Text |Xu Caiyu Li Anqi.
Edit |Li Qin, Yang Xuan.
Nvidia, a chip giant with a market value of one trillion dollars, has earned $18 billion in revenue in a single quarter with the continued growth of the large model industry, but the automotive business is obviously their shortcoming - only 2$600 million. To this end, in August this year, Nvidia recruited Wu Xinzhou, the soul of the original Xpeng Motors' autonomous driving, to his command and began the rescue journey.
36Kr Poweron learned from a number of people familiar with the matter that on December 4, nearly 4 months after taking office, Wu Xinzhou held an all-staff meeting in China at NVIDIA's Shanghai office, pointing out the reasons for NVIDIA's "backwardness" in intelligent drivingAt the same time, it began to rapidly expand its staff, requiring 100 new employees to be hired within two months, "considering the impact of the Spring Festival holiday, this goal must be achieved within half a year." ”
A few days ago, Wu Xinzhou took the official recruitment information issued by NVIDIA in the circle of friends, showing that the NVIDIA China team is seeking 25 positions in five departments: autonomous driving software group, autonomous driving platform group, system integration & testing group, map &** group, and product group, with a total of hundreds of positions.
According to reports, nearly 300 NVIDIA intelligent driving personnel in China participated in the meeting, plus 100 new recruits, which are expected to form the core team of NVIDIA intelligent driving.
We are too slow to catch up. This is the core theme that Wu Xinzhou wants to express at this meeting, and a person familiar with the matter revealed that the new Zhijia helmsman expressed dissatisfaction with the past work rhythm of the NVIDIA team and asked the team to improve the work rhythm.
Wu Xinzhou's "firefighting operation" has been carried out intensively within NVIDIA. Some insiders said that after the arrival of Wu Xinzhou, the key work processes of the NVIDIA intelligent driving team have undergone visible changes to the naked eye: the speed of software iteration has been accelerated, vehicle testing has been increased, and the integration method has changed.
Previously, NVIDIA's L2+ intelligent driving products were still in a relatively rudimentary state, and an engineer said: "It's like a demo, and now everyone has to work hard to implement it." In this context, NVIDIA, which originally did not have a size week of 996, has begun to "work overtime" in order to catch up.
This is also the beginning of NVIDIA's intelligent driving team to catch up. Wu Xinzhou urgently needs a strong team to help him complete the implementation of the technology.
Wu Xinzhou once served as the vice president of autonomous driving at Xpeng Motors, where he comprehensively led the core team, technical architecture and business direction of Xpeng Intelligent Driving, and also helped Xpeng deliver high-end intelligent driving products for high-speed NOA and urban NOA in China at the earliest.
On August 25 this year, Wu Xinzhou left Xpeng Motors and joined chip giant Nvidia. According to Poweron, Wu Xinzhou has taken over NVIDIA's intelligent driving software business and reports directly to Nvidia founder and CEO Jensen Huang.
One of the main purposes of Wu Xinzhou's arrival in Shanghai this time is to interview high-level intelligent driving candidates.
A person close to Nvidia's recruitment said that Nvidia has opened some M5-M6 level positions, close to Ali's P9-P10 level, "basically at the team leader level, and the total salary package can be more than 5 million (yuan)."
People familiar with the matter said that the most urgent recruitment of Nvidia this time is the manager position, followed by algorithm and engineering development, and the recruitment process of the product department is the slowest.
In fact, the expansion of NVIDIA's intelligent driving team has already begun. In August this year, the team has started the first round of recruitment, and the number of NVIDIA China's intelligent driving team has exceeded 200 after the end. In the recent all-staff meeting in China, the number of team members has approached 300.
The recruitment model of casting a wide net and fishing for more fish has indeed rapidly expanded its combat team, but due to the shortage of manpower, Nvidia has not had too high restrictions on recruitment.
Relevant people close to Nvidia's intelligent driving recruitment said that those who have passed the academic qualifications, have an industry background, and can speak English as a working language will enter the interview process, as for the candidate's previous company or project experience, Nvidia has not made any requirements, "It is clear that companies that make L2+ products will pay special attention to whether the candidate has participated in mass production projects, and Nvidia is very special."
The importance of candidates is also reflected in the salary and other links. Not only did the human resources department call on employees to recommend themselves, but also said that "push people to the past first, and then solve the problem of money, and don't limit salary".
In addition to team building, Wu Xinzhou is also initiating some changes in the workflow of intelligent driving.
