The telecommunications industry is undergoing a transformative shift propelled by AI, as networks evolve to meet growing demands for agility, efficiency, and innovation. Operators now confront pivotal decisions: Should they build separate AI infrastructure? Wait for future hardware breakthroughs? Rip out existing deployments and start fresh?

The answer might be simpler than expected - use what's already being deployed. Software-driven networks are proving to be the ideal architecture for AI integration. The operators who virtualized early aren't just saving on operating costs - they're sitting on infrastructure that can integrate AI capabilities more easily and run AI workloads today without fundamental redesign.

Automation now integrates into this framework, harnessing observability and comprehensive insights gathered across the network to feed intelligence back into the system, turning networks into proactive, self-optimizing assets ready for the future.

Advancing the Software-Driven Foundation

Traditional telecom infrastructure kept network functions tightly coupled to dedicated hardware. A router was a physical box. Baseband processing happened on purpose-built equipment. Security appliances were separate solutions and devices. While this hardware-based network provided reliable services and performance for decades, this meant every new capability required more hardware, more space, and more power consumption.

Virtualization flips this model by decoupling network functions from purpose-built hardware and running them as software on a COTS server, which is a general-purpose compute platform. This shift not only brings cost savings but also delivers unprecedented flexibility and AI readiness, enabling the flexible utilization of compute capabilities on COTS servers.

The progression happened in stages. Core network virtualization came first, letting operators run packet gateways and subscriber databases on standard servers instead of purpose-built appliances. Then came vRAN, which separated baseband processing from radio hardware and ran it as cloud-native software.

Where AI Actually Fits

More importantly, an end-to-end software-driven network serves as the ideal foundation for AI, based on its inherent data visibility and dynamic control. Compared to traditional networks where data is trapped in hardware-based silos, fully virtualized, software-based networks serve as a unified data source for AI, enabling seamless network orchestration to enhance performance and optimize infrastructure.

This infrastructure also provides the agility and flexibility to adopt innovative AI-driven services, creating new monetization opportunities that were previously impossible with a hardware-centric architecture. The fast-paced innovation with the latest processors does more than just power networks. It frees up core resources to run AI applications and services directly on existing telco infrastructure. By seamlessly integrating either CPUs or GPUs into software-driven networks, operators can flexibly expand and enhance AI capabilities.

Intelligent Automation: The Gateway to Autonomous Networks

To maximize the efficiency of software-driven networks and AI capabilities, Samsung provides CognitiV Network Operations Suite (NOS) designed for seamless end-to-end network automation and management. With this, operators can be equipped with a holistic view of the entire network leveraging a truly unified management framework that orchestrates all network elements.

It provides various technologies such as AI agents, digital twins, Large Language Models (LLMs), Retrieval Augmented Generation (RAG), and knowledge graphs under Samsung’s AI Agent Fabric, enabling operators to effortlessly manage their virtualized assets while paving the way for fully autonomous networks.

To achieve auto-pilot networks, Samsung introduced a new, unified knowledge and decision intelligence layer, Samsung’s AI Agent Fabric, which aggregates the capabilities of the AI agents and advanced Agentic AI technologies into a centralized hub. This Agent Fabric maintains the telco knowledge required to operate the network and also provides AI Agents that plan, evaluate, and act on the network based on their own reasoning.

Operators can also preemptively assess the impact of deploying new functions or modifying network configurations across the entire infrastructure by mapping both physical and logical topologies through a digital twin and conducting AI-based what-if analyses.

The evolution from 5G to 5G-Advanced and 6G hinges on three interconnected pillars: virtualization for flexible networks, AI integration across all network layers, and automation towards autonomous networks.

Samsung is building a unified, software-centric foundation where flexibility, AI, and automation converge to redefine what networks can achieve.