Guosheng Securities: Behind NVIDIA's financial report is the long logic of the AI narrative

According to the Zhitong Finance APP, Guosheng Securities released a research report stating that behind NVIDIA's financial report is the long-term sustainability and growth of AI hardware demand. From various perspectives, including technology drivers, demand structure, and competitive landscape, it can be argued that the long-term logic of the AI narrative has not wavered. The firm pointed out that in the short term, capital flows tend to favor AI application fields, and AI hardware valuations are under pressure. However, from the perspective of the industry cycle, the value of hardware will ultimately be repriced and return to reasonable levels. The firm noted that AI applications and hardware complement each other, and the growth of hardware demand and the implementation of AI applications will form a positive feedback loop.

Guosheng Securities' main points are as follows:

Financial Highlights: NVIDIA's FY25Q3 revenue was $35.1 billion, a year-on-year increase of 94% and a quarter-on-quarter increase of 17%, exceeding the company's previous guidance of $32.5 billion (±2%); net profit was $19.3 billion, a year-on-year increase of 109%. Among them, data center business revenue was $30.8 billion, a year-on-year increase of 112% and a quarter-on-quarter increase of 17%, which is the main growth driver, mainly due to strong demand for Hopper and Blackwell chips; gaming, professional visualization, automotive & robotics businesses all achieved year-on-year growth. The company's GAAP gross margin was 74.6%. Although the gross margin fluctuated in the short term, long-term profitability remains solid. In addition, the company's Q4 revenue guidance is $37.5 billion (±2%), further confirming the continuity of demand in the AI hardware market.

Sufficient growth potential for AI hardware, solid narrative logic. Behind NVIDIA's financial report is the long-term sustainability and growth of AI hardware demand. From various perspectives, including technology drivers, demand structure, and competitive landscape, it can be argued that the long-term logic of the AI narrative has not wavered:

Exponential growth in computing demand

The expansion of AI applications is significantly increasing computing demand. From the training to the inference stage, the scale of generative AI models (such as the GPT series) continues to expand. For example, OpenAI's latest release of GPT-01 indicates that the growth in the number of parameters in AI models is leading to higher demand for high-performance computing hardware (strengthening the demand for hardware upgrades), and the computing volume in the inference stage is further expanding in commercial implementation, with data center GPU demand remaining highly prosperous.

The prevalence and high-frequency demand of inference applications

Unlike training, inference applications have a higher frequency and broader coverage. As AI expands from core data centers to edge computing scenarios (such as real-time language translation, autonomous driving, and generative content recommendation), the demand for computing power shows a shift from centralized to distributed, and this structural change will drive the overall demand for AI hardware upward and further stimulate the demand for low-latency, high-bandwidth connection technologies (such as NVIDIA's Spectrum-X).

Fundamental position in the ecosystem chain

AI hardware is the cornerstone of the entire ecosystem. Whether it is CSP manufacturers or AI companies, strong hardware support is needed. Currently, NVIDIA's products cover training GPUs (H100, A100), inference GPUs (L40, Blackwell), network connections (Spectrum-X), and supporting facilities for optimizing AI software. The end-to-end solutions are forming an "NVIDIA ecosystem" on one hand, and on the other hand, they closely bind hardware demand with the development of AI technology The market's pessimism towards hardware stems more from short-term sentiment and funding factors rather than changes in the industry's fundamentals. In the long run, AI hardware, especially custom chips and optical communication modules, are key components supporting AI training and inference. The proliferation of AI applications will further drive the demand for high-performance hardware, accelerating the trend of optical module upgrades and iterations; in addition, the competitive landscape of the hardware industry is relatively stable, with leading companies having significant advantages in technological barriers and market share. Their profitability and growth potential will not change due to fluctuations in market sentiment. The future development of AI will continue to increase the demand for hardware, further highlighting the advantages of leading companies.

In the short term, capital flows are inclined towards AI application fields, and the valuation of AI hardware is suppressed (not due to fundamental reasons), essentially reflecting the market's "high-low cut." However, from the perspective of the industry cycle, the value of hardware will ultimately be re-priced, returning to a reasonable level. Therefore, AI hardware and application software are not in a competitive relationship but rather coexist and thrive together, with the growth of hardware demand forming a positive feedback loop with the implementation of AI applications.

Thus, from the current standpoint, focusing on the layout for spring 2025, we recommend gradually allocating to leading optical module companies represented by "Yi Zhongtian," which are expected to welcome a valuation switch in the coming year.

Recommended focus:

Computing Power - Optical Communication: Zhongji Xuchuang, Xinyi Sheng, Tianfu Communication, Taicheng Light, Tengjing Technology, Guangku Technology, Guangxun Technology, Dekeli, Liant Technology, Huagong Technology, Yuanjie Technology, Cambridge Technology, Mingpu Optoelectronics. Copper Links: Wolong Core Materials, Jingda Co., Ltd. Computing Power Equipment: ZTE Corporation, Tsinghua Unigroup, Ruijie Networks, Shengke Communication, Feiling Kesi, Industrial Fulian, Hudian Co., Ltd., Cambrian, Haiguang Information. Liquid Cooling: Invec, Shenling Environment, Gaolan Co., Ltd. Edge Computing Power Carrier Platforms: Meige Intelligent, Guanghetong, Yiyuan Communication. Satellite Communication: China Satcom, China Satellite, Zhenyou Technology, Haige Communication.

Data Elements - Operators: China Telecom, China Mobile, China Unicom. Data Visualization: Haohan Deep, Hengwei Technology, Zhongxin Saike.

Risk Warning: AI development may fall short of expectations, computing power demand may be lower than expected, market competition risks