T-Mobile has completed the world’s first AI-RAN call on a live commercial network, NVIDIA announced at GTC Berlin on October 21, 2026 — a working demonstration that AI compute can run inside the mobile network itself, not just in data centers behind it. Why it matters: if radio networks double as distributed AI infrastructure, the carriers’ towers, power, and fiber become part of the AI buildout — and a new revenue line for operators.
Berlin, October 21, 2026
What happened in Bellevue
The call took place in Bellevue, Washington, on T-Mobile’s live network. It ran on commercial NVIDIA Aerial RAN Computer (ARC) servers, Nokia’s NVIDIA CUDA-accelerated RAN software, and massive multiple-input multiple-output radios, connecting to an existing 5G Standalone core and a commercial smartphone. NVIDIA says nearly 20 telecom operators globally have publicly announced AI-RAN collaborations using Nokia’s anyRAN software on ARC, with Nokia bringing trials to BT, Deutsche Telekom, Elisa, and Orange.
Europe is trialing it too
Elisa has completed Europe’s first AI-RAN field trial, running voice, mobile data, and vision AI together on a single accelerated computing platform over 5G spectrum — a result NVIDIA says clears the way for commercial AI-RAN services in 2027. Deutsche Telekom has validated the Nokia-NVIDIA platform in laboratories in Bonn and Oulu, including AI-enhanced uplink channel estimation, and plans field trials in 2027. BT is preparing live AI-RAN trials in Birmingham, England, in 2027, and Orange is testing spectral-efficiency improvements and edge services on live networks. Separately, nine operators across Europe, the Middle East, and Africa are building or expanding NVIDIA-powered infrastructure across centralized and distributed sites.
New hardware and network agents
NVIDIA introduced two new ARC configurations, MGX 2U and MGX 4U, powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs with x86 and NVIDIA Vera CPU options; Dell Technologies, Quanta Cloud Technology, and Supermicro will offer ARC-ready servers. On the software side, Nokia is integrating NVIDIA Nemotron open models into its Autonomous Networks Suite to build agents for network operations, and Cisco has fine-tuned Nemotron models inside Crosswork AI for IP and optical transport networks. “Telecom networks were built to carry the world’s conversations. Now, they will also carry the world’s intelligence,” said NVIDIA founder and CEO Jensen Huang.
Why it matters
AI-RAN is a bet that the scarcest AI resources — powered, connected sites close to users — are things carriers already own. A first live call does not prove the economics, and most European activity is still trials pointing at 2027. But the direction is clear: the same GPU platform NVIDIA sells to data centers is now being pitched for the radio network, and the operator list is long enough that this is a real product line, not a demo. Everything AI, tested — the thing to watch is whether any operator announces paying AI services running on ARC, not just trials.
FAQ
What is AI-RAN?
AI-RAN extends accelerated computing into the radio access network, so the same platform that runs 5G radio software can also run AI workloads — serving models and applications from sites close to users.
Who made the first AI-RAN call, and where?
T-Mobile completed the call in Bellevue, Washington, on its live commercial network, using NVIDIA Aerial RAN Computer servers and Nokia’s CUDA-accelerated RAN software, per NVIDIA’s announcement.
Which other operators are involved?
Nokia is bringing anyRAN-on-ARC trials to BT, Deutsche Telekom, Elisa, and Orange, and NVIDIA says nearly 20 operators globally have announced AI-RAN collaborations; nine EMEA operators are building or expanding NVIDIA-powered infrastructure.
When could AI-RAN services launch commercially?
NVIDIA points to 2027: Elisa’s field trial is described as clearing the way for commercial services that year, and BT and Deutsche Telekom both plan live or field trials in 2027.
Sources: NVIDIA (GlobeNewswire via Morningstar), Unite.AI