T-Mobile and Ericsson have reported performance gains from trials of an AI-native radio access network feature running on T-Mobile’s 5G Advanced network.
The companies said the feature has moved into large-scale commercial trials using live 5G Advanced network traffic. Compared with legacy rule-based methods, Ericsson’s AI-native Scheduler with Link Adaptation delivered close to a 10% improvement in spectrum efficiency and up to a 15% increase in downlink throughput, in line with earlier tests conducted in smaller geographies.
Trials began early in the second quarter of 2025, expanding to several US markets, including Los Angeles, New York, New Jersey, and Salt Lake City. The operator is targeting commercial deployment in the third quarter of 2026.
Scheduler and link adaptation
The tested software uses AI in the RAN scheduler and link adaptation process, using a neural network running on Ericsson hardware to predict changing radio conditions. The software is designed to make scheduling and link adaptation decisions as radio conditions change. That includes environments with high traffic demand or weaker radio frequency conditions.
In network operations, the scheduler manages how radio resources are assigned to users. Link adaptation adjusts transmission settings based on signal quality and interference.
Ericsson said the feature is intended to support more stable performance for services like streaming and video calls during periods of heavy network use. The software applies AI directly inside the RAN software in addition to running in external network management tools.
Johan Hultell, head of Product Line RAN Software for Business Area Networks at Ericsson, said AI is central to the company’s approach to programmable networks.
Nvidia infrastructure work
The work forms part of T-Mobile’s broader AI-RAN activity with Ericsson, Nokia, and Nvidia. In September 2024, T-Mobile announced the AI-RAN Innovation Centre at its headquarters in Bellevue, Washington, where it is working with the three companies.
Ericsson has also demonstrated its Cloud RAN software on NVIDIA AI infrastructure with T-Mobile at the Bellevue centre. The company said the software is designed to run in commercial off-the-shelf hardware and NVIDIA AI infrastructure. It can also run on purpose-built hardware using Ericsson Silicon.
When deployed on commercial hardware, the software can use accelerator hardware and software for compute-intensive RAN functions where available.
Ericsson said the demonstration with T-Mobile used NVIDIA Aerial CUDA. The work is separate from the scheduler trial and relates to Cloud RAN software portability.
T-Mobile is using the Bellevue centre to test AI-RAN and Cloud RAN deployments with Ericsson, Nokia, and Nvidia. T-Mobile, Ericsson, Nokia, and Nvidia are also founding members of the AI-RAN Alliance, which was launched at Mobile World Congress Barcelona in February 2024. The group was formed to examine how AI can be integrated into cellular networks, including RAN and mobile edge systems.
Ericsson and T-Mobile said they will continue working together on other areas where AI can be applied to network performance and efficiency.
See also: Optus and Ericsson hit 3.4Gbps on a live 5G standalone network
Want to discover how IoT is transforming telecoms and connectivity? Join the IoT Tech Expo in Amsterdam, California, and London. Explore how innovations in 5G, edge computing, and IoT are shaping the future of networks and services. The event is part of TechEx and co-located with other leading technology conferences, click here for more information.
Telecoms Tech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
