AI-RAN

Nokia AI-RAN is the industry’s first AI-native RAN platform. It is built on accelerated computing and the anyRAN software foundation, providing a software-defined, programmable platform across three deployment paths with a shared software roadmap to 6G.

The difference between RAN and AI-RAN is architectural. Traditional RAN architectures were built for mobile broadband connectivity and largely human-driven traffic such as voice, video and data. AI-RAN combines a software-defined platform with accelerated computing integrated into the RAN.

A software-defined RAN decouples software updates from hardware upgrade cycles. AI-RAN is software-defined and programmable. It allows the network to improve continuously through software and supports the evolution to AI-native 6G.

Accelerated AI-RAN node is a high-capacity solution suitable for any radio access network, including standalone deployments. It combines advanced algorithms and AI-accelerated computing, supporting evolution from 4G and 5G to 6G. Nokia has developed the Accelerated AI-RAN node in collaboration with industry partners like NVIDIA.

Cloud AI-RAN is an AI-accelerated RAN built on GPU-powered COTS servers. It is suitable for both distributed and centralized deployments. Cloud-native AI-RAN enables AI Grid rollouts and sharing of accelerated computing infrastructure across AI workloads and RAN. Nokia has developed the cloud-native AI-RAN in collaboration with an ecosystem of industry partners.

AI inference must become increasingly distributed to the RAN edge to meet the growing demand for AI inference. While AI model training remains largely centralized in data centers, the capacity demands for AI inference workloads will increasingly land at the network edge, or in the distributed AI-RAN nodes.

Telecommunication providers can monetize AI-RAN by transforming their existing, widely distributed RAN infrastructure into an AI grid that supports AI inference at the edge. As physical AI systems scale, telecommunication providers can create new revenue opportunities by delivering deterministic connectivity and edge AI capabilities.

Typical AI workloads are generative AI, agentic AI and physical AI.

• Generative AI (GenAI) places heavy demand on uplink and requires robust interconnect performance.
• Agentic AI generates bursty, transactional traffic and requires highly responsive communications.
• Physical AI requires ultra-low latency, controlled jitter and deterministic connectivity to support real-time interaction with the physical world.

The AI-RAN Alliance is a global consortium that aims to transform networks through AI innovation and collaboration. It advances the integration of AI into radio access networks and fosters the development of AI-native infrastructures, focusing on network performance, energy efficiency and monetization opportunities for telecommunication providers. Nokia is a founding member of the AI-RAN Alliance.