(Report producer/author: Great Wall Securities, Hou Bin)
1. Industry giants in the era of artificial intelligence
Founded in 1993 as a designer and manufacturer of computer processors, NVIDIA has grown over three decades to become a leader in artificial intelligence computing. As the first chip company to invent the GPU, NVIDIA's 1999 GPU redefined computer graphics, stimulated the development market of PC games, promoted the development of modern artificial intelligence, and injected strong impetus into the innovation of the metaverse. At present, NVIDIA has developed into a full-stack computing company, with CPU, GPU, DPU as its main business, providing competitive products and services in data center, gaming, professional visualization, automotive and other fields, and its main customers include Microsoft, Google, Amazon, Alibaba and other world-renowned enterprises.
1.1 Take continuous innovation as the core to promote the continuous development of the company
Innovation as the body, cooperation as the wing, help NVIDIA to lay a leading position in the industry. Since its establishment in 1993, NVIDIA has not only released innovative products to promote the continuous development of the computer field, but also actively sought cooperation and mergers and acquisitions, gradually establishing its leading position in the industry: In 1995, NVIDIA's first product, NV1, was launched; The world's first 128-bit 3D processor, RIVA128, and the world's first GPU, GeForce 256, were released in 1997; In 1999, the company was successfully listed on NASDAQ; In 2001, it entered the integrated graphics market with nForce and launched the industry's first programmable GPU, GeForce3, and in the same year, NVIDIA was included in the S&P 500 index; In 2006, NVIDIA introduced the revolutionary CUDA architecture, enabling scientists and researchers to harness the parallel processing power of GPUs to tackle the most complex computational challenges; In 2008, NVIDIA introduced the Tegra mobile processor, which consumes 30 times less power than ordinary PC notebooks, consolidating NVIDIA's position in the industry; In 2009, at the first GPU Technology Conference, the Fermi architecture was launched; In 2016, the company launched the NVIDIA DGX-1, the first desktop deep learning supercomputer that can enhance AI applications, while the NVIDIA DRIVE released in the automotive industry enables powerful in-vehicle artificial intelligence, putting the automotive industry on the road to autonomous vehicles, and the two products further drive the AI revolution; In 2018, the company released the Turing architecture with real-time ray tracing to redefine computer graphics;
By 2023, the company will launch innovative products in a variety of fields, including but not limited to artificial intelligence, high-performance computing, robotics, autonomous driving, healthcare, professional vision, etc. Since its establishment 30 years ago, the company has been deeply engaged in technology and innovative application. The company's core business, GPU computing, is continuing to meet the needs of all walks of life for powerful computing power, and driving social innovation and progress.
While continuously launching new products and leading industry changes, NVIDIA has expanded its own strength through a series of capital acquisitions, expanded the company's development space and laid a solid foundation for today's industry position.
1.2 FY24 Q1 results greatly exceeded expectations, and the company's revenue grew rapidly sequentially
On May 24, 2023, the company released its fiscal year 2024 Q1 results report. In fiscal 2024, Q1 revenue was $7.19 billion, up 19% sequentially; Non-GAAP net income of $2.71 billion; Sequentially increased $539 million. GAAP net income of $2.04 billion increased 26% year-over-year and 44% sequentially. The company's revenue is mainly composed of four major segments: data center, gaming, professional visualization, and automotive. All four segments grew sequentially in the first quarter, with data center, gaming, professional visualization, and automotive achieving revenue of $42.8, $22.4, $2.95 million, and $296 million, up 18%, 22%, 31%, and 1% sequentially, respectively.
NVIDIA's financial report released in Q1 of fiscal 2024 showed that the company's gross margin (non-GAAP) was 66.8%, down 0.3 pct year-over-year and up 0.7 pct sequentially. The decrease in gross margin from a year ago was mainly dragged down by lower gross margins in the gaming and automotive businesses, and even though the gross margins of the data center business were higher, the overall average was still lower year-over-year. The sequential increase was due to a decrease in gaming expenses, while the data center business maintained a high gross margin due to the impact of Hopper-based products. The company's gross margin under non-GAAP caliber in Q1 2024 was 66.8%, higher than the 38.4% and 50% of competitors Intel and AMD in the same period; From the combined perspective of Q1 of fiscal year 2022 to Q1 of fiscal year 2024, except for Q2 of fiscal year 2023, NVIDIA's gross profit margin is higher than that of the two competitors.
