The rapid evolution of AI demands dramatic gains in the computing performance of GPUs used in data centers. Meanwhile, power consumption and heat density are rising sharply, and a "thermal wall," in which cooling capacity limits performance, is emerging. Efficiently dissipating the generated heat has become an urgent challenge. Its solution requires not only higher thermal conductivity of the materials, but also lower thermal resistance at the device-material interface to optimize the entire heat path. In this talk, we first introduce the measurement technologies Resonac is developing to accurately characterize interfacial thermal resistance. We then present the development status of thermal interface materials (TIMs) balancing heat dissipation with package reliability, including carbon-based and metal-based materials such as low-temperature sintered bonding, along with our efforts to enhance heat dissipation by focusing on the thermal transport properties of the bulk resin.

The rapid expansion of artificial intelligence (AI) is driving a dramatic increase in data center power consumption, making both power delivery capacity and energy efficiency critical societal challenges. To overcome this “power supply barrier,” power conversion systems must achieve higher efficiency, greater power density, and improved reliability. Wide-bandgap semiconductor power devices based on SiC and GaN are attracting significant attention as key enabling technologies, offering performance beyond the limitations of conventional silicon power devices. This presentation provides an overview of the roles that SiC and GaN power devices play in realizing next-generation power infrastructure for AI data centers. It also introduces the latest research and development trends in SiC/GaN power device technologies and discusses future directions for next-generation power electronics enabled by these advances.