SoftBank Corp. has announced a new vision, Telco AI Cloud, aimed at building next-generation social infrastructure for the AI era by leveraging its nationwide telecommunications foundation.
Telco AI Cloud is an AI infrastructure vision that integrates a large-scale AI data center platform powered by a GPU (Graphics Processing Unit) cloud, an AI-RAN-based MEC (Multi-access Edge Computing) platform, and a software stack for AI data centres called ‘Infrinia AI Cloud OS’. By optimising AI processing from training to inference and utilising its nationwide telecommunications infrastructure, SoftBank will build a distributed AI infrastructure that delivers low latency, high reliability, and sovereign capability (data sovereignty). Through its Telco AI Cloud vision, SoftBank aims to evolve beyond the traditional role of a telecommunications operator and become an AI infrastructure provider.
Telco AI Cloud is a vision uniquely enabled by SoftBank’s position as a telecommunications operator with a nationwide network infrastructure. Unlike hyperscaler-type centralised clouds, it enables the construction of a distributed AI infrastructure embedded within a telecommunications network.
Telco AI Cloud consists primarily of:
In its AI-RAN product ‘AITRAS’, SoftBank is developing an orchestrator (‘AITRAS Orchestrator’) that monitors in real time the demand for computing resources used for both AI processing and RAN (Radio Access Network) control. Based on multiple indicators – such as resource availability, application requirements, and projected power consumption – the orchestrator dynamically and flexibly allocates resources. Within Telco AI Cloud, RAN itself is managed as a unified AI application, enabling advanced cross-domain control of computing resources across both the telecommunications network and AI processing infrastructure.
In distributed AI infrastructure environments, variations in configurations across sites can create operational complexity. To address this challenge, SoftBank has developed ‘Infrinia AI Cloud OS’, a software stack that provides integrated management from GPUs and telecommunications networks to Kubernetes and AI workloads. This enables optimised AI processing from training to inference and supports secure, multi-tenant GPU cloud operations.