NVIDIA's net profit under both non-GAAP and GAAP caliber from fiscal 2021 to Q1 of fiscal 2024 exceeded that of Intel and AMD.
While releasing the first quarterly report for fiscal 2024, the company gave more optimistic performance guidance for Q2: revenue is expected to reach $11 billion, with a fluctuating range of 2%; Non-GAAP gross margin is expected to reach 70% (up or minus 0.5%). Management's guidance, if met, will deliver growth in the next quarter on both year-over-year and sequentially.
1.3 The four major sectors go hand in hand to accelerate the company's performance
From the perspective of NVIDIA's business composition, NVIDIA has four revenue segments: data center, gaming, professional visualization, and automotive. Among them, the data center and gaming business account for the main part, which together drive NVIDIA's continuous growth. And from the trend point of view, since fiscal 2022, the game business has gradually weakened due to the impact of terminal demand, thanks to the continuous development of the field of artificial intelligence to promote the demand for computing power, the market demand for high-end chips is increasing, NVIDIA data center's revenue is later on top, from Q2 2023 to exceed the revenue of the game business, in the future may become the pillar business of NVIDIA.
1.3.1 The revenue of the data center business increased significantly and is expected to grow into the company's core business
The company's business in the data center mainly revolves around CPU, GPU, DPU, and has developed three main types of products: Grace CPU architecture, Hopper GPU architecture and BlueField DPU architecture. The Hopper architecture advances Tensor Core technology through the Transformer engine, which can apply mixed FP8 and FP16 precision to dramatically accelerate AI computation of Transformer models. Hopper also improves floating-point operations per second (FLOPS) by 3x TF32, FP64, FP16, and INT8 precision compared to the previous generation. At COMPUTEX 2023, NVIDIA announced the DGX GH200 AI supercomputer, which will be used to drive generative artificial intelligence, recommender systems, and data analytics, with a new structure that provides higher bandwidth than the previous generation and is 5 times more energy efficient than competing products. NVIDIA DGX GH200 provides nearly 500x more memory for the GPU shared memory programming model via NVLink than a single NVIDIA DGX A100 320 GB system.
Thanks to the continuous development of generative AI and big language models, the company is doing well in the data center business. Rising demand for Hopper and Ampere processor-based GPUs driven Q1 2023 revenue of a record $4.28 billion, up 14% year-over-year and 18% sequentially, driven by strong demand for the company's products from downstream enterprises such as cloud service providers, consumer internet companies, and enterprises.
The company's data center business focuses on accelerating the most computationally intensive workloads such as artificial intelligence, data analytics, graphics, and scientific computing in hyperscale, cloud, enterprise, public sector, and edge data centers. NVIDIA's partners in this area include companies such as Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud; At present, the company is committed to building an accelerated computing platform with GPU, DPU and CPU as the next-generation architecture, and providing various software development tools around CUDA to continuously reshape the data center in the AI era.
1.3.2 Game Business
The gaming business, another pillar business of NVIDIA in addition to data centers, achieved revenue of $2.24 billion in fiscal 2024 Q1, down 38% year-over-year and up 22% sequentially; The year-on-year decline was mainly due to the macroeconomic slowdown; Sequentially driven revenue growth was driven by the company's announcement of a new generation of GeForce RTX 40 series GPUs for notebooks and desktops. The company's current major products in the gaming market include GeForce RTX series and GeForce GTX series GPUs for PCs and laptops, and GeForce NOW, a cloud service for low-power devices to play PC games.
GeForce RTX and GeForce GTX graphics cards are the mainstream graphics cards for PCs on the market, the market is almost monopolized by NVIDIA and AMD, and according to the full GPU graphics card market share report released by Jon Peddie Research (JPR) for the third quarter of 2022, NVIDIA's discrete GPU graphics card market reached 88%, while AMD's market share was only 8%. The company's nvida RTX series GPUs bring the next generation of graphics to gaming, providing real-time ray tracing and cinematic quality rendering. Nvidia's DLSS 3 in gaming in 2022 brings a new round of revolutionary changes in the field of neurographics, allowing players to greatly improve performance while maintaining excellent graphics and responsiveness during gameplay. In many of today's games and engines, DLSS 3 delivers more than 4x the performance of the GeForce RTX 40 series compared to traditional rendering techniques.
