With at least $11 billion in revenue over the next six years, Nvidia is pushing the self-driving front to 2026


The 2022 Nvidia GTC conference was still online, and huang didn’t drop any new nuks on autonomous driving this year, as the Orin chip released in 2020 only went into mass production this year, and the 1000T Atlan chip released in 2021 won’t be available until 2025.But don’t be disappointed. There’s no nukes, but there’s a lot of new commercial orders for its Nvidia Drive line, as well as major updates to its systems.Include:Acquired the order of BYD, the world’s second largest new energy vehicle company (based on Orin chip), released a new automated driving software and hardware development platform based on Atlan chip (Hyperion 9), launched a new global level high precision Map data acquisition platform (Drive Map),Nvidia’s autonomous driving business line is more software-focused, and puts the 2026 autonomous driving development platform ahead of the industry.Nvidia’s logic is that, having already led the industry in the computing power of autonomous driving chips, it needs to lead the industry in autonomous driving related software.In the next six years, the revenue of Nvidia Drive platform will be $11 billion. This kind of advance of Nvidia is reflected vividly in the order volume. According to the data disclosed in April 2021,Customer orders for the Nvidia Drive platform have generated more than $8 billion in revenue over the next six years, rising to $11 billion over the past year.If spread out, those orders could generate nearly $2 billion in annual revenue for Nvidia, a huge jump from $566 million in 2021 for the company’s entire automotive business.If you want to get a clear picture of nvidia’s vast customer base for autonomous driving, the following two charts are the best illustration.In the field of electric passenger vehicles, Nvidia has won orders from 20 of the 30 major global automakers (Mercedes-benz, JAGUAR Land Rover, Volvo, Ullipeng, Zhiji, Feifan, etc.) in the field of autonomous trucks,Nvidia has won orders from 7 of the 10 leading companies (Tucson, Chika, Navistar, etc.) and 8 of the 10 leading companies (Didi, Wenyuan, Xiao Ma, etc.) in the field of Robotaxi. In addition, in the field of autonomous driving data center, it is even more ridiculous.More than 28 of these partners among the 30 major enterprises have been acquired, most of which are based on Orin chip, so the popularity of this chip can be imagined. It is not an exaggeration to say that it occupies “half of the country” in the industry. The biggest uncertainty now is its supply capacity.This year GTC, Nvidia Orin’s partner circle continues to expand, adding: one Chinese and one American two electric vehicle companies: BYD is the world’s second largest new energy vehicle manufacturer (Huang Renxun’s words);Lucid is a U.S. -based premium electric vehicle company that is a direct rival to Tesla.Among these new cooperation, the most worthy one is BYD, which will sell more than 600,000 new energy vehicles in 2021, and some of them will sell more than 10,000 models per month.Jumping on the BYD bandwagon should provide nvidia with a huge boost in shipments of chips for autonomous driving.Still, BYD’s supplier strategy is sound.In terms of autonomous driving chips, BYD is also said to be cooperating with Huawei MDC, and byd has also invested in Local autonomous driving chip manufacturer Diping Line.In the future, BYD’s autonomous chip supply will be mainly composed of these three companies.But it’s a safe bet that the Nvidia Orin will be available on byd’s high-end model.According to byd’s plan, starting in the first half of 2023, some of its new energy vehicles will be equipped with Nvidia’s Orin chips and related hardware and software systems (known as the Hyperion 8 development platform) to enable autonomous driving and parking.Lucid, needless to say, is also a relatively well-known new force of smart electric vehicles in the United States. It has been on the market and has a market value of more than 40 billion DOLLARS. It is very promising for the future, but it is still overwhelmed by the big brother Tesla.On subsequent Lucid models, the DreamDrive Pro self-driving system will be powered by nvidia’s Orin, and its sensor suite will include 14 cameras, one lidar, five millimeter-wave radars, and 12 ultrasonic sensors.Other autonomous driving companies, such as Wenyuan Zhixing and YuanRong Qixing, are unicorns. Some of them specialize in Robotaxi, while others focus on both Robotaxi and mass production solutions. They are favored by capital recently, and they are also old friends of Nvidia and have in-depth cooperation based on the previous generation of chips.Although the volume is not yet started, they will contribute good shipments to Nvidia chips as their fleet expands in the future.Another very special skateboard chassis company that made it into nvidia’s GTC presentation this year is Slotting Technology.The skateboard chassis has recently gained a lot of traction with Rivian’s successful launch, and Nvidia is exploring what the combination of chassis and chips can do.There’s no denying that Nvidia’s ecosystem of chip business partners is in multi-flowering mode, with at least $11 billion in revenue over the next six years and growing over time.China will account for much of that revenue, and nvidia is expanding its tech support staff there, including those who develop self-driving software.Nvidia’s Drive software and hardware platform has become so popular among car and autonomous driving companies because of the hard power of its chips and the soft power of its Drive ecosystem and tool chain.As part of its “soft power,” Nvidia has perhaps the industry’s most complete self-driving reference development platform, DRIVE Hyperion, powered by its own powerful chips that can DRIVE a wide variety of sensors.At GTC, Nvidia updated the platform to Hyperion 9, which will go into mass production in 2026.Hyperion 9: Possibly the most comprehensive software and hardware development platform for autonomous driving Hyperion 9 is an open and modular platform designed from the outset to meet the highest level of functional safety and network security requirements.The platform will be based on dual Atlan chips (2000 TERabytes of computing power, equivalent to eight Orin chips), capable of L4 level of autonomous driving (including self-parking) and equipped