Ping An Securities: NVIDIA's Q2 FY2025 performance continues to grow rapidly year-on-year, with intense competition in the global large model field

According to the financial news app Zhitong Finance, Ping An Securities released a research report stating that NVIDIA (NVDA.US) continues to achieve high-speed growth in the second quarter of the 2025 fiscal year. The data center business performed strongly, with revenue reaching $26.3 billion, a year-on-year increase of 154% and a quarter-on-quarter increase of 16%. Currently, competition in the global large model field remains intense. Chinese large model manufacturers continue to iterate and upgrade algorithm capabilities. By the end of 2023, the development of domestic large models will enter an explosive period. According to the latest evaluation results of SuperCLUE in the first half of 2024, the vast majority of domestic closed-source models have surpassed GPT-3.5Turbo, and are expected to accelerate the application of domestic large models in various scenarios. At the same time, the continuous iteration and upgrade of large models will provide strong impetus for the development of the AI computing power market.

Key Points:

NVIDIA released its second quarter report for the 2025 fiscal year. According to NVIDIA's announcement, NVIDIA achieved operating income of $30.04 billion in the second quarter of the 2025 fiscal year, a year-on-year increase of 122% and a quarter-on-quarter increase of 15%. It achieved a net profit (GAAP) of $16.599 billion, a year-on-year increase of 168% and a quarter-on-quarter increase of 12%.

Ping An's View:

NVIDIA's high-speed growth in the second quarter of the 2025 fiscal year. According to NVIDIA's announcement, NVIDIA achieved operating income of $30.04 billion in the second quarter of the 2025 fiscal year, a year-on-year increase of 122% and a quarter-on-quarter increase of 15%. According to NVIDIA's outlook for the second quarter of the 2025 fiscal year at the FY25Q1 performance briefing, the company expected to achieve revenue of $28 billion in the second quarter, with a fluctuation of 2%. The actual revenue in the second quarter of the 2025 fiscal year exceeded the previous upper limit estimate (the upper limit was $28.56 billion), with revenue in the second quarter exceeding expectations. In terms of profit and expenses, the gross profit margin for the second quarter of the 2025 fiscal year (GAAP) was 75.1%, an increase of 5.0 percentage points year-on-year and a decrease of 3.3 percentage points quarter-on-quarter; operating expenses were $3.932 billion, an increase of 48% year-on-year and 12% quarter-on-quarter; operating profit was $18.642 billion, an increase of 174% year-on-year and 10% quarter-on-quarter; net profit was $16.599 billion, an increase of 168% year-on-year and 12% quarter-on-quarter. NVIDIA's performance in the second quarter of the 2025 fiscal year was impressive.

NVIDIA's revenue growth expectation for the third quarter of the 2025 fiscal year has slowed down, but is still expected to increase significantly. According to NVIDIA's outlook for the third quarter of the 2025 fiscal year at the FY25Q2 performance briefing, the company expects revenue for the third quarter of 2025 to be $32.5 billion, with a fluctuation of 2%. Based on revenue of $32.5 billion, it is estimated to increase by 79% year-on-year and 8% quarter-on-quarter. The expected year-on-year growth rate of revenue in the third quarter (79%) is slightly slower compared to the past five quarters (year-on-year revenue growth rates in the five quarters from FY24Q2 to FY25Q2 were 101%, 206%, 265%, 262%, 122%), but still a significant year-on-year increase. The gross profit margin is expected to be 74.4%, with a fluctuation of 50 basis points, and the full-year gross profit margin is expected to be around 75%; operating expenses are expected to be approximately $4.3 billion; other income is expected to be around $350 million; and the tax rate is expected to be 17%, with a fluctuation of 1 percentage point NVIDIA's Data Center Business Shows Strong Performance. In the second quarter of the 2025 fiscal year, NVIDIA's data center business revenue reached $26.3 billion, a year-on-year increase of 154% and a quarter-on-quarter increase of 16%. The strong performance of the company's data center business is mainly due to the use of the Hopper architecture GPU computing platform for training and inference in large models, recommendation engines, and generative AI applications, driven by strong market demand. Cloud service providers contributed approximately 45% of the revenue from the data center business. By segment, the revenue from the computing segment was $22.6 billion, a year-on-year increase of 162% and a quarter-on-quarter increase of 17%. The revenue from the networking segment was $3.7 billion, a year-on-year increase of 114% and a quarter-on-quarter increase of 16%, mainly driven by growth in InfiniBand switches and Ethernet switches. Over the past four quarters, revenue generated from inference demand accounted for over 40% of the company's data center revenue.

