- SoftBank and Ericsson have deployed an AI-driven optimisation system for massive MIMO base stations that automatically adjusts coverage in response to real-time traffic changes.
- Early trials at Expo 2025 Osaka delivered a 24% increase in downlink throughput, showing that AI can directly enhance network performance beyond hardware upgrades.
What happened: AI begins to reshape fundamental mobile network operations
Japan’s SoftBank Corp. and Ericsson Japan K.K. have taken a significant step in embedding artificial intelligence into the core of mobile network radios. According to reports on 29 January, The partners have begun deploying an AI-powered system that dynamically optimises massive multiple-input, multiple-output (MIMO) coverage in 5G networks, using machine learning models that respond to minute-by-minute changes in user distribution and traffic load.
The system, designed to operate at base stations serving large venues such as arenas and parks, collects user distribution data every minute and uses AI to automatically adjust both horizontal and vertical antenna patterns on massive MIMO radios. SoftBank and Ericsson first demonstrated the concept at Expo 2025 Osaka in Kansai, where the system achieved a 24% improvement in downlink throughput when traffic fluctuated sharply — a relevant metric for event-driven demand.
SoftBank’s move builds on a longstanding research collaboration with Ericsson that now includes AI-RAN development and broader 5G/6G innovation. The partners are now operating the AI system at multiple large-scale event venues across Japan.
Also Read: Telefónica Tech UK&I unveils AI-driven managed Security Service Edge for British and Irish firms
Also Read: Telco capex shifts to cloud, AI, and core networks as Ran spending stays muted
Why it’s important
This development signals a paradigm shift in wireless performance engineering: 5G and future 6G networks are increasingly becoming software-defined and AI-enabled, not just defined by more spectrum or raw hardware capacity. By embedding AI into the radio access layer itself, operators like SoftBank are proving that network performance can be tuned in real time based on user behaviour and instantaneous demand — a leap from static, pre-programmed coverage planning.
For technology enterprises, cloud partners and network equipment vendors, this matters because it points to a future where AI becomes an integral part of network infrastructure, shaping performance outcomes and reducing reliance on traditional hardware upgrades alone. Networks that can self-optimise in response to traffic dynamics will be better placed to support demanding use cases such as augmented reality, IoT, autonomous systems and high-density event connectivity. As 5G monetisation challenges persist, embedding AI into RAN operations emerges as a compelling lever for cost-efficient performance gains and differentiation.
