PANews reported on March 18 that according to financial reports, NVIDIA will hold the world's top AI summit, GTC2025, from March 17 to 21, local time in the United States. At that time, Huang Renxun will deliver a keynote speech, focusing on the future development of AI intelligent bodies, robotics, and accelerated computing. This conference focuses on the iteration of AI computing power, and will focus on the new generation of Blackwell Ultra GPU and Vera Rubin super chip architecture. According to official disclosures, the conference will release a number of technological innovations, including the new generation of GB300 and B300 computing power cards, CPO switches, and NVL288 cabinet solutions. Among them, the performance of B300 may be improved by more than 50% compared to B200.
With the rapid development of AI technology, the demand for computing power continues to rise. In order to meet the growing demand for computing power, hardware equipment is constantly iterating and upgrading, which also brings huge business opportunities to the field of new materials. Especially in the process of AI computing power iteration, the demand for high-performance and high-stability materials is becoming increasingly urgent. PTFE (polytetrafluoroethylene) materials have become a core innovation point due to their high-frequency transmission performance advantages. It is reported that in NVIDIA's new generation of GB300 AI servers and NVL72 architectures, PTFE-based multi-layer PCBs (more than 40 layers) are used in orthogonal backplane designs to replace traditional copper cable solutions. It is estimated that the global demand for PTFE resin in AI servers will exceed 10,000 tons in 2025, with a market size of tens of billions of US dollars.
