Nokia and NVIDIA’s Commercial AI-Native RAN: Tracing the Path from Disaggregated Open RAN to GPU-Powered Intelligence

The Long Road to Intelligent Radio Access

The telecom industry’s journey toward AI-native radio access networks did not begin with GPUs or large language models. It started with the push for openness and disaggregation in the RAN, a foundational shift that laid the groundwork for embedding intelligence at the edge.

From Proprietary Silos to Open Interfaces

For decades, RAN equipment remained tightly integrated and vendor-specific. The emergence of the O-RAN Alliance and Open RAN principles introduced standardized interfaces, enabling multi-vendor deployments and separating hardware from software. This disaggregation created the modular architecture necessary for later AI integration, allowing operators to mix components while maintaining interoperability.

Early Experiments with Intelligence: RIC and Automation

As Open RAN matured, the RAN Intelligent Controller (RIC) became a key vehicle for injecting intelligence. Non-real-time and near-real-time RICs enabled policy-driven optimization and basic machine learning applications for tasks like traffic steering and energy management. These efforts demonstrated the value of data-driven decisions but were limited by traditional CPU-based processing and lacked the computational scale for truly transformative AI.

The GPU Inflection Point

The introduction of GPU-accelerated baseband processing marked a pivotal evolution. NVIDIA’s Aerial platform brought CUDA and AI capabilities directly into the RAN stack, transforming base stations from dedicated signal processors into programmable computing nodes. This shift aligned perfectly with Open RAN’s emphasis on flexibility, enabling operators to run AI workloads alongside traditional RAN functions on shared infrastructure.

Nokia’s July 2025 Announcement in Context

On July 15, 2026, Nokia announced the industry’s first commercial AI-native RAN platform. Built on its AI-native anyRAN software and integrated with NVIDIA’s Aerial AI-RAN capabilities, the platform embeds AI computing into the baseband system. It supports Open RAN standards and multi-vendor environments, with a subscription model for ongoing AI enhancements without hardware refreshes.

This launch represents the culmination of years of incremental progress: from interface standardization to RIC experimentation, and now to full commercial AI-native operation at planetary scale.

Implications for Operators and the Broader Ecosystem

For operators, the platform offers a pathway to continuous performance improvements through software. It positions Open RAN not merely as a cost-optimization play but as the enabler of AI-driven spectral efficiency, energy savings, and new service capabilities. The multi-vendor compatibility ensures that early adopters of disaggregated architectures can evolve without rip-and-replace strategies.

Looking Ahead to 6G and Beyond

As the industry eyes 6G, the fusion of Open RAN openness with AI-native processing sets the stage for networks that are not just automated but inherently intelligent. Nokia and NVIDIA’s collaboration underscores how foundational Open RAN work has enabled this next chapter.

Sources

• https://www.semicone.com/article-490.html
• https://www.fidelity.com/news/article/technology/202607150300PRIMZONEFULLFEED1001212822