In the context of high market share, the revenue of the graphics card part of NVIDIA's game business is greatly affected by macroeconomic factors, and the shipment of downstream PCs is directly related to the order situation of graphics cards, which is cyclical and cyclical, and in the long run, the company's game business accounts for a large share and is relatively stable. GeForceNOW uses its GPU and advanced software to make games run smoother and have higher image quality to enhance the user's gaming experience. PC hardware determines whether the game's graphics are smooth and the frame rate is high enough, but many players' hardware does not meet the recommended configuration required to run the latest games smoothly, in which case NVIDIA introduced GeForceNOW. GeForceNow is an open platform that allows users to purchase a high-performance cloud-based PC and access to major game distribution platforms to experience masterpieces. The service is subscription-based, and as of May 2023, GeForceNOW members with priority rights in the US are priced at $9.99/month; As of 2021, the number of paying members exceeded 10 million, which is expected to bring stable cash flow to the company in the future.
1.3.3 Professional visualization areas
Professional Visualization Q1 revenue of $295 million in fiscal 2024, down 53% year-over-year, mainly due to the adjustment of inventory levels in the company's sales channels, the market is depleting inventory; Sequentially increased 31%, driven primarily by demand for mobile and desktop workstations.
NVIDIA not only launched professional graphics processors suitable for professional visualization in terms of hardware, but also actively worked closely with independent software vendors to work together on cloud, software and services to make GPUs and software better adapted, forming the company's product barriers in the field of professional visualization: On the hardware side, the company's GPUs provide computing platforms that increase productivity and introduce new capabilities to key workflows in areas such as design and manufacturing, digital content creation, and more. Among the industries that can be used in design and manufacturing include computer-aided design, architectural design, consumer product manufacturing, medical instrumentation, and aerospace; Segments such as specialized video editing and post-production, special effects for film, and broadcast graphics in digital content creation are all driven by the power of GPUs. In terms of cloud, software and services, NVIDIA has built a next-generation toolchain, covering enterprise collaborative design, cloud VR, and cloud gaming scenarios. With the support of ray tracing, machine learning and other technologies, these software services combine to form NVIDIA's products and technical barriers. The company has recently announced several new developments in professional visualization: 1) expanding its partnership with Microsoft to connect Microsoft 365 applications with Omniverse; 2) announced NVIDIA Omniverse Cloud, a fully managed service running on Microsoft Azure for development and deployment of industrial meta-applications; 3) Announcing six new ADA-based NVIDIA RTX GPUs for mobile and desktop workstations.
1.3.4 Automotive Business
In the automotive business, the company achieved revenue of $296 million in fiscal 2024 Q1, up 114% year-over-year and 1% sequentially. Strong year-over-year growth in the first quarter was driven by NVIDIA DRIVE Orin's rapid growth in a number of new energy vehicles. The automotive industry is increasingly transforming into intelligence, automatic driving requires on-board computers to make real-time calculations to make decisions, and at the same time requires on-board radar to scan the driving environment, and then transmit the image back to the computer, the computer's central processing unit for calculation, all links need powerful hardware to support, driving the company's automotive business revenue to increase steadily.
2. The new era of artificial intelligence has created more hardware needs
2.1 ChatGPT applications are accelerating, the demand for AI servers and computing power is increasing, and the data center industry will welcome a breakthrough
In the data center, GPUs are being applied to help solve today's most complex and challenging problems through technologies such as AI, media and media analytics, and 3D rendering. In technology areas such as high-performance computing (HPC) and visual cloud computing, these new use cases require different types of computing power to advance their advanced capabilities. Bringing GPUs to data center environments helps meet growing high compute demands and massive data demands. Today, GPUs are widely used in on-premises and cloud data center environments and are often virtualized for greater flexibility and efficiency. According to the forecast of China Business Industry Research Institute, the global AI chip market is expected to be 72.6 billion US dollars in 2025, and the number of chips will reach 23.8 million sets in the same period, with great growth potential.
