Is the AI Server Industry Promising? Analysis of China's AI Server Industry Development
-
Is the AI Server Industry Promising?
AI servers are data servers capable of providing artificial intelligence (AI) services. They can support both local applications and web pages, as well as deliver complex AI models and services for cloud and on-premises servers. AI servers facilitate real-time computational services for various AI applications. There are primarily two architectures for AI servers: a hybrid architecture that stores data locally and a cloud-based architecture that utilizes remote storage technologies and hybrid cloud storage (a combination of local and cloud storage) for data retention.
AI servers are typically equipped with acceleration chips such as GPUs, FPGAs, and ASICs. The combination of CPUs and acceleration chips meets the demands of high-throughput interconnectivity, providing robust computational support for AI applications like natural language processing, computer vision, and machine learning, thereby facilitating AI algorithm training and inference processes. In China, the upstream segment of the AI server industry chain consists of core components and software suppliers; the midstream includes AI server manufacturers; and the downstream encompasses various application markets, such as internet companies, cloud computing firms, data center service providers, government agencies, financial institutions, healthcare sectors, and telecom operators.
Artificial intelligence, abbreviated as AI, is a new technological science that studies and develops theories, methods, technologies, and application systems to simulate, extend, and expand human intelligence. AI is a key driver of the new wave of technological revolution and industrial transformation. As an essential component of intelligent disciplines, AI seeks to understand the essence of intelligence and create intelligent machines that can respond in ways similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. Since its inception, AI theories and technologies have matured, and their applications have expanded continuously. It is conceivable that future AI-driven technological products will serve as 'containers' of human wisdom. AI simulates human consciousness and thought processes. While AI is not human intelligence, it can think like humans and may even surpass human intelligence.
Servers are a type of computer that operates faster, handles higher loads, and is more expensive than ordinary computers. Servers provide computational or application services to other clients (such as PCs, smartphones, ATMs, or even large-scale equipment like train systems) within a network. Servers feature high-speed CPU processing, long-term reliable operation, powerful I/O external data throughput, and superior scalability. Generally, servers are capable of responding to service requests, delivering services, and ensuring service reliability. As electronic devices, servers have highly complex internal structures, though they are not vastly different from ordinary computers, comprising components like CPUs, hard drives, memory, systems, and system buses.
Analysis of the Current Market Development of the AI Server Industry
The upstream segment of the AI server industry chain primarily includes chips, PCBs, connectors, cables, power supplies, and various interfaces. The midstream consists of server brand manufacturers and OEM/ODM providers, with OEM/ODM likely transitioning toward JDM models in the future. The downstream comprises various client groups purchasing servers, primarily B2B and government clients, such as internet companies, cloud service providers, telecom operators, government agencies, and financial institutions.
In recent years, China's integrated circuit production has shown an upward trend, increasing from 156.49 billion units in 2017 to 359.43 billion units in 2021. In 2022, production slightly declined to 324.19 billion units. With the continuous expansion of chip manufacturing capacity, China's silicon wafer industry has experienced rapid market growth. From 2019 to 2021, the market size of China's semiconductor silicon wafers exceeded $1 billion annually, reaching $1.656 billion in 2021, a year-on-year increase of 24.04%. The explosive growth of global data volume has driven sustained demand for data storage. Solid-state drives (SSDs), being more efficient than traditional hard disk drives (HDDs), have become the mainstream storage devices for AI servers. Global SSD shipments have risen in recent years, reaching 420 million units in 2022.
Driven by economic globalization and national industrial policies, China's high-tech manufacturing sectors, such as power electronics, have developed rapidly, leading to robust growth in the power supply industry. In 2021, the market size of China's power supply industry reached ¥341.6 billion. With the accelerated industrialization of AI applications, global AI infrastructure spending continues to grow rapidly, and the market size of the AI server industry has expanded steadily. In 2022, the global AI server market size reached $18.3 billion, with shipments accounting for nearly 1% of total server shipments, approximately 145,000 units.
As the demand for digital infrastructure data load capacity rises in China, the domestic AI server market has maintained rapid growth. In 2022, the sales revenue of AI servers in mainland China reached $7.255 billion, with shipments totaling 284,000 units. Among global AI server purchases in 2022, the four major North American cloud providers—Microsoft, Google, Meta, and AWS—accounted for about 66% of the total. Meanwhile, China's AI development wave has intensified, with ByteDance accounting for 6.2% of annual purchases, followed by Tencent, Alibaba, and Baidu at approximately 2.3%, 1.5%, and 1.5%, respectively.
In 2022, China's AI server market by sales revenue was dominated by internet companies with 47.5%, followed by telecom operators at 19.6%. Other sectors like communications, government, manufacturing, education, and finance accounted for 8.6%, 6.3%, 5.1%, 5%, and 3.8% respectively. AI servers adopt heterogeneous architectures, categorized by chip types into CPU+GPU, CPU+FPGA, and CPU+ASIC combinations. Currently, GPUs remain the preferred choice for data center acceleration, though other non-GPU chips are increasingly being adopted, with projections indicating they will exceed 20% by 2025. Generally, ASICs offer the best performance but have weaker programmability and flexibility, while GPUs are better suited for training or general-purpose applications.
At the 2023 China Computing Power Conference, Lenovo unveiled two new AI server products: the Lenovo Wentian WA7780 G3 AI Large Model Training Server and the Lenovo Wentian WA5480 G3 AI Training-Inference Integrated Server. These servers are designed with an AI-oriented approach to help build greener and more efficient AI data centers, meeting diverse computing needs from algorithm training and inference to large-scale model training.
Future Market Size Projections for the AI Server Industry
Globally, the AI server market was valued at $21.1 billion in 2023 and is projected to reach $31.79 billion by 2025, with a CAGR of 22.7% from 2023 to 2025. The training and inference of AIGC large models require substantial high-performance computing (HPC) support, driving demand for AI servers. It is estimated that AI server shipments will grow by 8% in 2023, with a CAGR of 10.8% from 2022 to 2026.