On November 20th, Nvidia released its third quarter results: revenue of $57.01 billion, exceeding market expectations by $54.92 billion; Net profit of 31.91 billion US dollars, up 65% year-on-year! Affected by positive news, the company's market value increased by over 2 trillion yuan overnight.

From a business perspective, Nvidia's most important data center business achieved revenue of $51.2 billion, a year-on-year increase of 66%, accounting for approximately 80% of total revenue. This was mainly due to the strong demand for Blackwell architecture GPUs in generative AI and big language models. Among them, the "computing" (GPU) business contributed $43 billion in revenue; The network business (i.e. components that allow multiple GPUs to work together) contributed $8.2 billion.

In terms of other businesses, Nvidia Games' Q3 revenue was $4.3 billion, a year-on-year increase of 30%. Before the rise of data centers, gaming was Nvidia's largest business segment. The Q3 revenue of professional visualization business was 760 million US dollars, a year-on-year increase of 56%; The revenue from the automotive business was 592 million US dollars, a year-on-year increase of 32%.

Nvidia Chief Financial Officer Colette Kress stated in a statement that the company's current best-selling chip series is the Blackwell Ultra, which is the second generation of Blackwell chips. Nvidia CEO Huang Renxun also mentioned in a statement that "all cloud GPUs have been sold out. ”

Although the performance is good, the industry's concern about the "AI foam" has not completely subsided. Recently, many giants have sold off Nvidia stocks one after another, such as SoftBank Group selling Nvidia shares worth $5.8 billion; A hedge fund led by venture capitalist Peter Thiel also liquidated its holdings in Nvidia stocks in the third quarter.

At the performance briefing, Huang Renxun and Nvidia executives responded to analysts' questions by stating that demand for Nvidia products is still overwhelming and downstream customers are benefiting from artificial intelligence. In addition, the management emphasized that there is still room for further growth in Nvidia's previously proposed goal of generating $500 billion in revenue for the next two years.

Nvidia announced that the Rubin platform is scheduled for mass production in 2026, and the first chip has been delivered; In the DeepSeeker R1 benchmark test, the training speed of Blackwell Ultra is 5 times that of Hopper, and the performance per watt is 10 times that of H200.

Nvidia also announced partnerships with companies such as AWS, Humane, Suzuki, Intel, ARM, and Anthropic for a new artificial intelligence factory project involving up to 5 million GPUs.

*Disclaimer: The above content is reproduced on the WeChat official account of the semiconductor industry circle, which does not represent the views and positions of the company, but only for exchange and learning. If you have any questions or objections, please contact us.