On March 23, the annual NVIDIA GTC 2022 Conference kicked off.
The official not only issued the much-anticipated Hopper architecture H100 acceleration card, but also launched the CPU processor Grace designed for AI and supercomputing, and the AI computing system "DGX H100". But I have to say that at this year's GTC, automatic driving undoubtedly occupies a relatively large space.
Looking at global car companies, those who want to get involved in L3 or above autonomous driving, few can jump away from NVIDIA chips or automatic driving platforms. And Nvidia also lived up to the expectations of all parties, releasing a lot of good news in the field of autonomous driving at the 2022 GTC.
Not long ago, Nvidia released its annual report for fiscal 2021. Full-year revenue reached a record high of $26,914 million, up 61% year-over-year, and Non-GAAP net profit for the full year was $11,259 million. In terms of stock prices, Nvidia hit a peak of $333.76 in November 2021, a significant increase from $59.02 at the beginning of 2020.
From its own business point of view, the current sales pillar of NVIDIA, which started with GPUs, is still the game and data center sector, but it has to be admitted that with The Orin chip, the autonomous driving business is becoming Nvidia's new sales and profit growth point.
Orin is not yet available, Andlan is already unveiled
At GTC 2022, Nvidia officially announced that the autopilot chip Orin was officially put into production this month. For many car companies that have long announced that they will be equipped with Orin chips, it can be regarded as a reassuring pill. At the same time, Nvidia has launched drive Hyperion 9, a new generation of autonomous driving platform based on Atlan chips, and plans to mass-produce in 2026.
As a new autonomous driving platform, Hyperion 9 can support L3 autonomous driving and L4 parking in parking lots. The most obvious improvement compared to the 8th generation is the significant increase in the number of supporting perception hardware, including 14 cameras, 9 millimeter wave radar, 3 lidars and 20 ultrasonic radars in the interior part of the car, which can support 3 cameras and 1 millimeter wave radar. Specifically, since Hyperion 9 has 17 more perceived hardware, it will process more than twice as much data as the Generation 8 platform and twice the performance of the Generation 8 platform.
For the Hyperion 9 platform, the Atlan chip is the biggest highlight. Compared with NVIDIA's upcoming Orin chip, it has undergone a complete change in the overall architecture of the chip. The Atlan chip will integrate the Grace-Next CPU, the Ampere-Next GPU unit, and for the first time, the Bluefield Data Processing Unit (DPU), which will have better monetization in TERMS for AI computing and autonomous driving capabilities. Especially in terms of the computing power that measures the chip's power, the Atlan chip will also achieve a significant jump.
From an internal point of view, the computing power will be from Xavier's 30 TOPS, to Orin's 254 TOPS and then to Atlan's 1000TOPS, Nvidia has also made great progress in chip computing power; and looking at the global market, the current Mobileye, Huawei, horizon can mass-produce single chip is basically at the level of tens of TOPS. Horizon's latest journey 5 released last year, a single chip AI hashing power has just reached 128 TOPS. As for the Qualcomm Snapdragon Ride that can compete with Orin, the hash rate can reach up to 700TOPS, but this chip needs to be released in 2023.
Of course, computing power is not the only indicator of a chip, energy consumption, data transmission bandwidth, etc. also affect the performance of the chip. However, Atlan's ultra-high computing power has undoubtedly laid a solid foundation for its performance.
Multi-modal maps, NVIDIA's new weapon?
For the autonomous driving business, the importance of high-precision maps is self-evident. At GTC 2022, Nvidia also released the DRIVE Map multimodal map engine, which will assume the role of high-precision map collection, production, and update, and become a new weapon for Nvidia to strengthen its leading position in the field of autonomous driving.
According to Nvidia, DRIVE Map mainly collects data from cameras, radar, and lidar. Among them, the camera corresponds to the data of the visual perception level, including data such as lane separators, road markings, road boundaries, traffic lights, signs and benchmarks; millimeter-wave radar, which is mainly responsible for collecting aggregate point cloud data of radar echoes; and lidar, which can provide accurate and reliable 3D environmental data with an accuracy of up to 5 centimeters.
In order to meet the needs of different users and continuously improve the accuracy of high-precision maps, NVIDIA will provide two map engines, The True Value Mapping Map (DeepMap) engine and crowdsourced fleet map, to achieve the best balance between surveying and mapping accuracy, accuracy and cost and scale.
