Nokia just made a bold claim. On July 15, the company launched what it calls the industry’s first commercial AI-native RAN platform, built on its own anyRAN software and NVIDIA’s Aerial system. The pitch is simple: squeeze far more capacity out of spectrum operators already own, without a hardware overhaul.
What the Nokia AI-RAN Platform Actually Does
Spectral efficiency is the headline number. Nokia says the platform has already delivered gains above 20%, with a target of 50% by 2027 and more than 100% by 2028. Hit that last mark, and operators could roughly double the capacity of their existing spectrum.
Worth noting: those bigger figures are targets, not results. Nokia’s own roadmap puts pilots at the end of this year, with commercial availability arriving in 2027.
Operators won’t need to rip out hardware to get there. Nokia is offering the capability as a software subscription, with three ways to deploy it: a GPU plug-in card for existing AirScale sites, a standalone AI-RAN node, or a cloud-server build through partners.
Why This Matters for Nokia’s Business
This launch is really about fixing Nokia’s weakest division. Radio has been a headache for CEO Justin Hotard since he took over in 2025, and at last November’s capital markets day, he admitted the mobile business hadn’t delivered acceptable returns.
The NVIDIA partnership, backed by a $1 billion investment for roughly a 3% stake in Nokia, sits at the center of that turnaround. By leaning on NVIDIA’s silicon and CUDA software instead of custom chips, Nokia gets to redirect R&D spending toward software rather than hardware. Investors seem to like the story: Nokia shares have climbed through 2026 on AI and cloud momentum, and Omdia pegs the AI-RAN opportunity at more than $200 billion by 2030.
Is Nokia Really First?
Here’s where the claim gets complicated. Ericsson beat Nokia to market in June with a commercial AI-in-RAN subscription, one that runs on operators’ existing baseband silicon with no GPU required and is already live across more than 15 deployments.
Nokia’s “first” rests on a narrower distinction: a GPU-accelerated platform, which is a different architecture than AI layered on top of existing hardware. Both claims can be true at once. The bigger difference is dependency. Nokia’s CTO has acknowledged some of its Layer 1 software is tied to NVIDIA’s hardware, while Ericsson deliberately built its AI features to stay silicon-independent.
Conclusion: A Comeback Still in Progress
Nokia’s AI-RAN platform isn’t commercial yet, its biggest efficiency numbers are still two years out, and a rival already reached the market by a different route. Still, tying its radio business to NVIDIA gives Nokia a credible shot at fixing a unit it couldn’t repair alone, and the subscription model finally gives radio the recurring revenue hardware cycles never delivered. Whether that’s enough to call it a lead is the question worth watching through 2027.
