Recently, there has been big news in the autonomous driving industry.
The first is that Apple has shelved and canceled all development plans for self-driving electric vehicles. The project, which took 10 years and cost more than $1 billion, has sunk into the iceberg. According to reports, the project will begin to be tapered in size, while numerous members of the automotive team will be relocated to the AI department.
While Apple's autonomous driving is experiencing a vigorous "cutting pain", NVIDIA's intelligent driving team is "recruiting" in a low-key manner. Luo Qi, the former head of the overall software architecture and regulation and vehicle interaction technology of the intelligent driving L2+ business, joined the NVIDIA Automotive Division as engineering director, responsible for **, planning and control. And his reporter is Wu Xinzhou, the head of NVIDIA's automotive division who joined not long ago.
The outside world may focus more on NVIDIA's business in automobiles, and the last time it was paid attention to was Wu Xinzhou, the soul of Xiaopeng intelligent driving. In fact, NVIDIA's intelligent driving business is not a new business, but the results of the team's project do not seem to be well known. So how is NVIDIA's intelligent driving business developing?
Strong Nvidia automotive business
Nvidia has three automotive platforms, namely Nvidia Drive Infrastructure, Nvidia Drive AGX, Nvidia Drive Concierge, and Nvidia Drive Hyperion.
NVIDIA Drive Infrastructure is a complete workflow platform for data ingestion, curation, labeling, training, and validation. NVIDIA DGX systems provide the computing power needed to train and optimize deep neural network models at scale. NVIDIA Drive Constellation enables physics-based** on an open hardware-in-the-loop platform to test and validate autonomous vehicles before they hit the road.
The scalable, software-defined NVIDIA Drive AGX platform delivers advanced performance to enable autonomous vehicles to process massive amounts of sensor data and make real-time driving decisions. The open NVIDIA Drive software stack also enables developers to build perception, mapping, planning, and driver monitoring capabilities using redundant and diverse deep neural networks (DNNs). Through continuous iteration and over-the-air updates, the platform has become increasingly powerful.
With NVIDIA Drive Concierge, vehicle occupants have access to a range of intelligent services based on NVIDIA Drive ix and NVIDIA Omniverse Ace, **ATAR Cloud Engine. NVIDIA Drive Chauffeur is an AI-assisted driving platform based on the NVIDIA Drive** SDK that enables end-to-end driving. It can also provide proactive safety features to intervene in dangerous situations if you want to drive yourself.
NVIDIA Drive Hyperion is a complete development platform and reference architecture for designing autonomous vehicles. This architecture accelerates development, testing, and validation by integrating NVIDIA Orin-based AI computing with a complete suite of sensors. Drive Hyperion has a complete software stack for autonomous driving (DRIVE**) as well as over-the-air updatable driver monitoring and visualization (DRIVE IX). This allows new features and functionality to be added throughout the vehicle's lifecycle.
Judging from the distribution of such a platform, NVIDIA has actually accumulated a small amount in the software support of intelligent driving. Also as a computing power boss, Nvidia does not lack hardware support, but why hasn't Nvidia been able to "dominate" the autonomous driving industry? Some industry insiders told icviews that part of the reason why it could not be implemented was because Nvidia did not know enough about the automotive business and lacked a key No. 1 role. This has led to a situation where although NVIDIA has technology in the car, the intelligent driving team is still "weak".
The "weak" NVIDIA intelligent driving team
Although automotive is listed by NVIDIA as one of the three major businesses (data center, gaming, and automotive), its revenue share is far from the other two businesses. Nvidia's fourth-quarter 2023 financial report shows that the revenue of the automotive business is 2$8.1 billion, up 8% sequentially and down 10% year-over-year, accounting for 12%。
In the automotive business, autonomous driving teams are even more "weak".
In 2015, NVIDIA began the development of autonomous driving solutions. As the absolute leader in the chip industry, NVIDIA's intelligent driving directly won the industry's top users as soon as it debuted, and reached cooperation intentions with several major overseas car companies such as Mercedes-Benz and Jaguar Land Rover.
In June 2020, NVIDIA and Mercedes-Benz officially announced a cooperation under which NVIDIA will provide AI software architecture for Mercedes-Benz's next-generation models, including autonomous driving software solutions, intelligent cockpits, etc. After the partnership, Mercedes-Benz executives often appeared at NVIDIA's GTC (developers) conference. Huang has also made several high-profile appearances for Mercedes-Benz platforms, frequently appearing in promotional materials for Mercedes-Benz's next-generation S-Class models.
The cooperation between Nvidia and Mercedes-Benz is neither charged according to the project nor according to the IP license, but in addition to the basic research and development costs, it is divided according to the sales of Mercedes-Benz's new products. The cooperation with NVIDIA is a new attempt in the history of automobile development, and in the past, OEMs would not use the model of sharing profits according to sales volume to cooperate with leading companies. However, Nvidia has not shown the corresponding strength. It has been reported that when NVIDIA's intelligent driving team showed Mercedes-Benz the assisted driving ability of the real car, the side and rear reversing tried several times without success. The cooperation between Nvidia and Mercedes-Benz has become less stable, and Mercedes-Benz once asked for the introduction of new ** merchants.
