XPENG-W + NVIDIA = ?

Today is Wu Xinzhou's first day at NVIDIA.

On the afternoon of August 24th, the day before he reported to his new company, Wu Xinzhou, former Vice President of Xiaopeng Motors' autonomous driving division, retweeted a Weibo post by He Xiaopeng on his personal Weibo account. He Xiaopeng mentioned in the post, "It turns out that Mr. Huang holds a two-hour summary and outlook session for all employees on the day after each earnings report, and he often admits his specific mistakes for the quarter. It's not easy! The next opportunity lies in AIGC and AUTO, and our deeper collaboration is about to begin." The accompanying picture on Weibo is a group photo of He Xiaopeng, Huang Renxun, and Wu Xinzhou.

Wu Xinzhou officially announced his move by retweeting and saying, "Thanks to Xiaopeng for personally delivering me to Mr. Huang. According to Mr. Huang, I will still be working for Xiaopeng, but he won't have to pay me anymore." The cooperation between Xiaopeng Motors and NVIDIA is evident.

It is understood that Wu Xinzhou will serve as the Head of Automotive Products at NVIDIA, and typically, each business unit leader can report directly to Huang Renxun.

On the same day, NVIDIA released its earnings report for the second quarter of this year, with record-breaking quarterly revenue of $13.51 billion, a year-on-year increase of 101%. Specifically, the data center business set a revenue record of $10.32 billion, becoming the "strong support" for NVIDIA during this period; in the gaming business, revenue of $2.49 billion solidified NVIDIA's position as another pillar; as for the automotive sector, which originally accounted for nearly 5% of the business, its revenue share in the second quarter dropped to only 1.9%, and compared to the over 100% growth momentum in the previous two quarters, the year-on-year growth rate for this quarter was only 15%.

Regarding the unfavorable performance of the automotive business, NVIDIA CFO Colette Kress attributed it to an overall decline in automotive demand, especially in the Chinese market.

NVIDIA has always had high hopes for its automotive business. Huang Renxun once publicly stated, "In the future, NVIDIA's potential market size will reach $1 trillion, and in terms of specific segments, the revenue from the automotive industry will account for 33%, reaching $300 billion."

From 2015 to 2022, relying on its long-term accumulated technological advantages in the field of intelligent driving, NVIDIA has successively launched a series of high-performance chip products such as Tegra, Paker, Xavier, and Orin, and has made a large number of friends, including Tesla, Mercedes-Benz, Land Rover, Volvo, NIO, and Xiaopeng, all of whom are included in NVIDIA's "circle of friends."

However, NVIDIA's ambition is not just to be a chip supplier. Danny Shapiro, Vice President of NVIDIA's automotive division, once said, "NVIDIA is not just a technology supplier. From software to hardware, we hope to do more for the automotive industry and make future cars safer and more energy-efficient."

Previously, NVIDIA had directly intervened in the operations of original equipment manufacturers by providing high-performance chips to Mercedes-Benz and receiving a commission from Mercedes-Benz's sales revenue. For automakers, in order to gain the upper hand in negotiations with NVIDIA, they undoubtedly need chip design capabilities similar to Tesla's, which is something that most traditional automakers can hardly achieve in a short period of time.

New Role at NVIDIA

"As a technologist, what I most want to see is the implementation of technology." Wu Xinzhou revealed his vision when announcing his joining of Xiaopeng Motors.

At the beginning of 2019, Wu Xinzhou, who previously served as the Senior Director of Autonomous Driving at Qualcomm, joined Xiaopeng Motors as the Vice President of Autonomous Driving.

Wu Xinzhou graduated from Tsinghua University with a Bachelor's degree in Electronic Engineering. He later obtained his Master's and Ph.D. degrees in Electrical Engineering from the prestigious University of Illinois at Urbana-Champaign in the United States. After graduation, Wu Xinzhou joined Qualcomm in 2006 and worked there for 13 years, eventually becoming the Senior Director of Engineering and holding 160 U.S. patents. He was a key figure in Qualcomm's autonomous driving business.

According to the LatePost, Xiaopeng Motors started developing its own software after He Xiaopeng joined the company in 2017, but the early stages of their intelligent driving research and development were not smooth. It was not until Wu Xinzhou joined and restructured the technical architecture while integrating the team that Xiaopeng's self-developed path began to take shape. On the P7 model, which was launched in April 2020, Xiaopeng replaced the supplier's solution with their own algorithm, achieving this milestone more than a year earlier than expected and nearly two years ahead of NIO.

With Wu Xinzhou's arrival, Xiaopeng Motors gradually gained the reputation of being a "technology-oriented" company in the field of autonomous driving. The advantages Xiaopeng Motors demonstrated in autonomous driving translated into market competitiveness and contributed to an increase in sales. At one point, monthly sales of the P7 model reached nearly ten thousand units, making it the first pure electric new energy vehicle to achieve mass production of 100,000 units.

