SoftBank and Ericsson have validated AI RAN software on a live commercial 5G network in Japan, marking the first such demonstration of the vendor’s technology in the country. The two companies announced on 20 August that the trial tested an AI-native scheduler for link adaptation, a core function within Ericsson’s AI in RAN software suite. The validation applied artificial intelligence directly inside the radio access network to make real-time decisions on how data is transmitted to users.
The radio access network, commonly known as the RAN, connects user devices to the mobile network and plays a central role in overall network capacity and performance. By embedding AI into this layer, operators can potentially handle more traffic within the same spectrum bands without requiring additional frequency resources.
SoftBank and Ericsson Validate AI RAN on 5G
The trial was conducted on SoftBank’s commercial 5G network rather than in a lab setting, which means the AI RAN software was tested under real-world conditions with actual user traffic. SoftBank defined the requirements for evaluation areas based on traffic characteristics, while Ericsson trained the AI-native scheduler model using actual network data. Ericsson also implemented and tuned real-time channel-capacity prediction and selection of optimal downlink transmission rates.
The AI model was designed to learn and infer how to respond to complex, constantly changing radio conditions. This included adjusting link parameter settings for users located at cell edges and in areas experiencing high radio interference, where network performance typically suffers the most.
AI Scheduler Improves Network Efficiency
The results showed measurable gains in spectral efficiency and downlink user throughput. Under the tested conditions, spectral efficiency improved by up to approximately 25 per cent, while downlink user throughput rose by up to approximately 50 per cent compared with conventional technology. Across all evaluation areas, both spectral efficiency and downlink user throughput improved by an average of around 10 per cent.
These gains were confirmed specifically under difficult radio conditions such as cell boundaries and congested areas. Spectral efficiency measures how much data traffic can be transmitted within a given frequency bandwidth, and improvements in this metric directly expand an operator’s traffic-handling capacity without needing new spectrum. The AI RAN software runs on baseband equipment at mobile base stations, enabling real-time AI-driven decision-making at the network edge.
Trial Supports 5G Advanced and 6G Development
The project forms part of the broader collaboration between SoftBank and Ericsson toward building AI-native RAN capabilities. Both companies said the demonstration confirmed that AI can contribute to high-performance, secure and flexible networks aimed at 5G Advanced and future 6G. 5G Advanced refers to the next evolutionary phase of 5G standards before 6G is expected to arrive.
Mobile operators around the world are increasingly exploring AI-based network optimisation as data traffic continues to grow. Applications such as generative AI assistants, autonomous agents and immersive services are placing more varied and demanding requirements on wireless networks, making efficient use of existing spectrum resources increasingly important. The validation of AI-native RAN technology on a live 5G network represents a practical step toward integrating AI into everyday mobile network operations.