Nokia Unveiled AI-Native RAN Platform

Nokia announced it expects significant hikes in spectral efficiency over the next two years as it launched an AI-native RAN platform. The solution relies heavily on Nvidia systems.

The AI-RAN platform is already capable of boosting spectral efficiency by 20%, and Nokia stated gains of 50% and 100% would be delivered in 2027 and 2028, respectively. It hailed the platform as one of the most significant shifts in radio network architecture in decades. It hopes to deliver pilot deployments by the end of the year and commercial availability in 2027.

The platform uses Nokia’s anyRAN software so operators running other network vendors’ equipment today can use it; however, the bulk of the services available through Nvidia’s systems are reserved for those already using the Finnish provider’s AirScale range. A GPU-based AirScale capacity plug-in is announced as the most efficient path forward. It brings Nvidia computing capabilities into current networks, preserving existing investments and offering the option to employ chips from Marvell Technology.

Nokia also launched a standalone AI-RAN node employing GPUs which can run on 4G and 5G networks, with forward compatibility to likely 6G infrastructure. Various configurations are available comprising a one-node deployment, clusters and as part of a single logical base station. GPUs also form the heart of a commercial off-the-shelf server available from Nokia’s partners, which it stated would tackle any supply chain worries.

Nokia president and CEO Justin Hotard said that AI-RAN makes the network intelligent and would help operators maximize the capabilities of their current infrastructure. Jensen Huang, founder and CEO of Nvidia, hailed the step as the dawn of a new era for the telecoms sector as the companies turn RAN into a planet-scale AI computer.