On May 19, 2023, OpenAI released the official iOS app of ChatGPT, although it is currently only available in the App Store in the United States. However, OpenAI said that in the future, mobile ChatGPT applications will be launched in more countries and regions, and the Android version is also under development. Google released a new general-purpose language model PaLM 2 at the developer conference on May 10, PaLM 2 artificial intelligence big model is good at mathematics, software development, language translation reasoning and natural language generation, in the future will combine Google's multiple products to achieve comprehensive coverage from various collaboration tools such as documents, to email, search, cloud and other services. The iteration of AI is inseparable from the support of computer infrastructure, especially CPU, GPU, DPU, the development of AI and the upgrade of chips are mutually influencing, as the pace of AI-related applications accelerates, the demand for computing power will increase significantly, which in turn drives the demand for chips.
Taking ChatGPT as an example, the computing power requirements of ChatGPT are positively correlated with the amount of parameters, which puts forward higher requirements for the memory capacity and bandwidth of the hardware. According to the Green Energy Computing Data Center, the total computing power consumption of ChatGPT is about 3640PF-days. According to a data center recently landed in China as a reference, the computing power 500P cost 3.02 billion yuan to complete, to support the operation of ChatGPT, 7-8 such data centers are needed, and the infrastructure investment needs to be tens of billions of yuan. In 2021, the global cloud computing market represented by IaaS, PaaS, and SaaS reached 32.5%, which has basically recovered to the pre-epidemic level, NVIDIA keenly captured that the market size of AI and cloud computing in the new era is constantly increasing, and began to layout in the data center business field, and continuously iterated GPUs into the underlying chips that provide computing power for AI and cloud computing.
2.1.1 Multi-product layout data center business, NVIDIA is the industry leader
NVIDIA's data center products achieve comprehensive coverage from hardware to software, from edge to data center to cloud, and a complete ecosystem has built its own product moat.
NVIDIA's data center GPUs are accelerated computing platforms for data centers, enabling these modern data centers to handle workloads involving deep learning, machine learning, and high-performance computing (HPC) faster; The accompanying NVIDIA DGX system, the world's first custom AI supercomputer, combines innovative GPU-optimized software, breakthrough performance, and simplified management; On the hardware side, in the area of acceleration service platform, NVIDIA introduced the powerful NVIDIA DGX A100 built for deep learning, machine learning and high-performance computing, which combines multi-precision computing to accelerate deep learning, machine learning, and high-performance computing. The HGX A100 combines up to eight NVIDIA A100 Tensor core GPUs and NVIDIA NVSwitch into a unified accelerator for larger computing challenges; On the software side, NVIDIA introduced Virtual GPU Software (vGPU), which provides graphics-capable virtual desktops and workstations, accelerated by NVIDIA data center GPUs, and provides customers with flexibility, security, and improved IT management.
NVIDIA founder and CEO Jensen Huang announced in his NVIDIA Computex 2023 presentation that the generative AI engine "NVIDIA DGX GH200" is now in mass production and will be available by the end of this year. This is a new AI supercomputer that fully connects 256 NVIDIA Grace Hopper superchips into a single GPU, supports trillion-parameter AI large model training, is capable of handling large-scale recommendation systems, generative artificial intelligence and graph analysis, and provides linear scalability for giant AI models. NVIDIA GH200 is mainly aimed at users such as data centers and other industries with high computing power requirements, and the global data center business has steadily increased. According to the forecast data of the China Academy of Information and Communications Technology, the global data center market size exceeded US$67.9 billion in 2021, an increase of 9.8% over 2020. Market revenue is expected to reach $74.6 billion in 2022, with overall stable growth. The continuous development of artificial intelligence has put forward higher requirements for computing power, and further increased the requirements for hardware devices, and NVIDIA GH200 products have made a great improvement in performance for the previous generation to meet the needs of the industry.
2.1.2 Acquisition of Mellanox, NVIDIA presses the DPU development accelerator
DPU (Data Processing Unit) is a dedicated processor built with data as the center, which uses software-defined technology to support infrastructure layer resource virtualization and supports infrastructure layer services such as storage, security, and service quality management. DPU is the "third workhorse chip" after CPUs and GPUs, setting off a wave of industry boom. The emergence of DPUs is a phased sign of heterogeneous computing. Similar to the evolution of GPUs, DPUs are another classic example of application-driven architecture design; But unlike GPUs, DPUs target more low-level applications. The core problem to be solved by DPUs is the "cost reduction and efficiency increase" of infrastructure, that is, the load of "CPU processing inefficiency and GPU processing" is offloaded to a dedicated DPU, improving the efficiency of the entire computing system and reducing the total cost of ownership (TCO) of the overall system. The DPU product line developed by NVIDIA after the acquisition of Mellanox mainly includes BlueField, ConnectX, and Innova. It is planned to integrate GPUs on BlueField-4 to realize single-chip data centers/units, providing low-cost, high-performance secure data processing capabilities for edge devices.
