As mobile networks evolve toward 5G-Advanced and 6G, Artificial Intelligence–Radio Access Network (AI-RAN) platforms represent a fundamental shift in how radio access infrastructure is designed and operated, fully integrating AI into RAN hardware and software to deliver measurable gains in spectral efficiency, performance, and monetization.
Unlike conventional RAN architectures that rely on static configuration methods, AI-native RAN design employs data-driven, adaptive, and predictive control across physical, management, and operations layers, positioning AI as a core structural element rather than an add-on capability.
AI-native RAN doesn’t bolt intelligence onto existing infrastructure — it builds the network around intelligence from the ground up.
Nokia and NVIDIA have formalized a strategic partnership to bring this vision to market, producing what the companies describe as the industry’s first commercial AI-native RAN platform. T-Mobile U.S. has joined the partnership to integrate AI-RAN technologies into its 6G development, with field trials scheduled to begin in 2026.
Built on Nokia anyRAN software and the NVIDIA Aerial AI-RAN platform, the solution unifies AI and radio access workloads on software-defined, accelerated infrastructure. The architecture turns the radio network into a programmable AI-native platform where RAN functions and AI workloads run on a shared computing foundation, enabling communication service providers to improve performance, operational efficiency, and revenue opportunities simultaneously.
The platform’s compute infrastructure is cloud-centric and built on unified general-purpose hardware that supports concurrent cellular and AI workloads in containerized, multi-tenant environments.
A converged compute-communication fabric co-hosts baseband, MAC, and PHY protocol functions alongside diverse AI workloads on CPUs, GPUs, DPUs, and other accelerators. This shared hardware model supports both centralized and distributed intelligence across the RAN stack, forming the foundation for compute-enhanced wireless networks spanning 5G through 6G.
Performance results from trials and real-world deployments illustrate the platform’s practical impact.
Nokia’s AI-RAN platform has demonstrated more than 20% spectral efficiency gains in trials compared to conventional RAN architectures. The roadmap projects 50% spectral efficiency gains by 2027 and more than 100% gains by 2028, effectively doubling the capacity of existing spectrum assets.
Across real-world deployments, operators have reported roughly 30 to 40% performance improvements in throughput, latency, and reliability within dense mobile environments. These gains stem from AI-driven radio resource management, intelligent beamforming, MIMO optimization, interference mitigation, and energy-saving algorithms that continuously adapt to network conditions.
The improved spectral efficiency and reduced cost per bit allow operators to handle considerably higher traffic volumes in dense cells while maintaining or enhancing end-user experience.
Nokia’s long-term AI-RAN roadmap targets AI-native 6G using NVIDIA programmable merchant silicon platforms as a common foundation across 4G, 5G, 5G-Advanced, and future standards, ensuring investment continuity as network generations advance.
Together, the Nokia–NVIDIA platform signals a structural departure from legacy RAN design, replacing purpose-built, statically managed infrastructure with a programmable, intelligence-driven compute fabric capable of evolving alongside the demands of next-generation mobile networks.
