Semidynamics Launches 3nm AI Inference Silicon and Full-Stack Systems for Data Centers
Semidynamics has launched its high-performance 3nm AI inference silicon and full-stack systems aimed at enhancing data center capabilities. Achieving silicon readiness in December 2025, the company has developed an innovative memory subsystem to improve efficiency in AI workloads, supported by partnerships with notable organizations. This initiative is part of Semidynamics' strategy to provide a comprehensive stack for advanced multi-accelerator AI platforms in Europe, promoting digital sovereignty and technological autonomy in the region.

Semidynamics has announced its entry into full-stack AI infrastructure, unveiling high-performance inference silicon and vertically integrated systems. The company achieved silicon readiness at the 3nm node in December 2025, marking a significant milestone with TSMC that aims to deliver production-grade chips and systems.
The new silicon architecture includes an innovative memory subsystem to enhance efficiency in AI workloads, addressing bandwidth bottlenecks and supply constraints. Partnerships with organizations such as EuroHPC Joint Undertaking, HPE, Bull (Atos Group), Telefonica, Multiverse Computing, E4 Computer Engineering, and Megware support this initiative. Semidynamics’ strategy focuses on developing a comprehensive stack of chips, boards, and racks to cater to advanced multi-accelerator AI platforms in Europe, contributing to the region's digital sovereignty and technological autonomy.




Comments