On March 11, 2020, NVIDIA announced that it would acquire network chip company Mellanox, a technology company with a series of network cards, adapters, Ethernet switches, messaging accelerators and other products for US$6.9 billion (approximately 48.8 billion yuan), covering all categories in the server field, including RDMA, an important technology in AI data storage and transmission. Through this acquisition, NVIDIA has further accelerated its layout in the data center field and further broadened product and technical barriers. NVIDIA officials said that the acquisition "injects strong impetus into the creation of a new generation of data centers." Since it is currently in the early stage of DPU development, major manufacturers emphasize landing and ecological construction. Driven by many industry trends such as the gradual maturity of intelligent network card solutions, the steady growth of global general server shipments, and the technology landing of L3 intelligent driving vehicles, the global DPU industry market has shown a trend of increasing year by year. The large-scale production of DPUs by Intel, NVIDIA and other manufacturers will promote the explosive growth of the DPU market in 2023-2024. CCID Consulting estimates that the global DPU industry market will reach $3.05 billion in 2020, and this figure is expected to exceed $24.53 billion by 2025.
Since China has the world's largest number of Internet users and online ecology, the explosion of superimposed data has promoted the demand for computing power in the Internet industry, and DPU has huge development potential in the Chinese market. In terms of ensuring network security, DPUs have unique advantages to achieve full coverage from data security to data center security, which is in line with China's industry trend of gradually paying attention to network security, and expands a huge market development space; DPU can solve problems such as network protocol processing, data security, algorithm acceleration, etc., and the corresponding data center and cloud computing Qiayi has a considerable market size in China. Thanks to the growth in demand brought about by the development of data centers, edge computing, new energy vehicles and other industries, CCID Consulting expects that the Chinese DPU market will also usher in explosive growth in 2023-2024. In 2020, the size of China's DPU industry was only 390 million yuan, and by 2025, this value is expected to exceed 56.59 billion yuan. From the perspective of industry pattern, the DPU industry has a high market concentration. According to the data of Toubao Research Institute, in 2020, NVIDIA's share of the Chinese DPU market reached 55%, and the growing scale of DPU will lead to an increase in NVIDIA DPU shipments.
2.2 The game business is still the pillar business, and the industry as a whole is growing steadily
With the continuous development of the game industry, users' requirements for game experience have gradually increased, and GPU-based product hardware needs to be constantly updated and iterated. The GPUs of NVIDIA, Intel and AMD have almost broken the graphics card market with their excellent performance, especially the mainstream GPUs used by the two major game platforms of PC and game console. PC is divided into the sales of the whole machine by major computer manufacturers and the user's own assembly of the machine, from the perspective of the two ways combined, according to IT Home data, NVIDIA's independent GPU graphics card share reached 88% in Q3 2022, with a leading position; Although the game console market Sony and Microsoft's two current flagship products PlayStation 5 and Xbox Series X both use AMD's RDNA series GPUs, but from the overall trend, PC due to more diverse functions, in recent years in the game market share continues to rise, although NVIDIA's revenue in the gaming field in 2022 fell by 27% compared with 2021, but mainly due to macro factors: the demand for working from home during the epidemic drove notebooks The growth of computer consumption and the surge in the purchase of integrated graphics cards have prematurely consumed market demand to some extent, and in the post-epidemic era, the weakening of demand for laptops and the excess inventory of suppliers have led to a continuous decline in integrated graphics card shipments. But as the global economy improves, IDC forecasts that shipments of gaming PCs, including desktop and notebook computers, are expected to grow from 41.3 million units in 2020 to 52.3 million units in 2025, at a five-year compound annual growth rate (CAGR) of 4.8%. This will also drive the increase in NVIDIA graphics card shipments in the future.
