In April 2026, Japanese mobile operator KDDI deployed Samsung’s NIS (Network in a Server) at a LAWSON convenience store located inside KDDI’s headquarters in Tokyo, establishing a Private 5G network and conducting an AI-on-RAN PoC. The PoC featured two retail applications: Intelligent Robot Control, developed by KDDI Research, and AI Video Analytics, developed by NEC.

1. AI Video Analytics: AI running on the AI-RAN server receives and analyzes video footage from the convenience store in real time over the Private 5G network to understand on-site situations. For example, as illustrated in the figure below, it can quickly detect and respond to situations such as suspected shoplifting, a fire, or a customer’s lost property, enabling the store operator to manage the retail store more efficiently.

2. Intelligent Robot Control: Ensuring safety when a person approaches a product-stocking robot while it is operating in a convenience store aisle. Video footage captured by cameras installed in the convenience store is transmitted over the Private 5G network to Edge AI (AI running on the AI-RAN server), where it is analyzed in real time. The AI detects the distance between the person and the robot and continuously assesses the risk of collision. When a person approaches the robot, the AI instructs the product-stocking robot to slow down or stop. The robot then autonomously adjusts its operation to ensure safety. This use case demonstrates how the low-latency characteristics of 5G can be leveraged for real-time robot safety control.

The Samsung NIS used as the AI-RAN server in this demonstration was configured as a Supermicro general-purpose server equipped with an AMD EPYC CPU and an NVIDIA L4 GPU card. The server runs the 5G Core, 5G vCU, vDU, and AI applications on a single server (AMD CPU: 5G Core and 5G vRAN, NVIDIA GPU: AI applications)

Cost and Speed: According to KDDI, a key advantage of its AI-RAN (NIS) approach is that it can rapidly establish an Edge AI infrastructure by leveraging existing base-station assets and simply adding GPU cards, without requiring large-scale new infrastructure investments.

[Sources]

Fierce Network, Samsung's Network in a Server: Future-Ready Enterprise Use Cases, 2026.06.22
Samsung Networks, AI-on-RAN in Action: Transforming Retail, 2026.04.26
KDDI, 仮想化基地局がエッジAIの拠点に～コンビ二の品出しロボットをリアルタイムに制御～, 2026.02.26