Physical AI’s goal is to support all kinds of robots, reducing latency and cost by putting processing power on the network close to them
SK Telecom (SKT) and KT have begun deploying networks for physical AI at industrial sites like shipyards and petrochemical plants according to repoorts in local Korean media.
Physical AI systems can perceive, reason and act in the real world, but this generates a massive computational burden and the cost of devices rises, undermining commercialisation. In physical AI, the network infrastructure shares the AI computation at the edge or RAN, and supports real-time data analysis and control to improve performance and efficiency at lower cost.
A new study by global tech strategist firm Juniper Research predicts deployments of physical AI systems in manufacturing and logistics will reach 400,000 by 2030. This would reflect a growth rate of 3,500% from 2026, driven by advances in real-time processing and AI models, which promise to improve the capabilities and safety of physical AI operating in real-world environments.
The Korean pilot
The pilot will last two years and be overseen by the South Korea’s Ministry of Science and ICT and the National Information Society Agency (NIA). The contract to build the network was awarded to the two operators in mid June as part of the government’s Hyper AI Network Foundation Development.
The network will combine 5G SA and AI-based RAN to provide “ultra-low-latency, high-capacity communication”. The aim is to create an environment where multiple robots can operate reliably at industrial sites based on this network.
The Seoul Economic Daily quotes KT saying, “Based on our AI-based autonomous operation network technology and physical AI demonstration experience, we plan to expand our B2B AX business into various industries such as manufacturing, logistics, and energy.”
SKT was cited saying, “We plan to further advance the AI-RAN pilot network going forward and link it with SK Group’s AI data center infrastructure strategy.” Last October, industry commentator Sebastian Barros called SKT “the only telco that actually gets AI“.
SKT plans simultaneously to build and demonstrate AI-RAN equipment from four manufacturers — Samsung Electronics, HFR, Ericsson, and Nokia. It will then verify the structure most suitable for physical AI services.
Throughout, the two carriers will work with companies in manufacturing, shipbuilding and other sectors to develop and demonstrate physical AI services tailored to each industrial site.
Different types of robot
SKT will demonstrate three physical AI services — a quadruped patrol robot, a humanoid low-power mode and unmanned autonomous transport services. The demos will be at the SK Incheon Petrochemical and KG Mobility plant sites.
According to Korea Times, in the first year, SK Telecom will build AI-RAN networks in Incheon, using Samsung equipment for the patrol robot service at SK Incheon Petrochemical, and in Pangyo, using HFR equipment for the autonomous transport service. The second year will expand testing to KG Mobility’s Pyeongtaek plant using Ericsson’s technology.
The services will implement real-time analysis and response for hazardous areas within factories, as well as remote driving services based on digital twins. The humanoid low-power mode aims to improve the efficiency of battery usage by distributing the robot’s AI computation to base stations.
KT will work with HD Hyundai Samho to develop an AI-based autonomous system able to analyse and control shipyard robot and equipment data in real time. It will demonstrate three physical AI services, namely a welding robot, a painting robot and an autonomous operation robot for communication facilities. The plan is to verify the potential for improving safety and productivity at hazardous industrial sites.