NVIDIA's latest graphics card is a significant improvement over the previous generation RTX30 series. The 40 series graphics card adopts the NVIDIA Ada Lovelace architecture, the new SM multi-unit stream processor performance per watt is increased by up to 2 times, and the fourth generation of Tensor Core uses up to 4 times the performance when using DLSS 3 compared to using only traditional image rendering methods, making great breakthroughs in graphics rendering and low latency of games.
With the popularity of 5G technology, the cloud gaming industry has gradually emerged, which is an online gaming method based on cloud computing technology. All calculations in the game (including screen rendering, data synchronization, interaction logic, etc.) are performed on the cloud server, and the player's input instructions are accepted through the Internet, and the final picture result after processing is displayed on the player's front-end device. In an ideal cloud gaming scenario, the user's gaming device only needs basic video decompression and networking capabilities, without the need for any high-end processor or graphics card. To a certain extent, this can meet the play needs of low-end players, thereby opening up a larger game market. According to the statistics of Huajing Industry Research Institute, the global cloud gaming market revenue reached 11.41 billion yuan in 2021, its domestic market revenue in China was 4.06 billion yuan, and the overseas market revenue was 7.35 billion yuan, and the global cloud gaming market revenue is expected to reach 74.21 billion yuan in 2025.
2.3 Automotive intelligence brings industry increment, NVIDIA rapid layout
The automotive industry is gradually transitioning to intelligence, and this transformation is the most important incremental market for the automotive industry in the future. Intelligent car is composed of bicycle intelligence and Internet of Vehicles, which refers to a new generation of intelligent travel system that integrates new technologies such as information communication, Internet of Things, big data, cloud computing, artificial intelligence and other new technologies by equipped with advanced sensors, controllers, actuators and other devices to realize intelligent information exchange and sharing of the vehicle's intranet, external network, and vehicle-to-vehicle network, with information sharing complex environment perception, intelligent decision-making automation collaborative control function, and intelligent highway and auxiliary facilities together to form an intelligent mobile space and application terminal. Compared with traditional cars, the core difference between intelligent vehicles is that they have more advanced automatic driving assistance systems, intelligent cockpit systems and vehicle networking systems, and the most significant features are intelligence, networking and sharing. Intelligent cars gradually shift from simple transportation to intelligent mobile space through their software and hardware, so as to finally realize a "people-centered" intelligent mobile space.
At present, "intelligent car" is mainly composed of three major elements, namely intelligent interaction, intelligent driving and intelligent services. Among them, the intelligent part includes but is not limited to: intelligent driving, intelligent cockpit, intelligent networking, intelligent electric, and vehicle cloud services. NVIDIA's main drivers come from autonomous driving, software-defined vehicles, and new software and services business models. At present, the entire intelligent vehicle industry is still in a period of rapid development, taking the mainland as an example, according to data from the Ministry of Industry and Information Technology, as of 2022, the sales volume of new intelligent networked passenger vehicles equipped with assisted autonomous driving systems in mainland China reached 7 million, a year-on-year increase of 45.6%; 48% of NEVs are equipped with assisted autonomous driving systems. According to the "White Paper on the Internet of Vehicles" released by the China Academy of Information and Communications Technology, it is expected that by 2025, the scale of the mainland intelligent vehicle market will be close to one trillion yuan.
From the perspective of the whole world, IDC's latest "Global Autonomous Vehicle Forecast Report (2020-2024)" data shows that in 2024, the global L1-L5 autonomous vehicles (self-driving cars are divided into five levels according to the degree of automation, namely: L1 level partial driving assistance, L2 level combined driving assistance, L3 conditional automatic driving, L4 highly automated driving, L5 level full autonomous driving) shipments are expected to reach about 54.25 million units, 2020 to 2020 The compound annual growth rate (CAGR) will reach 18.3% in 2024; The market share of Level 1 and Level 2 autonomous driving is expected to be 64.4% and 34.0%, respectively, in 2024. IDC believes that despite the pioneering use of Level 3-L5 autonomous driving technology, Level 1-L2 autonomous driving will remain the largest market segment that will drive global autonomous vehicle shipment growth over the next 5 years. As the market for smart cars gets bigger and bigger, NVIDIA has kept pace with the times and started to lay out the automotive business. Semiconductor products such as chips are in great demand for automotive autonomous driving, and data from Jiwei Consulting shows that from 2021 to 2025, the global automotive semiconductor market will grow at a CAGR of 10%; By 2025, the global automotive semiconductor market size will reach $73.52 billion. At the same time, Gartner also pointed out that due to the strong demand for autonomous driving, lower emissions and higher efficiency, the automotive chip market will usher in a period of rapid growth. Gartner expects the market to reach $116.6 billion by 2030, up from $38.7 billion in 2020. There are huge development opportunities in the industry as a whole, which is one of the reasons why NVIDIA continues to deploy in this field. With the continuous expansion of the intelligent automobile industry, the revenue of NVIDIA's automobile business still has a lot of room for growth.
