T-Mobile and Nvidia Complete First AI-RAN Call on Live Network

What Happened

T-Mobile and Nvidia announced they have completed the first telephone call carried over an AI-RAN (artificial intelligence radio access network) on a live commercial network. The test demonstrates that AI processing can be embedded directly into wireless radio infrastructure, rather than handled by separate systems, representing a functional step toward AI-native cellular networks.

Background

Radio access networks form the backbone of mobile connectivity, managing the link between cell towers and user devices. Traditional RAN architecture relies on fixed, rule-based software to allocate spectrum, manage interference, and route traffic. AI-RAN replaces or supplements those fixed rules with machine learning models capable of adapting in real time to network conditions.

T-Mobile is the largest wireless carrier in the United States by subscriber count, with approximately 120 million customers. Nvidia is the dominant supplier of graphics processing units used in AI workloads, with its H100 and subsequent chips underpinning the majority of large-scale AI training and inference deployments globally. The two companies announced a partnership focused on AI-RAN development in early 2024, with Nvidia committing its Grace Blackwell platform to the effort.

The AI-RAN Alliance, an industry group that includes Nokia, Ericsson, Samsung, and other major telecom equipment vendors, has been working to establish common standards for integrating AI compute into RAN infrastructure. T-Mobile is a member of that alliance.

What the Test Involved

The completed call ran over T-Mobile's live network using Nvidia's AI compute platform integrated into the RAN layer. According to the companies, the system used AI models to manage radio resources during the call in real time. The test was conducted on existing spectrum holdings rather than a closed laboratory environment, which the companies described as a requirement for validating commercial viability.

Nvidia's contribution to the deployment centers on its accelerated computing hardware, which provides the processing capacity needed to run AI inference at the speed required for radio network decisions, which must occur in milliseconds. T-Mobile contributed network infrastructure, spectrum access, and operational integration.

What It Means in Practice

AI-RAN technology is designed to improve spectral efficiency, meaning more data can be transmitted using the same licensed frequencies. It is also intended to reduce energy consumption per unit of data by optimizing transmit power and antenna configuration dynamically. Carriers have cited both factors as operational priorities as data traffic volumes continue to grow.

The technology also opens the possibility of running additional AI workloads, such as edge inference for enterprise customers, on the same physical hardware as the radio network. Nvidia has described this as a way for carriers to monetize their tower infrastructure beyond connectivity services.

The practical deployment of AI-RAN at scale would require carriers to replace or upgrade substantial amounts of existing baseband hardware, a capital-intensive process. No timeline for broad commercial rollout has been announced by either company.

Industry Context

Several other carriers and equipment vendors are pursuing parallel AI-RAN development. SoftBank in Japan has conducted AI-RAN trials in partnership with Nvidia. Ericsson and Nokia, the two largest RAN equipment vendors globally, have each announced AI integration roadmaps for their respective product lines. The competitive landscape for AI-native network infrastructure is still forming, with no dominant architecture established.

The completion of a live-network call gives T-Mobile and Nvidia a concrete demonstration point ahead of upcoming industry events, including Mobile World Congress and future spectrum auctions where network capability claims carry commercial weight.

T-Mobile has not announced a specific date for expanded AI-RAN trials or a commercial deployment schedule, and Nvidia has not disclosed pricing or availability details for the hardware configurations used in the test.