with cockpit AI assistant applications.You may be unfamiliar with the performance of Atlan, to help you recall: the industry’s first SoC with a power of 1000TOPS;Based on the latest generation of Nvidia GPU architecture “Ampere”;It integrates the new ARM CPU core Grace, which has super data processing capacity.New deep learning and CV accelerators;Nvidia BlueField Data Processor (DPU);400Gbps wireless data transmission rate, equipped with a security gateway;Asil-d safety island;Sample film will be released in 2025.With such computing power, the number of sensing hardware that the Hyperion 9 can support increases dramatically. Data input from 14 cameras outside the car, nine millimeter-wave radars, three lidar and 20 ultrasonic sensors can be supported, as well as the in-car camera.By comparison, the Reference sensor design for the Hyperion 8 platform is less than stellar: 8 cameras, 4 fisheye cameras, 3 in-car cameras, 9 millimeter wave radars, and 2 Lidar.The Hyperion platform also has the strong advantage of cross-generation compatibility, such as seamless migration to Atlan from vendors currently developing on Orin.The Hyperion 9 platform includes computing architecture (Atlan chip), sensor suite, Nvidia Drive autonomous driving software system, and AI cockpit system. It is a comprehensive autonomous driving and intelligent cockpit software and hardware platform that can help automotive and autonomous driving companies build intelligent products efficiently.It’s this ability that keeps BYD from escaping the law of true fragrance, but byd is stuck with Hyperion 8, which won’t go into mass production until 2026, four years from now.03, Drive the Map: around the world Map and high precision data for automated driving it goes without saying that the importance of the data including the traffic data and Map data, the data can be used to Map and high precision, and can be used to feed automatic driving algorithm training center, can also be used to build automated driving simulation test system.Among them, it is not clear whether high-precision maps are necessary for autonomous driving. Mavericks like Tesla, who want to squeeze all the power out of vision technology, are not so active in the application of high-precision maps.But nvidia believes that if you do it, you do it all. So at GTC, Nvidia unveiled Drive Map, an extensible multimodal mapping engine that allows nvidia to collect high-resolution Map data, create and update high-resolution maps from around the world (legally permitted).According to Nvidia’s plan, its mapping fleet will complete detailed coverage of major roads in North America, Europe and Asia by 2024, with a total distance of 500,000 kilometers and accuracy of centimetres.The high-resolution map data collected will later be applied to millions of passenger cars.What’s more, the data can be fed into Nvidia’s own virtual-world simulation engine, Omniverse, to build a world-class digital twin known as the “metasverse.”In this digital twin, autonomous fleets can be trained.Why does Nvidia, a chip company, have the ability to map and build such a map engine?Drive Map came seven months after Nvidia announced in June 2021 that it was acquiring DeepMap, a high-precision mapping company, with the goal of enhancing nvidia’s autonomous driving hardware and software products with mapping and location capabilities.Drive Map supports field mapping by surveying vehicles and crowd-sourced fleet mapping, ensuring the accuracy, freshness and wide coverage of Map data.Drive Map contains multiple sensing and positioning layers with camera, millimeter wave radar and lidar data.The camera location layer consists of map attributes such as lane dividers, road markings, road boundaries, traffic lights, signs and posts.The millimeter wave radar positioning layer is the convergence point cloud of radar echo.It is particularly useful in poor light conditions and bad weather, where cameras and lidar can be blinded.The lidar positioning layer provides the most accurate and reliable representation of the environment, with an accuracy up to 5cm and a 3D representation of the environment, which is beyond the reach of cameras and millimeter-wave radar.With the help of this map, autonomous driving systems can rely on the detailed semantic information provided by the map to make and safely execute driving decisions in advance.Drive Map will be an important part of Nvidia’s “self-driving universe.”Nvidia: The benchmark for Autonomous driving chip companies, computing power in the early stage of the threshold, but now, the industry basically believes that the TOPS of computing power is no longer the only indicator to measure the quality of chips.As the technology advances, a number of autonomous chip companies have introduced a range of powerful chips, including toothpaste newcomer Mobileye.Even more, many enterprises support the need to cascade multiple chips to achieve large computing power, so large computing power is no longer an insurmountable “threshold”.Now industry to judge a automated driving chip of the products is good or bad, it depends on whether the chip in some automated driving the most attention on the processing of perception, decision-making link do timely and reliable enough to depend on the chip hardware software development ecology, behind the software support to the difficulty of the breadth and get started, depends on whether the chip tool chain is perfect,Depends on whether the chip test system is perfect.Moreover, the autonomous driving chip is no longer the chip itself, but to be closely combined with the autonomous driving system, the hard to hard, soft when it needs to have soft ability.As Nvidia showed at GTC, it has a steady “nuclear bomb” of self-driving chips and an edge in software, including Hyperion, a software and hardware development platform for autonomous driving, and Drive Map, a high-precision mapping engine.Most of the companies competing with Nvidia in China are still in the catch-up stage of hardware performance, and a few have caught up.But this is just the beginning. The software ecosystem, the development ecosystem, the tool chain, and so on are the key to succeeding in the competition. The focus is shifting.Otherwise, the trend of localization advantage and localization substitution will not move the entire domestic autonomous driving chip industry.

Leave a Reply

Your email address will not be published.