Some insiders said that in Wu Xinzhou's view, the biggest problem of the NVIDIA intelligent driving team is the unreasonable process. He said in front of all the employees in China: "We are like Duan Yu, NVIDIA has a very rich engineering foundation and engineers, but now there is no good workflow to achieve these things, and the six-vein divine sword is not effective."
The arrival of Wu Xinzhou has made visible changes in several main work processes of NVIDIA's intelligent driving: faster software iteration, increased vehicle testing, and changes in the most advanced integration mode.
The most significant is the sudden increase in working hours and workload. Nvidia has autonomous driving development teams in China, the United States and Germany, and the teams in the three places maintain the same ** library, "After he (Wu Xinzhou) came, in fact, there was no situation where anyone was involved in whom, everyone was rolled up, but the local team gave priority to solving local problems." ”
In terms of intelligent driving testing, Nvidia's vehicle testing has hardly left the company, and Wu Xinzhou, who has repeatedly emphasized the importance of "real car testing" in Xiaopeng, has broken the past model of Nvidia's intelligent driving.
According to Poweron, at present, Nvidia China has double-digit road test vehicles, all of which are Mercedes-Benz S-Class models, and is carrying out road tests in several cities such as Shanghai, and the number of test vehicles and cities is also increasing.
At the moment, Mercedes-Benz is NVIDIA's number one customer. Previously, in June 2020, NVIDIA and Mercedes-Benz officially announced their cooperation, and NVIDIA will provide AI software architecture for Mercedes-Benz's next-generation models, including autonomous driving software solutions, intelligent cockpits, etc.
At present, the team is all in Mercedes-Benz, and the first model to be launched will be the Mercedes-Benz S series. People familiar with the matter said that the first cities to land should be Beijing and Shanghai.
Under the soaring AI business, the single-quarter revenue of Nvidia's automotive business is even more dwarfed.
Nvidia's Q3 results for fiscal year 2024 show that revenue in the third quarter reached 181200 million US dollars, of which the two pillar businesses of data center and gaming have revenue of 145$100 million, 285.6 billion US dollars, accounting for 80% and 157%, while the automotive business contributes only 2$6.1 billion, accounting for about 14%。
More critically, NVIDIA's traditional model of providing chips + tool chains in the intelligent driving industry is being impacted.
First of all, the head of the new car company is generally brewing self-developed intelligent driving chips, and the person in charge of new car technology revealed that the self-developed intelligent driving chips of Weilai, Ideal and Xiaopeng are expected to be mass-produced in 2025 or even 2026, and the computing power of a single chip will not be a problem to 1000tops. For this, Nvidia CEO Jensen Huang has a clear prediction.
But for the remaining large number of car customers who do not make chips, Nvidia cannot sit idly by and watch the loss. They are facing the siege of domestic chip companies such as Horizon and Black Sesame. With its chip and algorithm capabilities, Horizon has captured more than 25 car companies in the mid- and low-end intelligent driving market, while companies such as Black Sesame are also actively complementing their algorithm capabilities.
Nvidia also sees the particularity of the automotive market, and they are no longer facing the customers of technology companies in the past, but manufacturing companies that lack software capabilities, which inevitably requires NVIDIA to provide richer algorithm capabilities. The arrival of Wu Xinzhou is obviously accelerating the implementation of NVIDIA's intelligent driving algorithm capabilities.
Mercedes-Benz is NVIDIA's number one customer, which is also a hard bone that Wu Xinzhou must gnaw.
Industry insiders who have been in contact with the project told Poweron that Nvidia and Mercedes-Benz will adopt a new cooperation model, that is, it will no longer be a traditional first-class business procurement cooperation, and the two sides will share the profits of the whole vehicle based on the cooperative model.
This is obviously a set of business ideas that match the trillion-dollar market value of Nvidia, and it is also expected to become a benchmark case for Nvidia and be widely promoted in the automotive market. But the premise is still that the three-year-long cooperation project is completed.
Entering 2024, domestic high-end intelligent driving products are entering cities from high-speed areas on a large scale, and the industry has a greater demand for chip companies' supporting algorithm capabilities. Horizon's next-generation chip J6 series is about to be launched, and its intelligent driving software algorithm in urban areas is also being developed rapidly. "Horizon is our most important opponent at home. An Nvidia person told Poweron.
On the other hand, although Huawei does not build cars, it also provides full-stack intelligent driving chips and software algorithm capabilities. Recently, Huawei and Changan announced that they will establish a new joint venture to inject Huawei's intelligent driving solutions and other technologies and open financing to the industry. Through capital binding, Huawei will obviously further eat up more market share of intelligent driving.
Under the fierce competition in the industry, there is indeed not much time left for Wu Xinzhou and the NVIDIA intelligent driving team.