Wu Xinzhou, who has been labeled as the person responsible for giving Xiaopeng the "intelligent driving" label, is exactly the talent Huang Renxun was looking for.

Although NVIDIA excels in hardware manufacturing and has made great strides in autonomous driving computing, it has obvious shortcomings in building services and ecosystems. Apart from leading automakers having sufficient self-developed capabilities, most automakers still hope to have a complete package solution, which is not NVIDIA's strong suit, but it is where Wu Xinzhou excels.

For example, Horizon, which is weaker than NVIDIA in terms of product strength, invests heavily in building an ecosystem to enhance its competitiveness. Horizon not only provides perception algorithms but also helps customers solve basic engineering problems, significantly reducing the workload of customer development and minimizing the adaptation cycle of algorithms and software. Therefore, the collaboration between automakers, Tier 1 suppliers, and Horizon allows projects to progress very quickly, from initiation to SOP (start of production).

Now, with the practical experience in intelligent driving software gained at Xiaopeng and his rich experience in implementation, Wu Xinzhou, who has joined NVIDIA, can also create matching software based on their excellent chips.

NVIDIA's Automotive "Circle of Friends"

Typically, a new role involves two types of demands: personal demands and organizational demands. However, in Wu Xinzhou's case, there may be a third demand, which is the deep cooperation between Xiaopeng Motors and NVIDIA.

According to Huang Renxun's plan, the future automotive business will be one of NVIDIA's three main pillars, alongside data centers and gaming. This means that the automotive business needs to reach the same level as the gaming and data center businesses. In the second quarter of 2023, the revenue from NVIDIA's data center, gaming, and automotive businesses were $10.32 billion, $2.49 billion, and $253 million, respectively. If the automotive business wants to surpass the $10 billion threshold, there is still a long way to go. From 2015, NVIDIA has entered the field of in-vehicle SoCs and in-vehicle computing platforms, providing fundamental computing capabilities for autonomous driving.

Currently, in the race for computing power around autonomous driving chips, NVIDIA has achieved the "top configuration" in the industry. NVIDIA releases a vehicle-grade SoC chip approximately every two years, continuously increasing the level of computing power. The Xavier chip released in 2020 has a computing power of 30 TOPS, the Orin chip released in 2022 has a computing power of 254 TOPS, and the Thor chip has a computing power of 2000 TOPS.

It is worth mentioning that the new generation autonomous driving chip, Thor, released by NVIDIA, is not only a chip for autonomous driving, but also a chip for intelligent cabins. It can be used for multiple purposes, achieving integration of the cabin and driving, becoming the central computing unit of the vehicle. The Thor chip has two notable features: one is its high computing power, which is twice that of Altan and eight times that of Orin; the other is its ability to integrate various automotive chip functions. It is widely believed in the industry that with the release of the Thor chip, NVIDIA is challenging the intelligent cabin chip race led by Qualcomm. The Thor chip is planned to be mass-produced in 2024 and installed in vehicles in 2025. Xike will be one of the first customers of Thor.

Currently, NVIDIA, including the third-generation Orin autonomous driving chip, has achieved a good market share. According to Sullivan's statistics, in 2022, NVIDIA's shipment volume accounted for 82.5% of the global high-performance autonomous driving chip market share.

Relying on its long-term accumulated technological advantages in intelligent driving computing power, NVIDIA has also established a large "circle of friends" in the automotive field.

Guosen Securities pointed out in a research report that NVIDIA's automotive customers can be roughly divided into three categories: first, new car manufacturers, including NIO (ET5, ET7), Xiaopeng Motors (P5, P7, G9), Li Auto (X01), WM Motor (M7), SAIC Zhiji, R Automobile, FF, Lucid Group, etc.; second, traditional automakers, including BYD, Mercedes-Benz, Jaguar Land Rover, Volvo, Hyundai, Audi, Lotus, etc.; third, autonomous driving companies, including GM Cruise, Amazon Zoox, Didi, Volvo Commercial Vehicles, Kodiak, TuSimple, Zhijia Technology, AutoX, Pony.ai, Wenyuan Zhixing, Yuanrong Qixing, etc.

The reason why NVIDIA can quickly acquire a large number of customers lies in the fact that there are no more chip options available in the market for enterprises that need to develop intelligent driving at Level 3 and above.

Reaching out again, Huang Jiaozhu wants both hardware and software

Huang Renxun firmly believes that the automotive industry will be NVIDIA's next billion-dollar business. However, NVIDIA, which started with hardware, faces competition from more experienced car manufacturers in the automotive industry. For autonomous driving, which represents the future trend, it is necessary to have "software capabilities" that match the "hardware capabilities" extremely well.

This poses requirements for NVIDIA's grand plan in the automotive industry: to control both the brain and the nerves of autonomous driving. The relationship between autonomous driving chips, platforms, and vehicle models is not easy to understand. NVIDIA provides a vivid analogy: the car is the body, the autonomous driving platform is the nervous system, and the autonomous driving chip is the brain. However, NVIDIA cannot rely solely on chip hardware to achieve this goal.

