Nokia Claims First Live AI-RAN Call With T-Mobile

On Tuesday T-Mobile put numbers on an AI trial with Ericsson. A day later its other radio supplier, Nokia, announced a milestone of its own with no numbers attached. Nokia says that with T-Mobile and NVIDIA it has completed “the world’s first AI-RAN call”, made on T-Mobile’s live production network at a commercial cell site in Bellevue, Washington, according to a blog post by Nokia’s chief technology officer, Pallavi Mahajan.

AI-RAN means running the radio access network, the part of a mobile network that links phones to masts, as software on the kind of GPU computing used for AI, so the same hardware can also handle AI work. The call shows Nokia’s NVIDIA-based approach working outside the lab. It doesn’t show what it does for speed, capacity or cost: none of the three companies released any performance figures.

The call ran on commercial kit at a working site

Nokia says the call used its Massive MIMO radio and 5G anyRAN software, running on servers built on NVIDIA’s Aerial RAN Computer platform, and that the site was connected to a commercial 5G core and commercial devices. NVIDIA’s account adds that the core was an existing 5G Standalone core and that the device was a commercial smartphone.

Mahajan goes further than the “world’s first” label. She calls it “the world’s first software-defined RAN call using accelerated computing” and “a moment from which there is no looking back”. Light Reading noted that “few details were provided” beyond the site and the equipment. It also recalled that T-Mobile’s chief technology officer, John Saw, said at Mobile World Congress in March that he would be “really upset” if he did not have Nokia’s AI-RAN system to test in the field within the year.

The call is the latest step in a short history. T-Mobile opened an AI-RAN Innovation Center in Bellevue in September 2024. Nokia and NVIDIA formed their partnership in 2025, and earlier this year Nokia reported lab tests of anyRAN on NVIDIA’s platform with T-Mobile and other operators, the blog says.

Nokia’s ‘2x spectral efficiency’ is an ambition, not a result

The headline promise is Nokia’s own. “With AI-RAN, we are on a path to 2x spectral efficiency,” Mahajan writes, describing it as “an ambition”: more capacity from the same radio spectrum. Nothing published on Wednesday measures progress towards it.

Ericsson’s figures make the contrast plain. In the New York trial TSN reported on Tuesday, T-Mobile said Ericsson’s AI-managed beamforming lifted peak-hour download speeds by 16%. Light Reading framed the two announcements as a split in strategy: Nokia is betting on NVIDIA GPUs, while Ericsson runs AI on its own silicon. T-Mobile is testing both.

Nokia says the next field trials will try more spectrum bands and test whether the platform can run AI inference workloads alongside the radio functions. That second test matters most to Nokia’s case, which is that shared hardware could bring operators “new monetization opportunities from AI services and return on investment”. Nokia gives no date.

BT plans British trials in Birmingham next year

The approach is also coming to the UK. NVIDIA says BT is planning live AI-RAN trials in Birmingham with Nokia and NVIDIA in 2027. Deutsche Telekom has validated the platform in labs in Bonn and Oulu and plans field trials in 2027. Finland’s Elisa says it has run voice, mobile data and video analytics together on the platform in a field trial. NVIDIA counts nearly 20 operators that have publicly announced AI-RAN work with Nokia’s software and its hardware.

For now the Bellevue call proves the system can carry live traffic. The first published throughput or spectral-efficiency figures from a live network will show whether Nokia’s 2x ambition is on track.