AI development (training) and deployment (inference) require enormous computational power and energy. With the rapid rise of generative AI
this demand is expected to grow exponentially
creating a global need for computing infrastructure optimized for AI-specific performance and power efficiency. Preferred Networks (PFN) not only utilizes general-purpose GPUs but also develops and integrates its own MN-Core™ series of AI processors with advanced networking and storage technologies to provide cutting-edge computing infrastructure to researchers and developers in and outside of PFN. Since 2024
PFN has also offered the Preferred Computing Platform™ (PFCP™) — the sole cloud service for AI workloads running on the MN-Core™ series.
Preferred Computing Platform™ (PFCP™) is a cloud service built and operated by PFN for deep learning and AI workloads. PFCP is the only platform that provides access to PFN's MN-Core™ series of proprietary accelerators, delivering high AI computing performance and efficiency. Users can exclusively utilize multiple servers equipped with eight MN-Core 2 boards each, with all nodes interconnected by a high-speed network optimized for deep learning. PFCP™ also offers a multi-tenant Kubernetes cluster environment extended specifically for AI workloads.
Operating since May 2020, MN-3 is PFN’s third-generation computer cluster that uses MN-Core, a highly efficient custom processor co-developed by PFN and Kobe University specifically for use in deep learning. PFN is currently working to increase MN-3's computational speed for practical deep learning workloads. MN-3 topped the Green500 list of the world's most energy-efficient supercomputers three times in June 2020, June 2021 and November 2021.
At PFN, we build and operate multiple computing infrastructures powered by the MN-Core series of processors, GPUs, storage systems, and other components to support the advanced AI computation required for our research and business activities.
The integrated computing infrastructures leverage both on-premises and cloud environments, all managed through PFN’s in-house machine learning platform.
PFN’s computing infrastructure is built and operated as a machine learning platform with Kubernetes as its core technology. Through a proprietary scheduler and numerous custom features, it efficiently executes AI and machine learning workloads. The technologies behind this platform are also available through Preferred Computing Platform (PFCP), PFN’s cloud service for AI workloads.
