Samsung is framing physical AI as a turning point for telecom networks. The company believes networks must move beyond basic connectivity. They must also coordinate computing, data, and automation in real time.
Physical AI refers to systems that sense and act in the real world. Examples include robots, cameras, drones, and industrial machines. These systems need fast responses and stable links. They also generate large amounts of video and sensor data.
Jungchul Kim, head of product strategy group for Samsung’s networks business, outlined the shift. He said physical AI will demand ultra-low latency, high reliability, strong uplink capacity, and autonomous network operations.
South Korea will provide an important testbed for this idea. Samsung will work with KT and SK Telecom under the government’s Hyper AI Network initiative. The projects will use standalone private 5G networks in demanding industrial sites.
The trials will cover shipyards and petrochemical facilities. At HD Hyundai Samho’s Yeongam Shipyard, KT will test robots for welding and painting. It will also test autonomous robots for telecom facility operations.
Meanwhile, SK Telecom will work at SK Incheon Petrochem. It will validate autonomous patrol robots and CCTV-based monitoring. These systems will send high-definition video while moving through the facility. AI tools will then analyze risks and support safety teams.
This model offers clear value for industrial operators. Faster detection can improve safety. Automated inspections can reduce human exposure in difficult areas. Better uplink performance can also support richer video and sensor streams.
However, the approach raises major network challenges. Operators must invest in edge computing and smarter orchestration. They must also decide where AI workloads should run. Some tasks belong on devices. Others need edge servers or cloud platforms.
Samsung will supply its Network in a Server architecture for the trials. The platform combines virtualized RAN, AI Core, and AI applications. It runs on commercial off-the-shelf servers, which may simplify deployment.
The larger question is how networks manage intelligence across many locations. Physical AI cannot wait for distant cloud responses in every case. It needs local decisions when timing matters. It also needs cloud resources for heavier analysis.
“The network will need to do more than connect AI applications; it will need to provide the intelligent, distributed infrastructure that enables them to operate reliably in the physical world,” Kim said.
For Samsung, these projects also support long-term AI-native 6G research. The trials may shape future network design. They could help carriers understand where connectivity ends and distributed intelligence begins.
