Fujitsu Limited

In this vendor showdown talk, Yoshihiro Kusano describes Fujitsu's roadmap for future HPC, centered on AI as the core of the company's technology strategy. He introduces FUJITSU-MONAKA, an Arm-based CPU designed for both data centers and edge deployments with high performance and energy efficiency, and then focuses on what follows the CPU-centric Fugaku supercomputer. The successor platform, FugakuNEXT, responds to growing data volumes and the expansion of AI by combining CPUs and GPUs in a node, targeting one hundred times application performance over Fugaku through application modernization. The design philosophy is complementary compute: CPUs handle irregular data structures, complex control flow, and low-latency inference, while GPUs handle regular data structures, massively parallel throughput, and large-scale AI training, coupled tightly through NVLink Fusion with cache coherency and zero-copy data flow. Fujitsu has begun designing MONAKA-X, the CPU for FugakuNEXT, on the Armv9-A architecture with more than 200 cores per socket, SVE2 and SME2 for on-CPU AI processing, and confidential computing support. Kusano cites AI-accelerated simulation, simulation-data-driven science, and agentic simulation as directions the architecture enables. In the question round, he confirms Fujitsu's full commitment to Arm on grounds of software maturity and ecosystem growth, indicates multiple SKU options spanning core to edge data centers, identifies MONAKA-X with its SME2 matrix units as the stronger inference processor, and states that SKUs both with and without NVLink Fusion are planned so other accelerators can also be supported.