Incident: On the evening of April 19, Inspur Information released its 2023 annual report. The company achieved revenue of 65.867 billion yuan, a year-on-year decrease of 5.41%; realized net profit to mother of 1,783 billion yuan, a year-on-year decrease of 14.54%; realized deducted non-net profit of 1,118 billion yuan, a year-on-year decrease of 35.87%; gross margin was 10.04%, a year-on-year decrease of 1.14 percentage points.
The company's profit side grew significantly in Q4 in '23. In terms of revenue, the company achieved revenue of 65.867 billion yuan in 2023, a year-on-year decrease of 5.41%, of which in the fourth quarter, achieved revenue of 17.77 billion yuan, an increase of 6.04%; by business, the company's server and components business achieved revenue of 65.24 billion yuan in 2023, a year-on-year decrease of 5.41%; IT terminals and spare parts achieved revenue of 393 million yuan. The year-on-year increase was 29.77%. Profit side: In 2023, the company achieved net profit of 1,783 billion yuan, a year-on-year decrease of 14.54%. Of these, in the fourth quarter of a single year, it achieved net profit of 996 million yuan, an increase of 86.49% over the previous year.
The company actively operates and prepares goods, and the cost ratio remains stable: 1) In 2023, the company achieved a cash flow of 520 million yuan from operating activities, a year-on-year decrease of 71.3%, mainly due to an increase in preparation and procurement payments in the current period. The company's inventory reached 19.115 billion yuan, an increase of about 27% over the previous year. The vigorous development of large-scale models and generative AI in 23 years has driven downstream customers' demand for AI computing power to continue to increase, and the company has increased its efforts to prepare related raw materials. 2) The company's sales expenses in 2023 were 1,455 billion yuan, down 3.62% from the previous year; management expenses were 765 million yuan, up 4.46% from the previous year; R&D expenses were 3,071 billion yuan, down 3.95% from the previous year, and the company's cost control remained good.
The company's leading server position is stable, seizing AIGC's computing power demand to unlock new opportunities. 1) The big model and generative AI in 2023 set off a global AI boom, and the AI computing power market grew rapidly. According to the “2023-2024 Evaluation Report on the Development of Artificial Intelligence Computing Power in China”, artificial intelligence is moving from perceptual intelligence to generative intelligence. In 2023, China's artificial intelligence server market will reach 65 billion yuan, an increase of 82.5% over the previous year; the scale of intelligent computing power is expected to reach 414.1 EFLOPS (10 billion floating point operations per second), an increase of 59.3% over the previous year. 2) Wave continues to maintain its leading position in servers. According to the latest data released by IDC, in '23, the company's server market share was second in the world and number one in China; the storage market share was the top three in the world and number one in China; and the market share of liquid-cooled servers was number one in China.
The company continues to make breakthroughs in hardware, computing power, algorithms, etc., to enhance the competitiveness of core products. 1) Hardware level: In April 2024, the company released the Super AI Ethernet Switch X400, which is the first domestic Ethernet switch built on the NVIDIA Spectrum-X platform. It has the highest throughput of 51.2T in the industry, which is 4 times higher than the previous generation of products. 2) Computing power level: In 2023, the company released a new generation of G7 computing power platform. The company's next-generation AI training server, NF5688G7, has improved the measured performance of the large model of the previous generation platform by nearly 7 times. In April 2024, the company and Intel jointly released the AI general-purpose server NF8260G7, which is the first time in the industry that the server can run on a general-purpose processor supporting 100 billion parameter large models. 3) Algorithmic level: The “Source 2.0” basic model released by the company in 2023 includes models with three parameters of 102.6 billion, 51.8 billion, and 2.1 billion. As the first fully open source 100 billion parameter model, “Source 2.0” excels in programming, reasoning, and logic.
Profit forecast: We expect EPS to be 1.39/1.64/1.97 yuan in 2024-2026, and the corresponding PE will be 26.93/22.92/19.05 times.
Risk warning: 1. The development of the big model falls short of expectations 2. Increased market competition 3. The development of the AI industry fell short of expectations.