According to Nvidia's plan, by the end of 2024, it will complete the creation of drive maps on major roads in North America, Western Europe and Asia, with a total road mileage of 500,000 kilometers. Of course, for the domestic market, which is more sensitive to map data and information collection, NVIDIA may need to cooperate with more domestic car companies and map dealers to expand its circle of friends to avoid the risks and resistance faced in the process of high-precision map collection.
While Nvidia gradually expands its circle of self-driving circles in the automotive industry, high-precision maps, especially crowdsourced maps, may become a new source of revenue and profit for Nvidia.
The circle of friends that continues to grow,
Can you let NVIDIA breathe a sigh of relief?
At this year's GTC, the official announcement of cooperation with BYD is undoubtedly a highlight in addition to the new product. As the largest new energy vehicle company in China and the second largest in the world, BYD will start production of vehicles equipped with the DRIVE Orin computing platform in the first half of 2023. Previously, including Weilai ET7, IM Zhiji Automobile and WM M7 will be equipped with four Orin chips, with a final force of more than 1000TOPS. In addition to domestic car companies, many multinational car brand and autonomous driving technology companies will use NVIDIA's Orin solution.
However, for Nvidia, it can bring so many vehicle companies under its own umbrella, and more or less there are not many products to choose from on the market, especially the AI chips required for automatic driving above the L3 level, Nvidia's solution has been recognized by more vehicle companies. But in addition to Nvidia, there are currently two forces that deserve NVIDIA's vigilance.
First of all, the traditional chip giant will not sit back and watch the big cake of car autonomous driving being swallowed by NVIDIA alone.
Mobileye, which was previously acquired by Intel, is also seeking a separate IPO. This IPO not only has Mobileye's own business, but also includes Intel's projects related to vehicle autonomous driving. In the future, in addition to mobileye's continued focus on system-level products such as autonomous driving platforms, mobileye may also enter the Robotaxi market.
As for domestic, companies such as Horizon and Black Sesame Intelligence are also growing rapidly. For domestic car companies, the previous case of the United States exerting extreme pressure on Huawei is in front of them. In terms of chip design, finding an alternative domestic solution for yourself is a major strategy to ensure the security of your own supply chain.
Secondly, car companies themselves also have the impulse to design chips.
Tesla has previously shown its own self-driving exclusive chip, compared to Nvidia's general chip, the exclusive chip has more advantages in the release of effective computing power. However, the investment in self-developed chips is huge, not only the investment of funds, but also the investment of up to 3 to 5 years of research and development and production cycle, and the chip design itself will also face the risk of design failure. In addition to Tesla, domestic Weilai and Xiaopeng have previously been rumored to form a chip design team. As for Huawei, Ali, and Baidu, they all already have high-process chips to market.
At present, we judge the quality of the chip, and the computing power is a very important indicator. For car companies, in the case that L4-level automatic driving is difficult to fully land in the short term, let the automatic driving control chip meet the recognized L4 requirements in terms of computing power, and then continuously unlock the automatic driving function in specific scenarios through continuous software upgrade in the later stage, so as to continuously approach the L4-level automatic driving function.
Therefore, while chip companies continue to improve the computing power of chips, it is more important to give car companies a more open and efficient ecology. In this regard, Nvidia is the market leader with its higher software openness and the need for R&D, such as NVIDIA's own CUDA and TensorRT, but the catch-up behind it is also menacing.
At last
At present, it is the best era of Nvidia, and the accumulation of GPU and AI technology in the early stage has allowed Nvidia's market value to increase significantly, from only $59.02 in early 2020 to the peak of $333.76 in November last year, becoming a new upstart in the global chip market.
But the high place is not cold: Nvidia's market value is higher than the sum of Intel +AMD+ Xilinx, but from its acquisition of ARM has been unanimously opposed by the industry, and at the same time AMD's acquisition of Xilinx and Intel's acquisition of the tower can be found that the pattern of the global chip industry is reconstructing. Intel and Qualcomm are also looking for a new position in the smart electric vehicle market through their own ways.
As for NVIDIA, it is far from the time when it can lie on the merits of history. With autonomous driving/driver assistance functions becoming standard on smart electric vehicles, the future autopilot chip market will be a big temptation that no one can stop.
Remember to click to share, like and watch, recharge me!