The telco industry finds itself at pivotal moments when it advances past the midpoint of the cycle toward next-generation networks and when it embraces transformative technologies. With the rise of artificial intelligence (AI) redefining industries across the board, today’s mobile networks are at the brink of the next era.

In response, the industry is discussing and testing how to bring AI into the network—including the radio access network (RAN), which constitutes a significant portion of the end-to-end network architecture and is the driving force behind its evolution. One prominent move was the establishment of the AI-RAN Alliance. The Alliance classifies the integration of AI with RAN into three distinct categories: AI-for-RAN, AI-on-RAN, and AI-and-RAN.

As a founding member and the Vice Chair of the Board of Directors, Samsung is actively engaged in all working groups of the Alliance. Samsung’s expertise in end-to-end software-based networks embraces an even broader view, while still aligned with the three categories defined by the Alliance.

AI-for-RAN: Enhancing Network Performance and Energy Efficiency

The AI-RAN Alliance defines AI-for-RAN as the use of AI in the RAN to improve performance, capacity, and energy efficiency. Samsung’s approach to AI-for-RAN involves implementing AI-powered solutions and applications across all network layers, as part of Samsung CognitiV Network Operations Suite (NOS). This is Samsung’s end-to-end network automation solution, deployed at central data centers, providing a comprehensive, intent-based automation with key applications such as:

• Energy Saving Manager (ESM), which predicts traffic patterns of cell sites and automatically finds and orchestrates diverse RAN energy saving features, has already achieved 35% total energy reduction across LTE, 5G C-band, and mmWave configurations in a U.S. commercial deployment
• RAN Speed Optimizer (RSO), which optimizes and recommends cell-specific parameters to improve throughput, has achieved up to 12% increase in average downlink throughput on a live 5G network
• Load Balancing Manager (LBM), which predicts network congestion and offloads traffic from congested frequency bands to others
• KPI Anomaly Detector (KAD), which identifies anomalies by analyzing historical performance of a cell or a cluster
• RAN Anomaly Insight (RAI), which automatically diagnoses network issues and identifies insights that allow faster problem solving

At the cell site, Samsung successfully completed a proof-of-concept (PoC) of AI-RAN in 2024, leveraging its virtualized RAN (vRAN) solution and a GPU card. From the PoC, Samsung verified the technical feasibility of Sounding Reference Signal (SRS) and Physical Uplink Shared Channel (PUSCH) estimation on AI-RAN. Our AI algorithm reconstructs channel signals from user equipment, often degraded by noise and interference, enabling more accurate channel estimation and leading to enhanced throughput and cell edge coverage. 

AI-on-RAN: Flourishing AI Services at the Network Edge

AI-on-RAN focuses on bringing AI applications closer to the edge by deploying them directly on the RAN. This can enable the development of new services or tap into edge computing to reduce latency and backhaul traffic.

Samsung’s Dr. Athul Prasad is serving as the Chair for the AI-on-RAN Working Group (WG3). Samsung has been actively contributing to AI-on-RAN by uncovering novel use cases that will enable innovation across consumer, enterprise, and government sectors. 

“As the Chair of Working Group 3, my focus is on guiding the industry toward the most impactful applications of AI, both in the current 5G era and as we prepare for the rise of 6G,” said Dr. Prasad. “We are exploring how AI can be effectively deployed across the RAN and brought closer to the network edge to deliver substantial benefits. By establishing clear interfaces and performance benchmarks, we are paving the path to the networks of the future that will revolutionize how operators enhance performance, optimize operations, and unlock new business opportunities.” 

Samsung is working with Tier 1 operators on AI-on-RAN using its vRAN. For example, AI-powered surveillance cameras and video monitoring applications can be deployed on the same commercial off-the-shelf (COTS) servers used for RAN functionalities. In construction or manufacturing, real-time monitoring enhances worker safety by detecting hazards instantly and ensures quality assurance through precise, automated production line inspections. This localized AI processing at the edge provides immediate insights, enables faster decision-making and tailored AI services across various industrial environments.

AI-and-RAN: Co-existence of AI and RAN Functionalities

The Alliance describes AI-and-RAN as the concurrent use of a combined computer-and-communications system to run both RAN and AI workloads to enhance platform utilization and create new monetization opportunities. 

