Fujitsu and NVIDIA Join Forces to Build “Physical AI” Platform

Fujitsu and NVIDIA have announced an expanded partnership to create a powerful new AI platform that blends advanced computing hardware, AI agents, and industry-specific applications.

The goal is to bring “Physical AI”—AI that can operate in the real world through robots, digital twins, and automation—into factories, hospitals, and infrastructure systems both in Japan and worldwide.

What’s New

The two tech giants are working together to deliver a full-stack AI solution. This means everything from CPUs and GPUs to orchestration software, AI agents, and real-world applications will be tightly integrated.

The initial rollout will start in Japan, where challenges such as labor shortages and an aging workforce are driving demand for robots and automation. Later, the companies plan to expand the technology globally.

Key focus areas include:

• Robots and factory automation.
• Healthcare and hospital support.
• Environmental monitoring.
• Customer service and digital twins.

Core Components of the Platform

The new platform combines Fujitsu’s and NVIDIA’s strengths into one end-to-end solution:

AI Agent Platform: Fujitsu and NVIDIA will co-develop a self-learning AI agent system.

It will be built on Fujitsu’s Kozuchi AI framework, combined with NVIDIA’s Dynamo and NeMo tools. Agents will be delivered as NVIDIA NIM microservices, making deployment easier for businesses.

Compute Hardware: Fujitsu’s new MONAKA CPUs will be linked directly to NVIDIA GPUs using NVLink-Fusion, a high-speed connection technology.

This setup allows multiple chips to function as one, boosting performance while saving energy.

Software Stack: The ecosystem combines Fujitsu’s orchestration software and ARM-optimized systems with NVIDIA CUDA, enabling companies to access all AI tools from one platform.

Physical AI for Real-World Use

The companies describe “Physical AI” as AI that doesn’t just run in the cloud but operates in real-world environments. For example:

Robots on factory floors that learn and adapt to tasks.

Digital twins that simulate and improve production lines.

AI systems that help doctors and nurses in healthcare settings.

Executives say the goal is “human-centric AI”—technology that works alongside people and supports industries facing worker shortages.

Partnerships with robotics companies like Yaskawa are being considered to bring these solutions into factories.

Link to Japan’s FugakuNEXT Supercomputer

The partnership is also tied to FugakuNEXT, Japan’s upcoming supercomputer project led by RIKEN, Fujitsu, and NVIDIA.

The new system will deliver 5–10× more performance than today’s Fugaku and, for the first time, use GPUs as accelerators.

This is part of Japan’s push toward zettascale computing—an enormous leap in computing power—by around 2030.

Industry Rollout

The companies plan to launch a reference ecosystem, including industry partners, developers, and integrators. Customers will be able to access AI agents as microservices, making it easier for companies to adopt the technology step by step.

The rollout will prioritize Japan first, followed by global expansion. Fujitsu brings local domain expertise, while NVIDIA provides its vast developer community and partner network.

Energy Efficiency and 2030 Goals

One of the biggest challenges in AI today is energy use. Training and running large AI systems consumes enormous amounts of power.

Fujitsu and NVIDIA are targeting energy-efficient AI chips and systems by 2030. By tightly integrating CPUs and GPUs, they expect to dramatically improve performance per watt, reducing costs for data centers, robotics, and national AI infrastructure.

News Gist

Fujitsu and NVIDIA expanded their partnership to build a full-stack Physical AI platform.

Combining MONAKA CPUs, NVIDIA GPUs, and AI agents, the initiative targets robots, healthcare, manufacturing, and zettascale supercomputing, with energy-efficient systems planned by 2030.

FAQs

Q1: What is “Physical AI”?

Physical AI refers to AI that operates in the real world, powering robots, digital twins, factory automation, and healthcare systems.

Q2: What did Fujitsu and NVIDIA announce?

They unveiled a joint end-to-end AI platform combining CPUs, GPUs, orchestration software, and AI agents, focused on robots and industry applications.

Q3: What is special about MONAKA CPUs and NVLink Fusion?

Fujitsu’s MONAKA CPUs connect directly to NVIDIA GPUs using NVLink Fusion, enabling high bandwidth, energy-efficient systems for AI and supercomputing.

Q4: How does this tie to FugakuNEXT?

The collaboration supports FugakuNEXT, Japan’s next supercomputer, targeting zettascale-class performance and 5–10× improvements over Fugaku.

Q5: Which industries will benefit first?

Manufacturing, robotics, healthcare, environment monitoring, and customer service are initial targets, with Japan as the rollout base.

Q6: What is the 2030 goal?

By 2030, Fujitsu and NVIDIA aim to deliver energy-efficient AI chips, global Physical AI infrastructure, and zettascale supercomputing systems.