"After Generative AI Comes Physical AI"... Synthetic Data Emerges as Core Infrastructure for Global AI

As the era of 'Physical AI'—AI that perceives, judges, and acts within real-world environments—begins in earnest beyond generative AI, the axis of global competition in the AI industry is shifting toward data infrastructure. To implement AI in the real world, such as in robotics, autonomous driving, and smart factories, it is essential to secure training data that closely resembles actual environments, and synthetic data is emerging as a key foundation supporting this effort.

Synthetic data refers to AI training data generated in a digital twin-based virtual environment instead of repeatedly collecting real-world data. The core lies in creating a training environment that closely mirrors industrial sites, not just with simple 3D modeling, but by reflecting physical elements such as material, surface texture, reflectance, lighting, and sensor errors.

The industry views the traditional approach of repeatedly training robots in the field as limited in terms of cost, time, and the acquisition of exception data. As a result, the 'simulation-to-real' (Sim-to-Real) structure—where training and validation occur first in virtual environments before being applied in reality—is expected to become more widespread. Accordingly, the importance of 'simulation-ready data,' which can be immediately utilized in actual robotics, autonomous driving, and smart factory environments, is growing, rather than just the sheer volume of data generated.

A representative example is the collaboration between global industrial automation company ABB and NVIDIA. ABB Robotics has established a partnership with NVIDIA to develop next-generation autonomous industrial robots and has announced plans to create an industrial AI simulation environment by combining its own robot simulation platform, 'RobotStudio,' with NVIDIA's 'Omniverse.'

ABB is enhancing its structure by integrating NVIDIA Omniverse into RobotStudio, its robot design, programming, and simulation software, allowing industrial robots to learn and be validated in virtual environments before being deployed in actual factories. NVIDIA is also focusing on developing Sim-to-Real technology to minimize discrepancies between virtual and physical environments through its Omniverse and Isaac Sim-based physical simulation technologies.

Industry insiders interpret this collaboration not as a mere technical alliance, but as a strategic move to secure dominance in the synthetic data and simulation ecosystem, which is a core infrastructure in the Physical AI era. The competition to develop technology that enables AI models trained in virtual environments to reliably operate in real industrial settings such as manufacturing and logistics is intensifying.

In Korea, efforts targeting the synthetic data market for Physical AI training are also expanding. However, the field of industrial synthetic data, which requires implementation of spatial structures, lighting, object interactions, robot trajectories, and physics-based simulations for industrial settings, is considered to have high technological barriers.

Among these, SKAI Intelligence is receiving attention as an industrial synthetic data company based on NVIDIA Omniverse. The company recently established a corporate-affiliated research institute (R&D center) to advance digital twin and synthetic data technologies and is working to strengthen its capabilities in designing Real-to-Simulation and Simulation-to-Real structures.

SKAI Intelligence is focusing on advancing its synthetic data infrastructure by linking not only simple 3D data generation but also industrial site structures, object interactions, robot trajectories, and physics-based simulations. In addition, other domestic companies such as Xiilab, N.LIGHT, and CrowdWorks are accelerating their entry into the related market based on their capabilities in digital twins, 3D CAD, and AI data development.

The market's growth potential is also steep. According to global market research firm Grand View Research, the global synthetic data market is expected to grow from USD 218.4 million in 2023 to USD 1.7881 billion by 2030. As the application of Physical AI expands in industries such as robotics, autonomous vehicles, and smart factories, demand for industrial synthetic data is also expected to rise rapidly.

An industry official commented, "In the era of Physical AI, synthetic data serves as the core training infrastructure that enables robots to experience countless exceptional scenarios before actual deployment," adding, "Future competitiveness will depend on how precisely the real world can be digitized and how effectively this can be linked to AI training data and real-world performance improvement."

This content was produced with the assistance of AI translation services.