IT service companies are moving to capture the physical AI market being applied to industrial sites. While hardware for physical AI—robots and large-scale equipment—is handled by manufacturers, the data and robot/AI platforms that serve as the brain of next-generation factories are seen as an area where IT service companies hold a competitive edge. The shared direction among these companies is developing a factory operating system (OS)that enables robots and equipment on industrial sites to judge tasks autonomously and divide work with human workers, and building self-operating 'intelligent factories' based on this foundation.
The four companies have staked out different layers
POSCO DXhas presented a blueprint for an 'Intelligent Factory' in which AI controls key steelmaking processes and large equipment operates autonomously without human intervention. A core axis of this effort is converting the tacit knowledge of soon-to-retire skilled workers into key data for physical AI. The company is pursuing 'equipment robotization,' applying physical AI to massive pieces of equipment weighing dozens of tons—such as large cranes that move steel products, reclaimers that transport raw materials in the yard, and ship unloaders for raw materials. These systems are currently in the trial-run stage and are expected to be commercialized and deployed at steel mills as early as the second half of this year.
SK AXannounced last month that it is fully rolling out its 'Manufacturing RX Full-Stack Service,' which shifts manufacturing sites toward a robot-centered model. Key features include verifying risk factors in a virtual environment before deploying robots, and integrating real-time monitoring by connecting robots, manufacturing execution systems (MES), and equipment data on the factory floor. An SK AX representative said, "We are accumulating field data in the semiconductor industry and verifying related systems and proof-of-concept models, and we plan to expand this to the shipbuilding industry as well."
LG CNSunveiled its robot transformation (RX) platform 'PhysicalWorks' in May, which manages the entire lifecycle from robot training to integrated monitoring and control. The goal is to shorten the deployment period for industrial robots—previously taking several months—to just one to two months. The company is currently conducting or has completed robot proof-of-concept (PoC) projects with more than 20 major industrial clients, including LX Pantos and Kurly, and recently signed a physical AI infrastructure contract worth 189.7 billion won with LG Electronics.
Samsung SDSis also preparing a platform to control robots deployed at industrial sites. Last month, the company made a strategic investment in U.S.-based Waldon Robotics through Samsung Venture Investment. During an earnings conference call, Samsung SDS stated, "Starting with our investment in Waldon Robotics, we will expand partnerships with global robotics companies and pursue a 'robot orchestration' business that integrates and controls diverse robots while linking them with manufacturing execution systems."
If you're considering adoption, first identify which layer your problem falls into
Placing what each of the four companies offers side by side reveals that they have staked out different layers: automating equipment itself (POSCO DX), pre-deployment verification and integrated monitoring (SK AX), the speed of getting robots onto the floor (LG CNS), and orchestrating and controlling diverse robots together (Samsung SDS). Which layer you should examine first depends on whether your bottleneck is that 'equipment can't make judgments,' 'risks go undetected before deployment,' 'deployment takes months,' or 'each robot is managed under a separate system.'
An LG CNS representative said, "Investment in smart factories and robotics is not about short-term expansion of production capacity, but a strategic response to rising labor costs, workforce shortages, and stricter quality and safety standards." If there's one thing worth checking this week, it's counting from past cases how many months it took—from order placement to actual operation—when your site last introduced a new robot or automation system. Once you have that number, it will also reveal which of the four layers above you should look at first.
Source: 'AI Brain' Linking Factories and Robots... IT Service Industry Races to Own the Factory OS