TodayToday's edition
NVIDIA added MGX 2U and MGX 4U Aerial RAN Computer servers for AI-RAN after T-Mobile's Bellevue call, but gave no power draw, price or shipping dates.
2 min readAI-assisted
NVIDIA's release artwork for its AI-RAN announcement; it is an illustration, not a photo of the Aerial RAN Computer servers or the Bellevue site.
Image: NVIDIA
Source · Press image from the NVIDIA newsroom, used for news reporting.
NVIDIA said on October 21 at GTC Berlin that its AI-RAN effort passed a milestone: T-Mobile completed the world's first AI-RAN call on a live commercial network in Bellevue, Washington, and that it is adding two new Aerial RAN Computer (ARC) server configurations, the MGX 2U and MGX 4U, for high-capacity radio sites.
According to NVIDIA's press release, the call ran on commercial ARC servers, Nokia's CUDA-accelerated RAN software and massive multiple-input multiple-output (MIMO) radios. It was connected to an existing 5G Standalone core and a commercial smartphone. AI-RAN runs radio processing and AI workloads on the same accelerated platform instead of on separate fixed-function radio hardware.
The hardware detail matters for power planners. The two new ARC configurations use NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs and will come with x86 or NVIDIA Vera CPU options. NVIDIA said they target dense, high-capacity workloads and are typically deployed at centralized RAN sites. Dell Technologies, Quanta Cloud Technology and Supermicro will offer ARC-ready servers. NVIDIA did not publish power draw, price or shipping dates in the release.
One live call so far; BT, Deutsche Telekom and Elisa trials and services are plans for 2027, per NVIDIA's release.
Illustration (AI-generated)
NVIDIA said nearly 20 telecom operators have publicly announced AI-RAN collaborations using Nokia anyRAN software and NVIDIA ARC. In Europe, the pilots are staged for 2027. BT plans live trials in Birmingham, England, in 2027. Deutsche Telekom has validated the Nokia and NVIDIA platform in labs in Bonn and Oulu, including AI-enhanced uplink channel estimation, and plans field trials in 2027. Elisa ran voice, mobile data and video analytics together on ARC over 5G spectrum in a European first, and said that clears the way for distributed deployment and commercial AI-RAN services in 2027. Orange is testing Nokia 5G software on ARC.
Nine operators across Europe, the Middle East and Africa are building or expanding NVIDIA-powered infrastructure, NVIDIA said. New deployments are at CRA in the Czech Republic, etisalat in the United Arab Emirates, Silknet in Georgia and TIM in Italy. Expansions are at Cassava, Deutsche Telekom, Fastweb and Vodafone, Kazakhtelecom and Orion Telecom.
Operators own land, grid connections and fibre at thousands of sites, and NVIDIA is pitching those sites as distributed inference capacity. For the supply chain, it opens a channel for RTX PRO 6000 Blackwell GPUs and Vera CPUs outside hyperscale halls. For grid and cooling teams, it means GPU servers drawing power at cell sites and metro hubs that were sized for radio equipment. That consequence is an inference from the announcement, not a figure NVIDIA stated.
NVIDIA also said Nokia and Cisco are building its Nemotron open models into network-operations software, and that other partners are using them for autonomous network agents. Those are software integrations and carry no hardware numbers.
The release is a vendor announcement. NVIDIA gave no site counts, no GPU counts, no power figures and no commercial rollout dates for the new configurations. The T-Mobile call is a single demonstration, not a network-wide deployment, and the European operators have so far announced pilots and trials rather than commercial service. Watch for ARC server pricing from Dell, QCT and Supermicro, and for the first 2027 field-trial results.