The Blackwell architecture GPU products will ramp up in the fourth quarter of the 2025 fiscal year. According to information from NVIDIA's FY25Q2 earnings conference, NVIDIA has sent Blackwell architecture GPU sample products to customers for testing in the second quarter of the 2025 fiscal year. The company made some modifications to the mask version of the Blackwell architecture GPU to improve product yield. The ramp-up of Blackwell architecture product production will continue from the fourth quarter of the 2025 fiscal year to the 2026 fiscal year. In the fourth quarter of the 2025 fiscal year, the company expects Blackwell architecture product revenue to reach billions of dollars. Additionally, the market demand for Hopper architecture products remains strong, with shipments expected to continue to grow in the second half of the 2025 fiscal year. Ping An Securities believes that the GPU chip market represented by NVIDIA still has strong market demand in the future. Furthermore, downstream tech giants' capital expenditures also maintain a high growth rate. In the second quarter of 2024, the combined capital expenditures of Microsoft, Meta, Google, and Amazon totaled $57.1 billion, a year-on-year increase of 66% and a quarter-on-quarter increase of 22%. The high growth rate of capital expenditures by tech giants also provides strong support for the continued high prosperity of the AI computing market.

Investment Recommendation: Ping An Securities is optimistic about investment opportunities in the AI theme. In terms of targets: 1) In terms of computing power, it recommends Industrial Fulian (601138.SH), Inspur Information (000977.SZ), Sugon Information (603019.SH), Tsinghua Unigroup (000938.SZ), Digital China (000034.SZ), Hikvision Digital Technology (688041.SH), and Loongson Technology (688047.SH), and suggests paying attention to Cambricon Technologies (688256.SH), Horizon Robotics (300474.SZ), and SoftBank Dynamics (301236.SZ); 2) In terms of algorithms, it recommends iFlytek (002230.SZ); 3) In terms of application scenarios, it strongly recommends Inspur Electronic Information (300496.SZ), Hengsheng Electronics (600570.SH), and Shengshi Technology (002990.SZ), recommends Kingsoft Office (688111.SH), and suggests paying attention to Wanxing Technology (300624.SZ), Tonghuashun (300033.SZ), and Colorful Guofen (300634.SZ); 4) In terms of cybersecurity, we highly recommend Qihoo 360 (002439.SZ).

Risk Warning: 1) Risks in the AI computing power supply chain are increasing. The US has escalated semiconductor exports to China, which will push the domestic AI chip industry chain to mature faster. However, if the iteration speed of domestic AI chips in China does not meet expectations, it will affect the development of AI computing power in China, thereby restricting the breakthrough of large models. 2) The application landing of large model products is lower than expected. Currently, domestic large models in China have begun to continuously land in education, healthcare, automotive, office, industrial, smart hardware, and other B-end and C-end application scenarios. If the market expansion of products falls short of expectations, there will be a risk of the application landing of domestic large model products being lower than expected. 3) The development of domestic large model algorithms may not meet expectations. Currently, although the capabilities of domestic large models are continuously improving, compared to global leading large model manufacturers such as OpenAI, there is a risk that the progress of domestic large model manufacturers in catching up may not meet expectations