KDDI 5G SA network throughput increases 52% from Samsung AI

KDDI has shown that Samsung AI RAN Speed Optimizer boosts 5G SA (Standalone) network performance by up to 52 percent during peak traffic.

Live commercial network tracking in Japan confirms that applying automated intelligence to individual cellular base stations yields measurable capacity gains over traditional cluster-wide configurations. Samsung and mobile network operator KDDI completed a multi-month field trial of the AI-powered RAN Speed Optimizer across Tokyo, demonstrating a new methodology for radio access network coordination.

Kazuhiro Furuhata, Chief Network Officer at KDDI, said: “Combining KDDI’s accumulated expertise in network innovation and Samsung’s technical leadership, this field trial proves that individual tuning for cells — a long-standing industry challenge — has now become a reality through the integration of AI.

“Moving forward, we remain committed to pushing the boundaries of AI-based technologies to continuously elevate network experience for our customers.”

The Tokyo deployment used a software framework that evaluates the distinct physical environment, interference patterns, and localised traffic demands of each independent cell. By executing continuous automated adjustments at the individual site level, the system addresses the specific propagation variables that macro-level cluster profiles fail to capture.

The field architecture spanned dense urban financial districts, residential suburbs, and peripheral rural zones to test the software across divergent RF environments. Operating on 100 MHz of mid-band 3.7 GHz Time Division Duplex spectrum, the trial platform collected continuous telemetry from hundreds of active cells. This operational data provided the foundation for training the predictive models against real-world congestion patterns and physical obstructions.

Peak load throughput metrics and spectral efficiency

Data gathered during peak utilisation hours indicated an average increase of 31 percent in 5G Standalone downlink throughput across the entire geographic trial footprint. In high-density urban sectors characterised by severe signal attenuation and complex multipath reflections, downlink performance improved by a maximum of 52 percent.

The software operates as a core application within the Samsung CognitiV Network Operations Suite, an engineering platform designed to consolidate automated infrastructure applications and software agents. The integration relies on an internal predictive engine that ingests live site environment data, computes the optimal radio frequency configuration parameters for that specific coordinate, and provisions the updates without manual script editing or field technician intervention.

This operational framework targets the overhead costs associated with traditional radio frequency optimisation cycles, which typically require iterative drive testing and manual cluster tuning. By transitionary shifting to automated self-optimisation, the software matches power levels, beamforming angles, and scheduling priorities to immediate localised demand, mitigating the risk of packet drops during sudden traffic spikes.

Foundations for next-gen RAN architecture

KDDI has historically deployed fully virtualised RAN architectures, which decouple the software processing layers from proprietary baseband hardware. This existing cloud-native environment provides the computational agility and programmatic interfaces required to ingest telemetry and execute real-time parameter changes across hundreds of sites simultaneously.

June Moon, Executive VP and Head of R&D, Networks Business at Samsung, commented: “Samsung continues to help operators like KDDI build intelligent and efficient networks by weaving in AI-powered innovation.

“Since 2024, we have been actively testing and training our AI-powered RSO technology in the field and we have demonstrated breakthrough progress in its capabilities. This trial with KDDI exemplified how Samsung’s AI-powered innovation can bring advanced optimisation to live commercial networks while facilitating steady, seamless connectivity in an array of different network environments.”

The engineering roadmap for the partnership involves expanding the automated optimisation models to cover broader commercial bands and complex multi-carrier aggregation scenarios.

Operators managing infrastructure capital expenditure face escalating power and processing demands as 5G Standalone coverage scales. Moving toward a software-driven, automated optimisation layer represents a necessary step to control operating expenses while maximising the data capacity of existing spectral allocations.

See also: Rocket Lab buys Iridium to expand global telecom connectivity

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the IoT Tech Expo and Cyber Security & Cloud Expo. Click here for more information.

Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.