Automotive chips have certain requirements for technology, and overall, automotive chips mainly focus on three aspects: reliability requirements, design life (more than 20 years), and high security requirements. NVIDIA regards autonomous driving as the most important layout area. Traditional companies such as Mercedes and Land Rover are also using neural network algorithms provided by NVIDIA to advance their autonomous driving platforms. The company's NVIDIA DRIVE Orin SoC (System-on-Chip) is a dedicated chip for autonomous driving that provides 254 TOPS (trillion operations) per second. As the central computer of the smart vehicle, the chip can power autonomous driving functions, confidence views, digital clusters, and AI cockpits. This is sufficient to handle camera, lidar, ultrasound, and any other sensor data that requires full automation.
NVIDIA is leading the way in automotive chips. At present, the company's main strength in the automotive field is the NVIDIA DRIVE family of embedded supercomputing platforms, which include NVIDIA DRIVE Hyperion, NVIDIA DRIVE Orin and NVIDIA DRIVE Thor. The supercomputing platform can process data from cameras, common radar, and lidar sensors to sense the surroundings, determine the location of the car on a map, and then plan and execute safe driving routes. This compact, energy-efficient AI platform supports autonomous driving, cockpit functions and driver monitoring, among other safety features. The combination of hardware and software together constitutes NVIDIA's product barrier:
1) NVIDIA DRIVE Hyperion Platform for Mass Production Autonomous Vehicles: This autonomous vehicle reference architecture accelerates development, testing, and validation by integrating DRIVE Ordin-based AI computing with a complete sensor suite. It also features a complete software stack for autonomous driving, as well as driver monitoring and visualization, enabling over-the-air updates to add new features and functionality throughout the vehicle's lifecycle. Not only that, but it is also cross-generational compatible, allowing partners to seamlessly migrate to NVIDIA DRIVE Thor and subsequent platforms using the DRIVE Orin platform they currently use;
2) NVIDIA DRIVE Orin SoC: As a high-performance smart car central computer that provides 254 TOPS (trillion operations) per second, it can power autonomous driving functions, confidence views, digital clusters, and AI cockpits. With the scalable DRIVE Orin product family, developers can build and scale from Level 2+ systems to Level 5 fully autonomous vehicle systems with a single development investment across the fleet;
3) NVIDIA DRIVE Thor: The company's next-generation centralized in-vehicle computing platform that runs advanced driver assistance applications and in-vehicle infotainment applications on a single, secure, and reliable system. The DRIVE Thor superchip is a powerful chipset developed by the company's new CPU and GPU combination that delivers outstanding 2,000 trillion floating-point operations while reducing overall system cost, with mass production scheduled to begin in 2025.
At the 2022 Intel GTC conference, the Orin series of chips was announced to begin mass production, and at the same time, NVIDIA launched a new generation of autonomous driving platform DRIVE Hyperion 9 based on Atlan chips, which is scheduled for mass production in 2026. NVIDIA and a number of car companies reached strategic cooperation, in March 2022, Chinese new energy vehicle manufacturer BYD announced a cooperation with NVIDIA in intelligent driving technology, from the first half of 2023, BYD will be equipped with NVIDIA DRIVE Hyperion platform on some of its new energy vehicles to achieve intelligent driving and intelligent parking of vehicles. In the Q1 quarterly report of fiscal year 2024, it was revealed that Chinese new energy vehicle manufacturer BYD will use NVIDIA DRIVE Orin series chips in new models, and Nvidia Chief Financial Officer Colette Kress mentioned at the investor conference that in the next five years, the automotive business is expected to bring Nvidia $11 billion in revenue.
