NVIDIA, Nokia and a group of telecom operators in Europe and the United States have unveiled advances in bringing accelerated compute to their networks, from central data centers down to base stations, with the goal of running AI models and applications closer to users. The announcement, made on October 21, 2026 at GTC Berlin, includes what NVIDIA describes as the first AI-RAN call on a commercial network, completed by T-Mobile, and two new configurations of its radio access network server.
AI-RAN with NVIDIA, Nokia and operators in 30 seconds
- T-Mobile completed an AI-RAN call on a commercial network in Bellevue, Washington, with NVIDIA and Nokia.
- BT, Deutsche Telekom and Elisa are preparing trials or commercial deployments for 2027, and Orange is testing Nokia’s 5G software on ARC.
- NVIDIA is introducing ARC in MGX 2U and MGX 4U form factors with RTX PRO 6000 Blackwell GPUs, which Dell, QCT and Supermicro will sell.
- Nokia, Cisco, TCS and NTT DATA are using Nemotron and Agent Toolkit to build agents that manage networks.
AI-RAN is the name given to radio access networks (RAN, the part of the mobile network that connects antennas to phones) that run on accelerated hardware capable of also handling AI workloads. NVIDIA’s piece is Aerial RAN Computer (ARC), and Nokia’s is its CUDA-accelerated anyRAN software. According to NVIDIA, almost 20 operators worldwide have publicly announced AI-RAN collaborations using that combination.
Jensen Huang, founder and CEO of NVIDIA, argued in the announcement that telecom networks, built to carry the world’s conversations, will also come to carry intelligence. He added that operators are turning their land, power and connectivity into distributed AI infrastructure.
From the first call in Bellevue to European trials in 2027
T-Mobile’s call used commercial ARC servers, Nokia’s CUDA-accelerated RAN software and massive MIMO radios, connected to an existing 5G Standalone core and a commercial phone. In Europe, the timelines are longer:
- BT plans live AI-RAN trials in Birmingham, England, with Nokia and NVIDIA in 2027. It is also using Nemotron to build agents that review photos from field installations and flag problems, aiming to avoid repeat site visits.
- Deutsche Telekom has validated the Nokia/NVIDIA AI-RAN platform in its labs in Bonn and Oulu, including an AI-enhanced uplink channel estimation, and plans field trials in 2027. It will also expand its Industrial AI Cloud, which hosts open models such as Nemotron and Soofi and serves robotics companies like Agile Robots.
- Elisa ran voice, mobile data and video analytics simultaneously on ARC in a field environment with 5G spectrum, a trial NVIDIA describes as the first of its kind in Europe. It expects distributed deployments and commercial AI-RAN services in 2027.
- Orange offers open models such as Nemotron as a managed service on its Live Intelligence platform, is deploying ARC for real-time multilingual translation aimed at emergency and technical-support teams, and is testing Nokia’s 5G software on ARC.
NVIDIA also cites operators in Europe, the Middle East and Africa building new infrastructure with its technology, such as CRA (Czech Republic), etisalat (United Arab Emirates), Silknet (Georgia) and TIM (Italy), as well as others expanding it, such as Cassava (Africa), Fastweb + Vodafone (Italy), Kazakhtelecom (Central Asia) and Orion Telecom (Serbia). The announcement doesn’t give details on those projects or GPU counts or site numbers.
New ARC servers and agents for running the network
The two new configurations, ARC MGX 2U and MGX 4U, are designed for dense radio and AI workloads, typically at centralized RAN sites. They use NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, with an x86 or NVIDIA Vera processor, and Dell Technologies, Quanta Cloud Technology (QCT) and Supermicro will offer ARC-ready servers. NVIDIA hasn’t given availability dates.
The other part of the announcement focuses on running the network with AI agents. Nokia is integrating Nemotron into its Autonomous Networks Suite and using NVIDIA Agent Toolkit to build agents that manage fixed, radio, core and transport networks. Cisco has fine-tuned Nemotron and integrated it into Crosswork AI, a shared agent layer for its Crosswork Network Automation software across IP and optical transport. TCS is building agents to validate radio site installations, detect and fix issues, and offer multilingual voice services, and NTT DATA is expanding its NextGenOSS platform with agents that detect network degradation, point to the likely cause and recommend what to do. Noticias.ai has covered the expansion of NVIDIA Agent Toolkit unveiled at the same conference, and cloudnews.tech has previously covered NVIDIA and Nokia’s alliance to bring AI-RAN from the lab to commercial rollout.
Frequently asked questions
What is AI-RAN?
It’s a radio access network that runs on accelerated hardware capable of handling both mobile network functions and artificial intelligence workloads at the same time.
Which operator made the first AI-RAN call on a commercial network?
T-Mobile, in Bellevue, Washington, using NVIDIA ARC servers, Nokia software and a commercial phone, according to NVIDIA.
Which European operators are testing AI-RAN with NVIDIA and Nokia?
BT, Deutsche Telekom and Elisa have trials or deployments planned for 2027, and Orange is testing Nokia’s 5G software on ARC.
Who will manufacture the ARC MGX servers?
Dell Technologies, Quanta Cloud Technology and Supermicro will offer ARC-ready servers, though NVIDIA hasn’t given dates.
Sources:
