ABB and NVIDIA: Delivering Industrial-Grade Physical AI

ABB and NVIDIA are making bold moves in the realm of industrial AI, with  ABB Robotics announcing the integration of NVIDIA Omniverse libraries into RobotStudio, its software programming, design and simulation suite for industrial robotics.

The collaboration enables manufacturers to deploy physical AI in real-world robotics applications through accurate digital simulation and synthetic data generation.

By combining NVIDIA accelerated computing and simulation technologies with ABB robotics software, the companies are bidding to close the gap between virtual training environments and real-world deployment.

Marc Segura, President of ABB Robotics, explains: "Using NVIDIA accelerated computing and simulation technologies, we have removed the last barriers to making industrial and physical AI a reality at a global scale by closing the sim-to-real gap.

"For more than 50 years, ABB Robotics has led the evolution of intelligent industrial automation, from pioneering the first generation of fully-electric industrial robots to advancing digital twin simulation through RobotStudio and shaping a new area of autonomous and versatile mobile robots. Today’s announcement with NVIDIA brings physical AI to industry at scale."

Combining RobotStudio with Omniverse simulation

The partnership focuses on linking RobotStudio with the simulation power of NVIDIA Omniverse libraries, which provide a platform for physically-accurate simulation.

Developers use these environments to create digital twins, allowing engineers to test processes and automation workflows in a virtual environment before deployment in the real environment.

Through RobotStudio and Omniverse, developers generate synthetic data. Synthetic data refers to computer-generated datasets used to train AI models when real data is limited or difficult to capture. In robotics development this allows robots to practise tasks across many simulated scenarios.

The combined platform introduces RobotStudio HyperReality, which uses simulation and data feedback from real operations to refine foundation models. These models train ABB robots anywhere in the world. Real-world operational data feeds back into the system so simulations remain accurate and continuously improve.

Deepu Talla, VP of Robotics and Edge AI at NVIDIA, comments: "The industrial sector needs physically-accurate simulation to bridge the gap between virtual training and the real-world deployment of AI-driven robotics at scale.

"Integrating NVIDIA Omniverse libraries into RobotStudio brings advanced simulation and accelerated computing to ABB Robotics’ unique virtual controller technology, accelerating how manufacturers of all sizes bring complex products to market."

The ‘sim-to-real’ gap

The sim-to-real gap describes the difference between results achieved in simulation and performance in the real world. In robotics development, virtual testing often fails to reflect real conditions such as lighting variation, material behaviour or environmental complexity.

This gap limits the ability of manufacturers to design advanced manufacturing processes entirely within virtual environments. When simulations do not match physical performance, engineers must conduct repeated physical testing.

ABB Robotics is addressing this challenge by integrating Omniverse libraries into RobotStudio. The system delivers robotics simulation and synthetic data generation that reach up to 99 percent accuracy between virtual models and real-world operations.

ABB maintains an advantage through its virtual controller which runs the same firmware as the physical device. Because the virtual controller uses identical firmware to the hardware controller, simulation behaviour closely matches physical robot behaviour.

This is combined with ABB Robotics’ Absolute Accuracy, reducing positioning errors from 8-15 mm to around 0.5 mm, enabling precise motion control. This precision supports high-accuracy industrial applications in both digital simulation and physical manufacturing environments.

Manufacturers can therefore design, test and optimise production lines virtually, reducing setup and commissioning time by up to 80% and cutting costs by up to 40% by removing the need for physical prototypes. The speed at which complex products such as consumer electronics reach production lines increases by 50%, according to ABB Robotics analysis. 

Meanwhile, ABB Robotics is assessing the integration of the NVIDIA Jetson edge computing platform into its Omnicore controller, enabling real-time AI inference.

Industrial testing and early deployments

RobotStudio HyperReality is set to support industrial clients across multiple sectors. Select customers are already testing the tech before a wider release to ABB Robotics’ 60,000 RobotStudio users in the second half of 2026.

Foxconn, the world’s largest electronics contract manufacturer, is piloting the first joint use case in consumer electronics assembly, requiring precise pick-and-place and assembly control. 

Using RobotStudio HyperReality, Foxconn trains assembly robots in a virtual environment using synthetic data. The robots practise different production scenarios in simulation before moving to real production lines with 99% – removing the need for physical training and reduces setup time and cost.

Dr Zhe Shi, Chief Digital Officer of Foxconn, explains: "Precision is everything in consumer electronics manufacturing and, until now, this level of accuracy and fidelity just wasn’t possible in simulation and digital twins.

"We’re incredibly excited by the potential of ABB Robotics and NVIDIA’s collaboration, which enables parallel engineering for better designs, faster production ramp-up and greater product evolution through advanced AI inference and understanding."

Elsewhere, WORKR, a California-based robotic workforce company that delivers manufacturing solutions, is extending the tech's reach to small and medium-sized businesses across the US. 

At the upcoming NVIDIA GTC 2026 in San Jose, WORKR will demonstrate AI-powered robotic systems built on ABB technology, trained using synthetic data generated through NVIDIA Omniverse libraries.

Ken Macken, CEO and Founder of WORKR, adds: "This collaboration is about making industrial AI deployable today. Together with ABB and NVIDIA, we're proving that advanced automation can work for manufacturers of any size."