With the gradual rise of new energy vehicles and the gradual update and iteration of fuel vehicles in the world, it is expected that more car manufacturers will cooperate with NVIDIA in the future, and the shipments of hardware in the automotive field of NVIDIA will usher in an increase together with the shipments of new energy vehicles, and the company's revenue in this field is expected to further increase. According to Jibang Consulting, global new energy vehicle sales are expected to increase from 2.046 million units in 2019 to 14.51 million units in 2023, and the year-on-year growth rate of new energy vehicles is expected to reach 36.2% in 2023.
2.4 NVIDIA wants to create professional visualization as the third pillar industry
Professional visualizations include, but are not limited to, architecture, engineering and construction, education, manufacturing, media and entertainment, and more. NVIDIA's professional services in this field cover professional graphics rendering, cloud XR applications, AI data science and big data research, and launch a variety of solutions at the software and hardware levels. In this area, NVIDIA has a strong market share. The company's main product in this field, Omniverse, has promoted the development of enterprise collaborative design, and CloudXR, as NVIDIA's industry AR/VR application product, can be used for digital content creation, medical care, architectural design and other uses; On the other hand, driven by the high bandwidth of 5G, GPUs have a wider range of applications in professional rendering, video creation and editing, engineering design and other sub-fields. Take architectural engineering and media entertainment as examples:
In Architecture & Engineering: NVIDIA Omniverse builds a real-time graphics and simulation platform based on NVIDIA RTX GPU and Pixar Common Scenario descriptions, while leveraging Omniverse's AEC Experience feature set to provide companies with a range of tools to improve the conceptual design process. And Omniverse introduces a new type of shading dye, Omniverse View: the module is accelerated by multiple NVIDIA RTX GPUs and provides extremely high scalability for GPU arrays, providing high-quality real-time output even for huge 3D models. Omniverse View displays 3D content aggregated from different applications within Omniverse, or directly in the 3D application being used. It also supports commercial game engines and offline renderers.
In media and entertainment: Omniverse systems can speed up the film and television production process in virtual production, rendering, artificial intelligence, and more. For example, virtual production sets are created, iterated and collaborated in real time by connecting virtual production sets directly to artists using NVIDIA certification systems, network solutions, and the NVIDIA Omniverse Enterprise platform; NVIDIA RTX has RT Core for ray tracing, Tensor Core for AI noise reduction, supersampling and other features to create beautiful, lighting-accurate renderings in real time, and GPU acceleration is now supported by all major renderers, including Autodesk Arnold, Chaos V-Ray, Maxon Redshift, Isotropix Clarisse, and DreamWorks MoonRay, Pixar RenderMan XPU, and NVIDIA Omniverse RTX renderers; NVIDIA Omniverse also enables teams to create content at breakneck speed with one-click interoperability between high-end content production tools and seamless collaboration in an interactive simulation environment.
According to the data of Bejes Information Consulting, the total size of the global data visualization market reached 36.34 billion yuan in 2021, and the size of China's data visualization market reached 7.846 billion yuan, accounting for 21.59% of the total share of the global data visualization market. The data visualization market is expected to grow steadily at a CAGR of 8.97% during the forecast period 2021-2027, with the total global data visualization market expected to reach $61.516 billion in 2027. NVIDIA currently maintains its position as a leading manufacturer in the field of professional visualization, and with the popularization of more and more intelligent technologies, more and more professional visualization application scenarios will appear in education, architecture, games, film and television industries, and NVIDIA's revenue is expected to further increase.
3. Profit forecasts
Key assumptions
Based on the current development trend of the big wave of the AI industry, the demand for AI computing power, large models and AI servers and other application fields is growing rapidly, and the demand for GPU chips such as A100 and H100 will increase with the requirements of the training and inference layers. OpenAI will continue to catalyze the development of the global AI industry and bring significant improvements to NVIDIA's performance. We expect the company's FY2024-FY2026 operating income to be 337.29/432.96/56.268 billion US dollars, a year-on-year growth rate of 25.04%/28.36%/29.96%. The company is still in the stage of continuous upward performance, and we will be optimistic about the development trend of NVIDIA in the long term.
(This article is for informational purposes only and does not represent any investment advice from us.) For information, please refer to the original report. ）
Selected report source: [Future Think Tank]. 「Link」