Samsung Electronics has achieved a significant advance in AI-RAN capabilities through a collaborative effort with NVIDIA, successfully completing a multi-cell test integrating Samsung’s virtualized RAN software with NVIDIA’s accelerated computing platform. Validated at Samsung’s research and development center, this integration represents a key step toward the commercial deployment of AI-enhanced networks and will be showcased at Mobile World Congress 2026. The demonstration features an AI-based downlink performance boost leveraging AI algorithms for improved throughput and higher spectral efficiency for operators. “As AI-powered capabilities become integral to meeting the demands of evolving networks and growing traffic needs, Samsung’s vRAN takes center stage with its software-based architecture,” said Keunchul Hwang, Executive Vice President and Head of Technology Strategy Group, Networks Business at Samsung Electronics.
Samsung and NVIDIA Validate AI-RAN Performance with vRAN Integration
This validation in a realistic network environment represents a significant step toward the commercial deployment of enhanced wireless networks, and builds upon Samsung’s existing leadership in 5G solutions encompassing chipsets, radios, and cores. Samsung and NVIDIA are also jointly pursuing innovations to accelerate data exchange between CPUs and GPUs, leveraging Samsung’s vRAN and NVIDIA’s unified processor for optimal performance and cost. The collaboration extends to the integration of Samsung’s vRAN software with NVIDIA’s ARC Compact, equipped with NVIDIA Grace CPU and L4 GPU, completed last month as a further move towards commercialization. This software-based approach provides operators with greater flexibility and performance, supported by an ecosystem of partners including NVIDIA.
NVIDIA’s Soma Velayutham affirmed the importance of this work, stating, “Operators today need AI-native, software-defined infrastructure to stay ahead of evolving connectivity demands. Samsung’s successful multi-cell validation and innovative AI beamforming solution on NVIDIA AI Aerial mark an important milestone towards AI-RAN commercialization.” By combining large-scale vRAN deployment experience with NVIDIA’s computing platform expertise, Samsung aims to lead the advancement of AI-powered, software-based networks, including multiple-input multiple output technology for increased data speeds and capacity.
The demonstration highlights an AI-based system—an AI MIMO 1 beamformer—that utilizes AI algorithms to achieve higher spectral efficiency, allowing operators to maximize capacity from their existing spectrum allocations. Samsung’s commitment to advancing AI enhancements within its software-driven network portfolio is evident through this collaboration with NVIDIA platforms.
The foundation of this enhanced functionality lies in the shift from dedicated hardware processing to sophisticated software-defined infrastructure. Virtualization decouples the radio functions (RF) from specialized hardware, allowing operators to rapidly scale, optimize, and allocate resources for specific services. This architecture enables dynamic network slicing, where dedicated virtual network segments can be provisioned for diverse use cases, such as mission-critical public safety communications or low-latency industrial IoT applications, ensuring guaranteed quality of service (QoS) irrespective of network load.
Furthermore, the deep integration with accelerated platforms like NVIDIA GPUs enables real-time processing of massive data streams required for advanced techniques such as beamforming and Massive MIMO signal processing. Traditional Signal Processors (SPs) struggle with the computational complexity of these AI algorithms, but leveraging heterogeneous compute architectures allows the network to achieve near-instantaneous adjustments to beam patterns, maximizing signal strength and mitigating interference across wide geographical areas.
A key challenge addressed by this collaboration is the latency introduced by coordinating data between disparate computing units. By implementing unified memory architectures and high-speed interconnects, Samsung and NVIDIA are minimizing the data transfer overhead, which is critical for maintaining the ultra-low latency requirements essential for next-generation services, including autonomous vehicles and remote telesurgery.
From an industry perspective, the movement toward AI-RAN marks a fundamental evolution from simply increasing spectrum capacity to optimizing spectral efficiency itself. The ability to use sophisticated machine learning models not only improves throughput but also predicts traffic fluctuations, preemptively reallocating resources across the cell structure before congestion occurs, thereby ensuring sustained, high-quality connectivity for all users.
As AI-powered capabilities become integral to meeting the demands of evolving networks and growing traffic needs, Samsung’s vRAN takes center stage with its software-based architecture.
Keunchul Hwang, Executive Vice President and Head of Technology Strategy Group, Networks Business at Samsung Electronics
