日本ガイシ (NGK Insulators) presented a range of ceramic packaging materials for next-generation semiconductors at a technology exhibition held in Japan from September 9 to 11, 2026, focusing on two increasingly prominent challenges in AI systems: heat dissipation and maintaining high-speed signal quality.
These technologies come at a time when AI designs are moving toward larger chips and greater complexity in interconnecting units, including chiplet architectures. In these environments, improving the chip's performance alone is insufficient; packaging-material properties, such as thermal conductivity, the coefficient of thermal expansion, and signal loss, become influential factors in system stability and component integration density.
Aluminum Nitride Substrate for Managing Heat
日本ガイシ presented an aluminum nitride substrate intended for applications supporting chips or wafers measuring 300 mm or larger. The material's stated thermal conductivity is 170 watts per meter-kelvin, while its coefficient of thermal expansion reaches 4.9 parts per million per kelvin, a figure close to silicon's coefficient of thermal expansion.
This similarity helps reduce stress caused by differences in expansion between the chip and packaging materials during operation and temperature changes. According to the presentation, the company also aims to use this material in processes requiring greater heat-dissipation capability, while addressing challenges related to manufacturing precision and surface quality as size increases.
Co-Packaged Optics (CPO)
The exhibits also included a ceramic package for co-packaged optics (CPO) applications, in which silicon photonic circuits are integrated near processing components rather than being kept separate from the main package. The package combines high thermal conductivity with reduced signal loss, two important factors when transferring data at high speeds.
日本ガイシ is participating in an industrial research project to develop data-center components and systems and, in this context, presented a concept for an active optical package (AOP). The design is based on integrating a silicon photonic circuit inside a ceramic package, using a polymer fan-out structure and an optical coupling method between the circuit and the fiber, including precision optical components.
Why Does This News Matter?
The practical message from these exhibits is that AI expansion is placing as much pressure on the packaging layer as on the processor architecture. Higher power density increases the need for materials that transfer heat efficiently, while the shift toward optical links and interconnected chips makes the electrical and optical properties of packaging materials part of the performance equation.
However, the presentation does not yet establish that these solutions have been adopted on a broad commercial scale or are available as standard products. The source also did not provide prices, production schedules, or customer names. Therefore, the current value of the development remains at the level of materials and packaging technologies being prepared to meet the requirements of AI systems and data centers, rather than representing an announcement of a completed commercial platform.