Nvidia and Fujitsu announced a broad partnership to build a full-stack artificial intelligence infrastructure. The plan centers on AI agents and robotics. The companies said they will combine Fujitsu hardware and software with Nvidia GPUs and interconnect technology.
Fujitsu framed the effort as a long-term push to boost enterprise competitiveness. The two firms said they will customize AI agents for industries such as healthcare, manufacturing, next-generation computing, and consumer services. The partners aim to establish the infrastructure in Japan by 2030.
Nvidia said the work will lean on its software and networking tools. The companies will link Fujitsu CPU designs with Nvidia GPU accelerators using NVLink Fusion. That technology lets systems combine different CPU and GPU vendors into a single high-bandwidth fabric. Nvidia called NVLink Fusion a key enabler for tightly coupled AI compute.
What they will build
Fujitsu described plans to integrate its FUJITSU MONAKA CPU series with Nvidia GPUs. The collaboration will use Fujitsu software for AI workload orchestration and Nvidia platforms such as NeMo and Dynamo. Fujitsu will also tune its Takane model and Kozuchi infrastructure for multi-agent scenarios. The companies said the stack will support continuous learning and secure multi tenancy for enterprise use.
The work will target robotic systems and industry-specific AI deployments. Fujitsu said it intends to work with partners in robotics and machinery to put AI agents into real-world applications. The goal is to address issues such as labor shortages in manufacturing and to offer enterprise-grade solutions that keep customer data under local control.
Why this matters
The move signals a push by Japan to build national AI infrastructure and to field homegrown solutions that can scale. The partners hope to unlock new markets for AI in areas that require tight integration between hardware and software. Observers note that combining many chips with fast interconnects is one path to match larger foreign AI systems while coping with export limits on certain advanced chips.
Nvidia and Fujitsu spoke about broader research and computing work. Executives mentioned high-performance computing and quantum research as future expansion areas. Fujitsu said it will try to create an ecosystem that mixes ARM-based designs, CUDA compatibility, and a one-stop support model for enterprise AI transformation.