Huang Yefeng, a senior industry analyst in the semiconductor industry, said that from a broad perspective, there are two parts related to autonomous driving cars. First, the data center needs to process data, train AI, simulate digital twins, and build virtual worlds. Second, the AI computer on the car itself processes various sensor data to perceive the driving environment and achieve autonomous driving.

NVIDIA's overall autonomous driving development and deployment platform is called Nvidia DRIVE. From the perspective of NVIDIA's platform layer, DRIVE mainly involves Nvidia AI and Omniverse. Currently, NVIDIA's three major platforms are Nvidia HPC, AI, and Omniverse.

It is understood that Omniverse is a virtual simulation platform that can reflect the real size of objects and comply with physical laws. It can help car companies conduct realistic simulations before product development or factory construction, and expose various potential problems to the maximum extent during the early design stage.

Omniverse has cooperated with multiple car companies:

Volvo and General Motors use Omniverse USD Composer to connect and unify their asset workflows. The latter also assembles car components into digital twin cars in a virtual environment. Mercedes-Benz and Jaguar Land Rover engineers use Driver Sim in Omniverse to generate synthetic data to train AI models and verify active safety systems through virtual NCAP driving tests. The former also uses Omniverse to build, optimize, and plan assembly lines for new models. Toyota uses Omniverse to build digital twins of its own factories.

Wu Xinzhou's Full Stack

In 2021, NVIDIA's automotive business revenue was $566 million, accounting for 2% of the company's total revenue. In 2022, automotive business revenue increased to $903 million, accounting for approximately 3%. In the second quarter of 2023, the proportion dropped to only 1.9%.

This is far from the two goals set by Huang Renxun. One is a software company in the intelligent automotive field, and the other aims to achieve a 30% market share in the field of autonomous driving.

Like Qualcomm, compared with data center and gaming revenue, NVIDIA's automotive business revenue is relatively small. Although the autonomous driving chip Orin has been mass-produced and delivered since 2022, the scale effect has not yet been demonstrated. The reason is that in the past, chip manufacturers mainly provided hardware development reference designs to downstream customers to help them quickly provide overall software and hardware solutions for car companies. Nowadays, software is becoming a new entry barrier. For this reason, in 2021, NVIDIA acquired the start-up company DeepMap, which specializes in high-precision mapping. By integrating their technology solutions, NVIDIA aims to expand its ability to provide full-stack solutions.

The turning point occurred in the collaboration with Mercedes-Benz.

In June 2020, Mercedes-Benz announced a 10-year partnership with NVIDIA to develop fully autonomous vehicles. Leveraging their respective expertise in high-performance computing and high-end automotive manufacturing, the two companies are jointly creating a new software-defined car. NVIDIA is responsible for providing the software and hardware full-stack solution, collaborating with Mercedes-Benz. This collaboration is considered by NVIDIA as the "largest single business model transformation" in the company's history, as both parties will share the revenue from future user purchases of features and subscription services.

Public data shows that for every $1 of hardware sold by NVIDIA, software and related system providers will generate $8 in revenue. Huang Renxun has repeatedly emphasized that "the business model of car manufacturers will fundamentally change. By 2025, many car companies are likely to sell cars at or near cost price, primarily providing value to users through software." In Huang Renxun's view, the most important factor for future car buyers is software that is constantly iterated, upgraded, and enhanced.

Industry insiders believe that NVIDIA recognized the opportunity of artificial intelligence many years ago and invested heavily in developing a complete AI stack, including software, services, and hardware. The intelligent automotive field is no exception. With the technological cooperation between Xiaopeng Motors and Volkswagen Group, NVIDIA sees opportunities but also feels the risks.

It is rumored that due to the lack of full-stack software capabilities, the collaborative development process between NVIDIA and Mercedes-Benz is not going smoothly, and the project is experiencing delays. NVIDIA and Mercedes-Benz have had to introduce solution providers that can offer Orin platform development. In addition, Mercedes-Benz's projects in China are even seeking help from other suppliers. Mercedes-Benz is continuously strengthening its software research and development capabilities in China. It is reported that to ensure the strengthening of R&D capabilities in China, Mercedes-Benz expects its Chinese R&D team to reach a size of 2,000 people by the end of 2023, nearly doubling its size compared to 2020.

The reason Huang Renxun recruited Wu Xinzhou under his command may be because he values his experience in full-stack systems, especially in software mass production.

For NVIDIA, whose market value has exceeded $1 trillion, it is not satisfied with just supplying components. At the 2023 Snowflake Summit, Huang Renxun stated that artificial intelligence is entering the era of Software 3.0 (relying on data, algorithms, and computing engines). Huang Renxun believes that with the shift from the "hardware-sales-software" strategy of the past to the "software-sales-hardware" strategy, a new software landscape will emerge. NVIDIA hopes to sell more software that runs exclusively on its GPUs.