the comprehensive capabilities of autonomous driving. Therefore, although the appearance is different, the essence is that intelligent driving is returning to mass production and moving towards intelligent driving equality and value payment.
Take the Xiaomi SU7 that has just been launched as an example, the standard version of the model priced at 215,900 yuan adopts the NVIDIA Orin N intelligent driving solution, with a computing power of only 84TOPS, this solution is also applied on the Denza N7 intelligent driving version, and Xiaomi is on the platform, standard high-speed NOA, one-key valet parking and intelligent parking assistance and other functions; All the scenes of high-speed and urban NOA, compared to Xiaomi SU7, it is obviously not enough "volume".
Therefore, this also means that if car companies roll up prices and want to popularize high-end intelligent driving functions, the pressure will naturally be transmitted to upstream suppliers, and how to reduce costs and increase efficiency will be the unavoidable keywords of the intelligent track. Among them, in addition to ensuring the sales scale of models, the ability of each player to collaborate with software and hardware, and the ability of large-scale mass production will become important influencing factors.
The price war of high-end intelligent driving solutions is heating up
During the 2024 China Electric Vehicle 100 Conference, a number of intelligent driving companies have shown their own cost reduction plans for intelligent driving products. Among them, DJI Automotive released the line platform, claiming to be on the basis of "7V+32TOPS", the launch of "7V+100TOPS" and "10V+100TOPS" two configurations, as low as 7000 yuan can achieve the urban NOA function, DJI Automotive person in charge said: The program is expected to make 150,000 yuan level passenger cars have the ability to open the city navigation function, accelerating the comprehensive standard configuration of high-end intelligent driving.
Baidu released the Apollo pure vision high-level assisted driving product, which is based on Baidu Apollo Lite, using the second generation of pure visual perception system newly upgraded by 4D BEV Transformer, and at the same time cooperating with the occupancy network OCC technology to do end-to-end real-time reconstruction of the static environment to obtain three-dimensional structural information with higher resolution than the lidar point cloud, so as to realize the replacement of lidar.
In addition, Horizon, a leading domestic intelligent driving technology company, in addition to announcing the upcoming intelligent driving computing solution Journey 6 series (Journey 6 covers the high-end, mid-range and forward-looking all-in-one markets), will also create a next-generation autonomous driving system through a high-end intelligent driving model room, a set of "elegant, calm" anthropomorphic system. In essence, Horizon's high-end intelligent driving model room is to improve the anthropomorphism of the autonomous driving system through algorithm optimization and the release of the ultimate performance of the combination of software and hardware, and the final evaluation criteria are the scene pass rate, commuting efficiency and behavior anthropomorphism.
At the same time, can this series of high-end configurations increase the volume on the consumer side, and under the industry's cost reduction and efficiency increase, pure vision or multi-sensor? For a time, the industry began to deploy "lightweight" supply solutions, simplifying intelligent driving products, de-sensing hardware, selecting high-precision maps, using cost-effective computing platforms, etc., rolling up a cost-effective solution that "needs both high-end and cost-effective".
Whether it is a software algorithm provider or a hardware computing solution provider, from the perspective of the entire industry chain, there has been an obvious trend in recent years - a single-point algorithm breakthrough or a breakthrough in computing power performance is not the key to high-end intelligent driving, how to fully cooperate with the two to make the system efficiency reach the ceiling, and can meet the needs of the market and users, and realize the large-scale mass production and popularization of the system is the key. It can be seen that companies such as NVIDIA and Momenta are facing the combination of software and hardware with full-stack technical capabilities to make up for their past shortcomings. Regarding how to break through and flow downwards for high-end intelligent driving, the industry is gradually forming a consensus that software and hardware synergy + large-scale mass production capacity is the key to solving the problem of high-end intelligent driving.
Software and hardware synergy + large-scale mass production capability is the key
According to the estimation of the Gaogong Automobile Research Institute, by 2025, the market penetration rate of domestic ADAS will exceed 76.4%, the penetration rate of L2 will exceed 45%, and the penetration rate of L3 and L4/L5 is expected to reach 8% and 1% respectively.
So, now everyone is starting to focus on the mainstream L1~L2/L2+ market. If intelligent driving system suppliers want to seize as much market share as possible in this mainstream market, whether they have the ability of full-stack self-research and whether they can give diversified product solutions based on standardized product platforms will be their core competitiveness.
In this regard, Horizon, an intelligent driving technology company, is indeed one of the few players in China that has the integration of software and hardware, and has the ability to mature full-stack layout intelligent driving software. Not only that, Horizon Robotics' journey product shipments have reached 5 million pieces, achieving the industry's largest mass production results. On March 26, Horizon also announced that it officially submitted a prospectus to the Hong Kong Stock Exchange, opening the IPO of Hong Kong stocks, and the prospectus also showed very bright business growth, revenue and gross profit margin performance. And this is inseparable from Horizon's corporate strategy and product technology layout ideas.
We know that at present, high-level intelligent driving represented by urban NOA needs to face the dual challenges of complex embedded supercomputer system technology and extreme engineering capabilities, and Horizon has gradually built a high-end full-stack technical capability combining software and hardware, and unique forward-looking algorithms, hardware computing facilities, and engineering mass production The capability can efficiently support the high-end intelligent driving system to achieve fast and good mass production, and can meet the development needs of intelligent driving in the next 3~5 years.
Yu Kai said: "What is a truly competitive next-generation autonomous driving system, we look at it through three dimensions, the first dimension is called the standard scene pass rate, the second is commuting efficiency, and the third is behavioral anthropomorphism, that is, the behavior aspect makes people feel comfortable and makes people feel anthropomorphic." Therefore, the anthropomorphic system we proposed to be "elegant, unintimidating, and calm" is actually the goal to be pursued by the next generation of autonomous driving systems. ”
At the same time, from the perspective of Horizon, from SoC to operating system, it will be in its "killer" application, that is, high-level autonomous driving, Horizon will create a model room. The so-called model room is that Horizon will create a full-stack software solution, and let the software solution be mass-produced on some key models, so that Horizon can provide favorable support to a large number of Tier1 partners, so that they can cover more models during mass production, and when they are copied to more models, the entire delivery efficiency will be improved.
Obviously, moving towards high-end intelligent driving is too risky for weaker OEMs to engage in full-stack self-development, but at the same time, they also want to be able to participate as deeply as possible. For the parts that they are not good at or can't do, they will seek help from partners in the industry chain, for example, they are responsible for the development of some upper-level application algorithms, and the rest is handed over to the supplier to complete. In terms of supplier selection, they tend to find suppliers who are relatively strong in the field of cooperation and have a relatively open cooperation mentality. Therefore, this often requires intelligent driving suppliers to have the ability of software and hardware collaboration, the ability of large-scale mass production, and the ability to build an open industry ecosystem.
Therefore, whether judging from the industry trend or the current technical dilemma, Horizon Robotics' full-stack deployment of intelligent driving software capabilities on the basis of intelligent driving computing solutions and operating systems is not only a wise move, but also the only way for high-end intelligent driving to move towards large-scale mass production.
But it must be admitted that making the intelligent driving system experience more anthropomorphic and more like an old driver is not something that can be achieved in a day. At the same time, how to make the urban NOA from usable to easy to use, the road of high-end intelligent driving of domestic automobiles has just reached a turning point, as the head supplier of intelligent driving technology, Horizon has occupied the opportunity through the full-stack technical capability layout of the combination of software and hardware, and the opening of the Hong Kong stock IPO will also provide a stronger force for it to stand firm.