An example of this is seen with base stations. They handle massive amounts of data traffic, but when traffic demand eases, computing power is often left underused or even unused. By leveraging these idle resources, operators can run AI workloads such as inferencing on the same infrastructure. With Samsung’s flexible vRAN, operators can maximize their competitive advantage at the edge, compared to a centralized architecture.

Another notable example of converging AI and vRAN infrastructure is Distributed AI. This not only includes orchestrating and scheduling different AI tasks across multiple virtualized Distributed Unit (vDU) servers but also splitting a Large Language Model (LLM) across vDUs to enable joint resource pooling for delivering high-quality AI services.

Samsung’s Broader Approach to AI-powered Networks

At Samsung, we are taking a broader approach to AI-RAN and AI-powered networks centered around an end-to-end software-based architecture, guided by two complementary perspectives: Network for AI and AI for Network. Our approach allows deeper exploration of how Network and AI drive mutual advancements, transforming the way networks operate and empowering operators to redefine their networks as AI-powered data centers.

• Network for AI refers to the optimal architectural foundation that best serves AI. We believe that AI needs to extend across the entire network, beyond the RAN. Therefore, Samsung’s end-to-end software-based network on vRAN COTS servers is the most suitable solution and the essential path to AI-native networks, driving innovation in 5G, 5G-Advanced, and 6G.

• AI for Network is about applying AI functions to enhance network performance and efficiency, shifting from reactive to proactive network management and automation. Samsung CognitiV NOS delivers full observability across network planning, installation, operation, and optimization by leveraging diverse AI-powered applications. It provides operators with comprehensive insights from massive amounts of network data while enabling closed-loop optimization and autonomous intelligence critical for network evolution—spanning mobile core, radio access, and transport elements.

Samsung’s Leadership in AI-powered Networks

Samsung’s leadership in AI-RAN is evident, as we have built the foundation for Network and AI to co-exist and create a powerful synergy, fueling new possibilities that will enrich our society. Samsung’s key strategy moving forward is “Network as a Data Center”. The inherent flexibility and distributed nature of this architecture will ensure a smooth evolution to AI-native networks.

This approach is all made possible by Samsung’s vRAN, which provides a flexible, multi-purpose platform for network evolution and lays the groundwork for AI-native networks. Proven in commercial deployments with global Tier 1 operators’ networks, Samsung vRAN has achieved significant real-world results, delivering enhanced network performance and efficiency with AI-powered capabilities. 

Samsung’s end-to-end software-based network beyond vRAN offers the easiest, most effective approach to an AI-powered network. As a multi-purpose platform, our solution allows operators to easily integrate Central Processing Unit (CPU) and Graphics Processing Unit (GPU) at any layer of the network—central, edge, and cell sites—enabling AI capabilities where needed. While operators with legacy networks need separate servers and GPU platforms that require additional investment, with Samsung’s vRAN, they can utilize the existing CPUs already equipped in vRAN COTS servers, or easily add GPUs for specific RAN or AI workloads.

Samsung’s approach to AI encompasses all network protocol layers—from Layer 1 (Physical Layer), Layer 2 (Data Link Layer) to Layer 3 (Network Layer)—while also extending across the core and even Service Management and Orchestration (SMO), boosting network performance and efficiency, as well as improving user experience. 

More specifically, AI-based channel estimation can increase user throughput in Layer 1, and AI-based link adaptation helps maximize user experience by finding the optimal modulation and coding scheme (MCS) levels for each user in Layer 2. For Layer 3, AI inactivity timer adaptation improves UE energy saving by optimizing terminal connection efficiency. In the core network, AI enables efficient use of resources by analyzing user mobility patterns and optimizing the paging operation.

The Path Forward

The evolution to software-based, AI-powered networks represents a strategic inflection point. Networks are becoming data centers and platforms for continuous innovation and Samsung's unique approach—pairing end-to-end software architecture with integrated AI capabilities—enables networks that adapt and optimize continuously.

As data traffic accelerates and AI applications proliferate, competitive advantage belongs to operators whose networks can learn and self-optimize. Samsung's “Network as a Data Center,” powered by AI, provides operators the practical tools to achieve this transformation.

Samsung recently published a technical white paper on its vision for the future of network evolution. To learn more about Samsung’s end-to-end software-based networks powered by AI, please refer to below.

WhitePaper: https://www.samsung.com/global/business/networks/insights/white-papers/0909-mobile-networks-evolution-in-the-ai-era/

Video: https://youtu.be/E453koE7E_U?si=1bNmaCOsETVJbAun