T-Mobile US and Ericsson Demonstrate AI-Enhanced Beamforming on 5G-Advanced Network

Image: Illustrative image · Wikimedia Commons · Public domain · Source

T-Mobile US and Ericsson are trialing AI-managed beamforming on 5G-Advanced, achieving up to 50% increase in advanced beamforming and improved network efficiency during peak traffic.

T-Mobile US, in collaboration with Ericsson, is testing an AI-driven approach to radio resource allocation on its 5G-Advanced network, focusing on enhancing beamforming techniques to better manage wireless traffic. The trial involves the use of a supervised-learning AI model trained on extensive T-Mobile network data, which runs on Ericsson’s RAN compute and silicon platforms. This model predicts rising demand in real time, allowing the network to dynamically redirect sounding reference signal (SRS) resources — brief transmissions from devices that enable the network to precisely measure and direct beams between the cell site and user devices.

Traditional resource allocation for SRS has been rigid and rule-based, but T-Mobile’s AI system adapts in real time, optimizing beamforming in response to current traffic conditions. Tests on newly activated sites in New York City showed significant performance improvements during peak usage periods. Key gains included approximately a 50% increase in advanced beamforming using full-band SRS, a 16% boost in downlink user throughput, a 12% increase in spectral efficiency, and a 10% rise in total downlink data volume.

These results build on earlier achievements by the partners, who in May reported around a 10% improvement in spectral efficiency thanks to an AI-native scheduler and enhanced link adaptation on live network traffic. T-Mobile Chief Network Officer Ankur Kapoor highlighted that the AI approach suits their 5G-Advanced network particularly well due to its reliance on Time Division Duplex (TDD) spectrum, which uses the same band for both uplink and downlink and where T-Mobile leads, especially in midband frequencies.

The partners emphasize that as mobile data traffic continues to grow—driven heavily by video content and 5G adoption reaching over 3.1 billion subscriptions worldwide—networks must evolve to be increasingly intelligent and adaptive. Ericsson’s Jaco du Plooy noted the trial as a key step towards self-optimizing networks empowered by advanced silicon technology.

Looking ahead, Kapoor envisions that 6G networks will require even greater capabilities to sense, learn, and react in real time to support emerging applications like distributed AI, robotics, and vision AI, with T-Mobile’s expanding nationwide 5G-Advanced network serving as a proving ground for these future technologies. This trial underscores the growing role of AI in optimizing network resource management and enhancing user experience under rising mobile data demand.