Nokia Makes First AI-native Platform Available

Nokia has introduced a commercial AI-RAN platform. This solution is designed to integrate artificial intelligence functions directly into the radio access network. The solution combines the Finnish group's anyRAN software with NVIDIA's Aerial AI-RAN platform and aims to shift part of the evolution of mobile networks from hardware renewal to software upgrades.

According to Nokia, the platform has already achieved spectral efficiency improvements exceeding 20%. The roadmap aims to reach 50% by 2027 and exceed 100% by 2028. The goal is to increase the amount of traffic that can be handled using the already installed radio spectrum and infrastructure, reducing the cost per bit.

Initial trials are expected by the end of 2026, while commercial availability is slated for 2027.

Three Hardware Options for Adopting AI-RAN

The platform is based on a common software-defined architecture and supports 4G and 5G networks, along with future evolutions towards 6G. Nokia has planned three implementation modes, designed to adapt to the different existing infrastructures at operators.

1. A plug-in unit equipped with GPU for existing AirScale installations. This approach allows the addition of accelerated computing capability to already operational base stations without replacing the entire infrastructure. The solution uses NVIDIA technologies and also includes support for AI-accelerated merchant silicon components developed by Marvell.

2. An autonomous GPU-based AI-RAN node. The system can be used individually, in cluster configurations, or alongside AirScale installations, forming a single logical base station.

3. Cloud-native architectures utilizing Cots servers equipped with GPUs, provided through Nokia's ecosystem partners. This configuration is aimed at operators intending to run RAN functions on standard computing infrastructures.

Open RAN Compatibility and Multi-vendor Networks

Nokia states that the anyRAN software will support three new baseband platforms based on accelerated computing. The existing portfolio will also be made compliant with O-RAN standards, allowing the platform to be used with compatible radio units manufactured by other vendors.

The goal is to enable operators to introduce AI-RAN gradually, without requiring an immediate replacement of the installed base. The same software platform can then be used across different hardware configurations and in various cloud environments.

Updates Through Software Subscription

The platform will also be offered through a subscription-based commercial model. Operators will be able to receive new artificial intelligence algorithms, network optimization features, and spectral efficiency updates without waiting for traditional hardware renewal cycles.

The strategy aims to transform RAN into a more programmable infrastructure, where part of the capacity increases and new functions are introduced via software. The start of trials will serve to verify how replicable the results obtained in tests can be in the commercial networks of operators.

"AI-RAN is the most important innovation in radio networks in decades. It makes the network smart, extends artificial intelligence to the physical world, and allows telecommunications operators to get more from existing infrastructures, including a software upgrade path to 6G. Nokia's anyRAN software, based on NVIDIA's Aerial AI-RAN platform, allows greater performance from the already available spectrum and can be used with existing Nokia radio units or with devices compliant with O-RAN standards. For operators, this means better performance, higher returns on investment, and a faster introduction of new services," says Justin Hotard, President and CEO of Nokia.

"Telecommunications are entering the era of artificial intelligence: the radio access network is the next infrastructure for AI. Together with Nokia, we are bringing NVIDIA CUDA and artificial intelligence to the baseband, transforming RAN into a global-scale AI computer. For operators, this represents a generational shift, enabling greater capacity and efficiency from the current spectrum while simultaneously laying the foundations for new AI-based services and the era of 6G," comments Jensen Huang, founder and CEO of NVIDIA.

"The launch of Nokia's AI-RAN marks an important step in turning AI-RAN from an industrial vision into a commercial reality. The introduction of the new AI-RAN node, along with the AirScale plug-in unit for capacity enhancement and cloud-native implementation options, offers operators tangible opportunities to adopt AI-native networks based on existing infrastructures and their transformation goals. By combining accelerated AI computing with a software-defined architecture and a clear product roadmap, Nokia helps operators increase capacity, improve network economics, and accelerate the transition to an AI-native RAN," explains Rémy Pascal, Practice Leader, Mobile Infrastructure at Omdia.