January 23, 2026

SoftBank to Begin Building High-Speed AI Communications Network This Year

Askar - stock.adobe.com

Starting in 2026, SoftBank will build a new high-speed communications network integrating mobile base stations with artificial intelligence platforms. The system will handle both communications control and AI data processing simultaneously, enabling local AI operations without relying on large-scale centralized data centers. This shift is expected to promote decentralized power consumption and accelerate the development of infrastructure for autonomous vehicles, robotics, and other next-generation technologies.

The company has developed a GPU-powered base station system, AITRAS, in collaboration with US semiconductor leader NVIDIA and UK-based Arm, a subsidiary of SoftBank Group. SoftBank plans to deploy the system at several base stations across Japan by 2026. By incorporating edge AI capabilities that enable local processing without sending data over the internet, the new network is designed to significantly reduce reliance on external data centers.

Base stations capable of running AI are expected to improve the accuracy and responsiveness of autonomous driving systems. They will also enable real-time analysis in locations such as factories and offices, where data can be processed close to its source, and results returned instantly.

The Coming Data Explosion and the Need to Rethink Japan’s Network Infrastructure

Data processing demand in Japan is projected to increase more than 300-fold in the decade leading up to 2030 as AI applications spread. Currently, around 80% of data centers are concentrated in the Tokyo metropolitan area and the Kansai region. Heavy reliance on cloud-based processing can lead to network congestion and excessive power consumption due to long-distance data transmission. Decentralizing processing is therefore seen as essential for improving energy efficiency and reducing costs.

SoftBank operates more than 300,000 base stations nationwide. Traditionally, these stations have run on dedicated platforms optimized solely for communications processing, with most information processing handled by CPUs. However, CPUs deliver less than 10% of the processing performance of GPUs and are ill-suited for high-speed parallel tasks. By introducing GPU-based systems, SoftBank aims to fundamentally redefine the role of base stations in the AI era.