The computing power battle escalates, Nvidia is not only designing chips, but also designing data centers

Author: Li Xiaoyin Source: Hard AI The computing power battle is escalating! Nvidia plans to further expand its "AI territory" through data center design. According to The Wall Street Journal, Nvidia's founder and CEO Jensen Huang is planning to transform Nvidia into a comprehensive AI solution provider by offering services such as software, data center design, and networking technologies, rather than just a GPU supplier. Specifically, Nvidia is investing in the development of a high-speed network protocol called InfiniBand based on its proprietary software CUDA for large-scale AI training. In addition, Nvidia is also building an AI-optimized Ethernet business, with the company's CFO Colette Kress expecting this business to generate billions of dollars in revenue within a year. In other words, in the future, Nvidia will release CPU and networking chip products for data center deployment. These fine-tuned products can seamlessly work together to meet the software and hardware needs of specific industries such as healthcare and robotics. During last week's earnings call, Jensen Huang stated: "We have a fairly unique ability to integrate and design an AI factory because we have all the components. Unless you have all the components, you can't introduce a new AI factory every year." Srini Pajjuri, an analyst at investment firm Raymond James, commented that as competition in AI data centers intensifies, customers may start seeking "like-for-like" Nvidia technology solutions. Therefore, for Nvidia, the "time to market" for the AI factory is the most important thing. As Nvidia's competitor, AMD has already taken the lead in this field. Prior to Nvidia's announcement of its second-quarter earnings, AMD acquired data center design and manufacturing company ZT Systems for $5 billion, and in recent years has also acquired programmable chip company Xilinx and data center networking company Pensando