How about the sales of Xiaopeng? A few days ago, Xiaopeng Automobile announced the delivery results in September, according to the data, Xiaopeng Automobile total delivery in September reached 3478 units, an increase of 145%, a new high single-month delivery in 2020; Xiaopeng P7 single-month delivery of 2,573 units, a single-month delivery of a new high, and from the point of view of the current production capacity of Xiaopeng P7, the results are very worthwhile to be affirmed.
At the same time, Peng Auto said that in all the vehicles delivered, more than 93% of users chose the intelligent version of the model equipped with XPILOT intelligent driver assistance system. In other words, the majority of consumers are very interested in the intelligent driver assistance system of Xiaopeng Auto, and it may even be said that this system is the decisive factor influencing the decision of these users to buy Xiaopeng Auto. So let's talk about the intelligent driver assistance system of Xiaopeng Auto.
First of all, we need to dial back to 2019 and look at the feud between Xiaopeng and Tesla. In March of last year, Tesla sued its former engineer, Guangzhi Cao, for allegedly stealing trade secrets related to the Autopilot system and bringing them to Xiaopeng.
The key figure in the lawsuit, Guangzhi Cao, is a former computer vision scientist at Tesla. As per Tesla, he worked on the development of the Autopilot system and was one of the 40 people on the Autopilot system team who had access to the source code. However, on January 3, 2019, Cao Guangshi suddenly submitted his resignation to Tesla, and then jumped to Xiaopeng Automobile as the head of perception.
Tesla said that between November 2018 and January 3, 2019, Cao Guangshi backed up Tesla's entire repository, Autopilot and neural network source code libraries, and uploaded them to his personal iCloud account, totaling more than 300,000 files and directories related to the Autopilot system, and that he then deleted 120,000 files from his work computer and disconnected his iCloud account and cleared all browser history. Tesla argued that Cao Guangshi and Xiaopeng did not have the legal right to use the technology related to the Autopilot system, which was the result of more than five years and hundreds of millions of dollars of investment in research and development as well as the efforts of Tesla's team. But at the time, Xiaopeng Automobile responded that before and after Cao Guangzhi's implantation, the company did not find any violations of the law as claimed by Tesla.
But later in July, Cao Guangzhi admitted in documents filed with the court that he had indeed downloaded Tesla technical files, i.e., uploaded files containing the relevant source code to his personal iCloud account, but denied that he had transferred any Tesla Autopilot-related trade secrets to Xiaopeng. At this point, whether Cao Guangshi provided Xiaopeng with his backed-up Tesla source code data has become a point of contention in the case.
In fact, in addition to the field of autopilot, in many technology-intensive industries, there are also various disputes over theft of trade secrets due to 'talent mobility'. Even if there are no disputes, when a company's technical core staff jumps to another company of the same type, it will also cause various speculations and concerns in the industry. As the former vice president of self-driving research and development of Xiaopeng Automobile, Mr. Gu Junli.
It is understood that Gu Junli is a well-known machine learning expert in Silicon Valley, Ph.D., served as director engineer of international chip giant AMD during 2012-2016, responsible for the research and development of big data and deep learning software; and joined Tesla in February 2016, built the Tesla Machine Learning team, developed the Autopilot 2.0 product and realized a large-scale deployment in multiple generations of Tesla cars; 2017, the Tesla Autopilot team has been working on the development of the Tesla Autopilot 2.0 product. The large-scale deployment of the Autopilot 2.0 product and achieved large-scale deployment in multiple generations of Tesla cars; on October 23, 2017, Xiaopeng Automobile announced that Dr. Junli Gu, a famous machine learning expert, became the vice president of Xiaopeng Automobile Autopilot Research and Development, reporting directly to Xiaopeng Automobile Chairman He Xiaopeng. But in March of this year, the Valley Junli left, the successor is the former Qualcomm autonomous driving R & D team leader Wu Xinzhou.
It can be seen that in the past two years to Tesla, Apple, Qualcomm, represented by the Silicon Valley autopilot giants, and Azalea, Peng as a representative of the new car-making forces of the flow of personnel is becoming more and more. In addition to its "close" relationship with Tesla, how far has Xiaopeng developed its intelligent driver assistance system? Taking the P7, the current production model that best represents its technological prowess, as an example, we can roughly get a clear picture of the situation.
