Fujitsu expanded its collaboration with NVIDIA to build a full-stack AI infrastructure that combines industry-specific AI agents with high-performance compute. The plan integrates Fujitsu’s FUJITSU-MONAKA Arm-based CPU series with NVIDIA GPUs using NVLink Fusion to target sectors such as healthcare, manufacturing, and robotics in Japan first, with global expansion to follow. The companies said the joint stack aims to let enterprises deploy secure, self-evolving AI systems while maintaining autonomy over their data and models.
The collaboration focuses on three workstreams: a multi-tenant AI agent platform built on Fujitsu Kozuchi with orchestration tied to NVIDIA’s Dynamo and NeMo; co-development of next-generation computing that marries MONAKA CPUs and NVIDIA accelerators via NVLink Fusion for scale-up AI; and a partner ecosystem to seed vertical use cases and deployments. Fujitsu and NVIDIA plan to package agents as NVIDIA NIM microservices to streamline inference rollout for enterprises.
Initial targets include digital-twin manufacturing, robotics and other “physical AI” applications to address labor constraints, and cross-industry automation. NVIDIA describes NVLink Fusion as a path for semi-custom AI systems that tightly couple third-party CPUs with NVIDIA accelerators, while Fujitsu frames MONAKA as its next-gen Arm CPU for data-center AI and HPC workloads slated toward the latter half of the decade.
• Co-develop an industry AI-agent platform (Kozuchi + NVIDIA Dynamo/NeMo), delivered as NIM microservices for easier enterprise deployment.
• Integrate FUJITSU-MONAKA CPUs with NVIDIA GPUs over NVLink Fusion for high-bandwidth scale-up AI.
• Optimize the stack “from silicon to system,” aligning Arm CPU memory/IO with GPU interconnects for AI factories.
• Build a partner ecosystem to accelerate vertical use cases in manufacturing, healthcare, and robotics.
• Begin in Japan with a roadmap to global rollout.
“Fujitsu’s strategic collaboration with NVIDIA will accelerate AI-driven business transformation in enterprise and government sectors… we will develop and provide full-stack AI infrastructure, starting with sectors such as manufacturing where Japan is a global leader,” said Takahito Tokita, CEO of Fujitsu.
🌐 Analysis: MONAKA background & Fujitsu’s HPC lineage
Fujitsu positions FUJITSU-MONAKA as a next-generation Armv9 CPU for data-center AI/HPC, featuring a many-core design, advanced packaging, and an announced timeline that points toward availability around 2027. Public materials indicate goals such as high energy efficiency, expanded memory channels, and modern IO (e.g., DDR5/CXL, PCIe 6), aligning the CPU for coupling with GPU-dense nodes via fabrics like NVLink Fusion.
Fujitsu’s legacy in HPC underpins this move. The company co-developed the A64FX processor and RIKEN’s “Fugaku” supercomputer, which hit #1 on TOP500 in June 2020 and remains a large-scale Arm system noted for efficiency and scientific output. That expertise in system architecture, interconnects, and large-scale operations provides credibility for building CPU-GPU topologies aimed at AI factories.
Side note: Japan’s government-backed role
Japan’s public agencies (including MEXT for Fugaku and programs via METI/NEDO supporting next-gen compute R&D) have historically co-funded strategic HPC initiatives. MONAKA communications reference development within broader national programs to advance domestic compute—context that helps explain the NVIDIA-Fujitsu alignment around an AI infrastructure foundation for Japan by 2030.
🌐 We’re tracking the latest developments in AI infrastructure. Follow our ongoing coverage at: https://convergedigest.com/category/ai-infrastructure/
