What is Physical AI?
Physical AI refers to the ability of artificial intelligence not only to analyze digital data but also to perceive information from the real world and translate it into physical actions. In this approach, robots and other machinery can evaluate ambient conditions, make decisions based on derived data, and execute appropriate physical movements.
In traditional industrial robotics, movements are largely executed via predefined programs and fixed operating conditions. The Physical AI approach, however, aims to combine artificial intelligence, sensor data, robot control systems, simulation, and real-time data into an integrated operational framework.
Through this architecture, the goal is to enable robots to adapt more easily to changing production conditions, execute tasks with greater flexibility, and collaborate more effectively with humans.
What Are Fujitsu and the Robotics Companies Planning?
According to Fujitsu’s announcement on July 16, 2026, the company is exploring business opportunities with FANUC, Yaskawa Electric, and Kawasaki Heavy Industries to apply Physical AI across various industries. At the core of this initiative lies a common control platform designed to bridge the digital and physical worlds.
The goal of this platform extends beyond controlling a single robot. Fujitsu plans to consolidate diverse technologies—such as AI, robotics, control systems, simulation, and data analytics—onto a unified infrastructure.
Furthermore, the company aims for this infrastructure to support interoperability among different robots and equipment, facilitating the creation of more advanced autonomous control systems.
This approach is particularly crucial because robots do not operate in isolation in modern factories. Production lines consist of numerous interconnected components, including robotic arms, conveyors, sensors, PLCs, machine vision systems, and high-level manufacturing software. One of Physical AI's primary objectives is to enable these disparate systems to operate in a more coordinated manner based on real-time data.
Where Do NVIDIA Technologies Fit into the System?
A key component of this initiative is NVIDIA's Physical AI ecosystem.
Fujitsu disclosed that it will leverage NVIDIA's AI, world model, simulation, and robotics technologies. Specifically, the company plans to utilize platforms such as NVIDIA Cosmos, NVIDIA Omniverse, NVIDIA Isaac, and the Newton physics engine.
NVIDIA Cosmos forms a foundational framework for world models and physical AI applications used to understand real-world environments and predict future events.
Meanwhile, the NVIDIA Omniverse and Isaac platforms serve to train, test, and optimize robots in virtual environments before deploying them into the physical world.
A prominent concept here is the Sim2Real (Simulation to Reality) approach. Allowing a robot to first learn and validate a specific task in a simulated environment before transferring it to physical hardware is critical for reducing development lead times and ensuring controlled testing on actual machinery.
Why Are FANUC, Yaskawa, and Kawasaki Important?
The participation of FANUC, Yaskawa Electric, and Kawasaki Heavy Industries is notable because it directly connects Physical AI with core industrial robotics infrastructure.
While FANUC focuses on advanced robot control technologies and open platform support, Yaskawa develops autonomous robot technologies like MOTOMAN NEXT powered by NVIDIA GPUs. Yaskawa is also actively working on ROS 2 compatibility and promoting the use of robots as open platforms.
Kawasaki Heavy Industries complements the robotics side of the initiative with its extensive industrial robotics technologies.
Rather than unifying these companies' robots into a single monolithic product, the objective is to build a common control infrastructure where diverse robotic and digital technologies can seamlessly interoperate.
What Can Physical AI Deliver in Factories?
One of the primary initial application domains in Fujitsu's roadmap is factory environments.
The company aims to optimize the overall planning of manufacturing operations by factoring in production variables and real-world shop-floor conditions, enabling robots to autonomously adapt to these dynamic factors.
For instance, when conditions on a production line shift, rather than relying on a system that merely executes a predefined command sequence, a more flexible architecture can be deployed—one that analyzes environmental data, evaluates the production plan, and determines optimal actions.
The primary objective of this approach is not to make robots completely autonomous entities, but rather to integrate sensing, decision-making, and physical execution processes with a more sophisticated AI foundation.
Ultimately, this aims to deliver higher flexibility, increased automation rates, and enhanced productivity across manufacturing operations.
Applications Beyond Manufacturing: Logistics and Healthcare
The scope of Physical AI is not limited strictly to factory floors.
In logistics, Fujitsu aims to automate material handling operations based on logistics plans that continuously adapt to real-time sales and inventory data. This structure can assist robots in warehouses and distribution facilities to adjust their tasks according to evolving operational demands.
In healthcare, robots are envisioned to autonomously handle tasks such as transporting pharmaceuticals and laboratory samples based on instructions generated by hospital management systems. Patient reception and guidance services are also among the potential use cases evaluated by Fujitsu.
Thus, the goal is to apply Physical AI not only to manufacturing equipment but also to diverse robotic systems operating across various physical domains.
Targeting an Open and Common Control Infrastructure
A striking aspect of the initiative is the emphasis on standardization and openness for the control infrastructure to be developed.
Fujitsu plans to build a platform open to participation from external companies and research institutions. This platform is intended to serve as a standardized interface between digital systems and robotic hardware.
However, interconnecting robots and equipment more deeply brings forth heightened security challenges. Risks such as cyberattacks, system-wide failures, production downtime, and sensitive data exposure become increasingly critical as physical AI systems proliferate.
Consequently, the advancement of physical AI relies not only on building smarter robots, but also on establishing secure control frameworks and robust data architectures to manage them safely.
Physical AI Could Mark a New Era in Industrial Robotics
The initiative between Fujitsu, FANUC, Yaskawa Electric, and Kawasaki Heavy Industries does not yet represent a finalized physical AI platform or a fully autonomous production system. At this stage, the companies are exploring joint business opportunities to develop and deploy these technologies across multiple industries.
Nevertheless, the initiative signals a major strategic direction in industrial robotics. It reflects an evolution where robots transition from machines executing pre-programmed movements into intelligent systems that capture environmental data, analyze it via AI, and adapt their physical behavior to changing conditions.
By converging AI, robotics, simulation, control systems, and real-time data onto a unified platform, Physical AI could pave the way for broader adoption not only in factories, but also in logistics, healthcare, and other physical operations in the coming years.
This is precisely where the core significance of the current initiative lies: Merging the digital capabilities of AI with the physical motion and control capacities of robotics represents one of the next pivotal steps in industrial automation.