Autonomous driving requires ASIL-D hardware SoCs, and NVIDIA's Xavier and Orin can both meet the safety standards of ISO 26262 ASIL-D rated systems. Whether it is Tesla, Xiaopeng or Ideal, their first choice in preparing for self-developed autonomous driving is NVIDIA. Visible, NVIDIA's autopilot chip must have a lot of advantages.
Xiaopeng P7 equipped with XPILOT 3.0 automatic driving assistance system, the main chip is used is NVIDIA DRIVE Xavier, with L4 level of automatic driving computing ability, single chip computing power of 30 TOPs, power consumption of 30 watts. Although the arithmetic is not as good as Tesla's FSD 70TOPs performance, it is already 12 times the arithmetic of the EyeQ4 platform currently used by the mainstream, which is already enough to support its needs in L3 mass production and future testing at the L4 level.
In order to ensure the normal operation of the system, Xiaopeng P7 is simultaneously matched with 14 cameras, 5 millimeter wave radar and 12 ultrasonic sensors, forming the industry's only 360-degree dual-sensing fusion system. Among them, the far, middle and near forward-looking trinocular camera and the side camera cover more than 180-degree field of view, which can realize the lane line recognition of narrower curves, and the recognition of close-range congestion vehicles; in the parking scenario, the two side-forward cameras can recognize the side-forward parking spaces, and the industry's first time to realize the auto-parking without the need to cross over the parking space.
At the positioning level, Xiaopeng P7 uses Gaode's high-precision maps, with an accuracy of decimeters, which is the basis for realizing lane-level positioning; at the same time, the use of triple high-precision positioning hardware (GPS+RTK+IMU) enables global positioning accuracy of up to centimeters; in addition, the use of real-time positioning and map construction technology (SLAM) enables P7 to achieve a relative positioning accuracy of less than 0.3%, and the use of triple technology is complementary to each other. The triple technology complements and overlaps each other, which can effectively improve the stability and safety of autonomous driving under overpasses, tunnels or bad weather.
Small Peng car has such technical strength, and its emphasis on R & D investment is closely related. According to the prospectus data, as early as 2019, Xiaopeng Automobile invested 2.07 billion yuan in research and development of automatic driving and intelligent operating systems, and R&D expenses accounted for 89.2% of the company's total revenue, which is about double the 1.05 billion yuan invested in 2018, and is also one of the most important companies in the new car-making forces that pay most attention to R&D investment.
In terms of automatic driving data collection, Xiaopeng Automobile has so far collected about 2.9 million pictures with annotations (road character interference, rainstorms, nighttime, blockage, etc.), and also y learns and researches neural network technology, and develops natural language processing and natural language understanding functions.
In addition to this, at the end of last month, Xiaopeng Automobile also announced that it had reached an agreement with Guangzhou Kai Tak Investment Holding Company Limited, a wholly-owned enterprise of Guangzhou Development Zone Administrative Committee, for Kai Tak to provide 4 billion yuan of financing to support Xiaopeng Automobile to accelerate the expansion of its business, including the construction of a new intellectual base in Guangzhou Economic and Technological Development Zone and the purchase of factory equipment and other investment projects. This will also be the second self-built factory of Xiaopeng Automobile after the Zhaoqing factory.
On September 24, the first 100 units of Xiaopeng G3i from Xiaopeng Automobile were also officially shipped to Norway, which is the first batch of orders obtained by Xiaopeng Automobile in the European market, and is expected to reach Norway in November 2020. Norway has always been one of the countries with the highest acceptance of electric vehicles in Europe, and in August this year its electric car market share has reached 70.2%, up from 49% last year. 53% of the car market is occupied by BEV models. BEVs alone accounted for 53 percent of the market. As of August, 54,000 electric cars had been sold for the year. If Xiaopeng G3i can stabilize the market in Norway, its globalization process will be guaranteed.
Currently in the domestic market, whether it is the connection with Tesla, or the comprehensive strength of the Xiaopeng P7, it is the most direct competitor of the Tesla Model 3. Compared to previous years, Tesla's products in the market is completely ahead of its rivals in the same level of one or two generations of level of existence, but now in the development, upgrading and improvement of the driver assistance system, the car companies have more and more similarity, which may also be a technology to get rapid development of an important stage of the performance of it.