AI-native network architecture: pervasive AI at every layer

From AI-Assisted to AI-Native

5G-Advanced introduced AI/ML as an optimization tool for specific functions (beam management, CSI feedback). 6G takes a fundamentally different approach — AI is not an add-on but a core architectural principle. An AI-native network embeds machine learning at every layer: AI-based waveform generation at the physical layer, reinforcement learning for MAC scheduling, graph neural networks for routing optimization, and large language models for intent-based network management. The network continuously learns from its environment, self-optimizes, and autonomously adapts to changing conditions…