By Diplomacy Journal Lee Kap-soo

Samsung Electronics is working with NVIDIA to advance AI-RAN technologies, the company said on March 12. The collaboration underscores Samsung’s commitment to fostering a robust ecosystem and diversifying the computing platforms available.

This endeavor aims to support a smooth and easy adoption of AI in mobile networks by expanding the ecosystem of Central Processing Unit (CPU) and bolstering partnerships with Graphics Processing Unit (GPU) companies.

To maximize and bring the power of AI into Radio Access Networks (RAN), Samsung has made significant technological progress leveraging its in-house AI and radio expertise since the beginning of 2024.

One of the pivotal milestones achieved was the interoperability between Samsung’s O-RAN compliant virtualized RAN (vRAN) and NVIDIA’s accelerated computing, which took place in Samsung Research’s lab in late last year.

Samsung has successfully demonstrated a proof-of-concept to verify how NVIDIA’s accelerated computing can be seamlessly integrated into software-based networks to help enhance AI capabilities.

This achievement further solidifies Samsung’s strides in moving ahead with its unique innovation of combining AI and RAN. With the baseline, Samsung can seamlessly deliver AI-RAN by integrating its vRAN (virtualized Distributed Unit, vDU) with NVIDIA’s accelerated computing into a commercial-off-the-shelf (COTS) server where Samsung’s vRAN software is installed.

 
In addition, the companies will continue to explore the best-of-breed combinations of AI-RAN options leveraging Samsung vRAN with NVIDIA’s Grace CPU and/or GPU based AI platform using Compute Unified Device Architecture (CUDA) technologies. All of these are optimal for each network deployment environment — from rural, sub-urban to dense urban.

At MWC 2025, Samsung presented its leadership in AI-For-RAN innovations, as exemplified through two AI-RAN demonstrations, endorsed by the AI-RAN Alliance and developed in collaboration with various members including NVIDIA. The demonstrations included AI-based physical uplink shared channel (PUSCH) estimation and non-uniform modulation, showing a glimpse in innovative ways to infuse AI into mobile networks.