Japan telecom giant KDDI trialed an AI-powered radio access network (RAN) optimization platform from infrastructure partner Samsung on a live 5G standalone (SA) network that produced measurable gains in network download throughput.

The trial used Samsung’s RAN Speed Optimization (RSO) platform on KDDI’s network using 100 megahertz of time division duplex (TDD) spectrum in the 3.7 GHz band across 407 cell sites around Tokyo. The test showed a 31% increase in 5G downlink speeds during “peak hours,” and maximum gains of up to 52% in downlink speeds in “dense urban areas.”

Samsung explained that its RSO platform, which is part of the vendor’s AI-infused CognitiV Network Operations Suite, uses an AI-based prediction model to automatically analyze cell site data and recommend optimized parameters to that specific location. This can help optimize cell site performance and resource usage, tenants that have been associated with broader AI-RAN developments.

ABI Research noted in a report that AI-RAN will generate more than $6.1 billion in investment revenues by 2032, though most of that is not expected to begin until 2029. However, the analyst firm also highlighted the dearth of operators among AI-RAN Alliance members, which “signals some industry skepticism regarding the near-term value of AI-RAN.”

“The real growth in AI-RAN will only come when performance benchmarks are validated in the field,” ABI Research analyst Samuel Bowling wrote. “Operators need evidence that AI-RAN can deliver technical and financial outcomes at scale, with a justifiable cost model. Without that, adoption will continue to lag behind vendor enthusiasm.”

AI-RAN running and collaborations

KDDI’s results with Samsung echo those of other operators.

T-Mobile US earlier this year noted that AI-RAN work with Ericsson resulted in incremental network efficiency gains. That work was through the integration of AI-controlled RAN services designed to improve 5G network efficiency into commercial trials across the carrier’s 5G-Advanced network.

Specifically, the pair moved AI-native scheduler and link adaption into “large-scale commercial trials” on live 5G-Advanced network traffic. That software taps neural networks running on Ericsson “optimized” hardware that can predict changing radio conditions in real time, with the trial showing a 10% improvement in spectral efficiency and 15% increase in downlink speeds compared with “legacy rule-based methods.”

The latest trial also continued long-standing work between KDDI and Samsung, including the vendor having been tapped to provide KDDI with the 5G SA core and work on deploying open RAN systems.

KDDI just last week also swept Samsung into work with Nvidia and Keysight Technologies to develop a high-fidelity RAN digital twin aimed at supporting AI-driven optimization and autonomous network operations in the 6G era.

That project is designed to create a virtual replica of real-world RAN conditions, allowing operators and vendors to train, test, and validate AI models without disrupting live commercial networks. KDDI said the platform will support large-scale simulation of changing radio environments, traffic patterns, and network scenarios, giving AI systems a safer and faster environment for development than direct experimentation on production infrastructure.