NVIDIA's intelligent driving team needs a "savior", and this person may be Wu Xinzhou.
Wu Xinzhou graduated from Tsinghua University with a bachelor's degree, and then entered the University of Illinois at Urbana-Champaign (UIUC) for further studies, and received a Ph.D. in electrical engineering in 2004. He then joined Flarion, a communications startup, as a technical engineer responsible for PHY MAC system design and design. At Flarion, he was involved in the development of several technologies and joined Qualcomm with the acquisition of Flarion by Qualcomm. After that, he accumulated 13 years of technical experience at Qualcomm, and since 2015, he has been responsible for Qualcomm's autonomous driving research and development as director of engineering, focusing on high-definition maps, cameras, radar, and deep learning.
With such a number one position joining, is there any hope that the tech-savvy Nvidia will take off again on the automotive track?
Nvidia's ambition is 30 billion
Chips will always be NVIDIA's competitiveness.
Even if it hasn't been unbeatable, Nvidia has never been a nobody in the automotive industry. In 2019, Nvidia released a new generation of autonomous driving chip product Orin, and at that time, all autonomous driving companies above the L4 level in China became buyers of NVIDIA. Now, Nvidia's list of automotive partners can find almost all "star" car companies. For example: Mercedes-Benz, Jaguar, Land Rover, Volvo, Hyundai, BYD, Polestar, NIO, etc.
Compared with other companies, NVIDIA has a huge algorithm ecological barrier, as early as 2015, Nvidia proposed the use of "end-to-end" neural network for autonomous driving, and there is a preliminary demo. As a leader in autonomous driving, Tesla has brought this end-to-end AI autonomous driving system to life. Musk tested Tesla's FSD Beta V12 live on the streets of Silicon Valley, attracting more than 40 million people online**. Although Nvidia and Tesla have different specific solutions to achieve "end-to-end", they have the same goal of using biomimetic artificial intelligence as the ultimate goal, so that every decision made by the autonomous driving system is not bound by rules. The fact that the "layman" Nvidia was able to put forward such an idea in 2015 is enough to prove its technical strength.
What's more, NVIDIA's automobile-related platforms that have been laid out, combined with its own hardware chips, have enough potential to rewrite the autonomous driving market pattern.
Some lidar companies said that on NVIDIA's drive computing platform, it is easier to retrieve all their lidar plug-ins; If a company is using a platform such as X**ier (their previous generation of autonomous driving computing chips) to make solutions, then their own perception and other software can be compatible with any of their computing units.
Nvidia is entering the automotive market in a "moisturizing and silent" way - letting the software of auto parts manufacturers be used on their hardware basis. NVIDIA's full-stack system solutions provide a sharper insight into the implementation of autonomous driving. The full-stack system covers multiple aspects such as hardware, software, and ecosystem, making NVIDIA more competitive in the field of autonomous driving. In addition, NVIDIA is committed to building a solid ecosystem where more autonomous driving sensor hardware and software capabilities can take root and grow in this system.
Although the current revenue is not high, Nvidia founder Jensen Huang is optimistic about the prospects of the automotive business. Huang has said that in the future, Nvidia's revenue will reach $1,000 billion, of which the automotive business can account for 30%. This means that Nvidia plans to get 30 billion in revenue in the automotive field.
Since Wu Xinzhou, the former vice president of autonomous driving of Xpeng Motors, joined NVIDIA, dozens of people have joined or will soon join the NVIDIA intelligent driving team, including Liu Lan Gechuan, the former head of autonomous driving AI of Xpeng Motors, Parixit Agera, the engineering VP of the former Xpeng Motors North America team, Han Feng, director of multimodal perception fusion algorithms, and Houman T**akoli, director of software architecture.
At the end of 2023, NVIDIA will open recruitment for five departments in China, including the autonomous driving software group, autonomous driving platform group, system integration and testing group, map and ** group, and product group, with a total of 25 positions. As vice president of NVIDIA's automotive business, Wu requires job seekers to have a strong professional background; Strong self-drive and the pursuit of excellence in technology and products. The opening of so many jobs in China at one time shows that Nvidia wants to focus on autonomous driving in China.
In 2024, NVIDIA announced that it has reached a cooperation agreement with four Chinese companies to jointly promote progress in the field of intelligent driving. The four car companies are Ideal, Great Wall Motor, Zeekr and Xiaomi.
It is reported that in the strategic cooperation with Li Auto, Nvidia chose to purchase the NVIDIA Drive Thor centralized on-board computing platform as the intelligent driving system platform for its next-generation models. The application of this platform will help to improve the ideal level of intelligent driving, and bring users a more convenient and safe travel experience. The other three said that the next generation of autonomous driving systems will all use Nvidia's NVIDIA Driveorin computing platform.
No matter how much this has to do with Mr. Wu, it's clear that Nvidia's efforts in China have paid off. In this track that Apple silently withdrew, can Wu Xinzhou lead Nvidia, which has a market value of 2 trillion, to the **? Both Huang and Musk want to know the answer.