SoftBank Launches Telco AI Cloud Distributed Infrastructure

BARCELONA, Spain — SoftBank Corp. presented a framework for distributing artificial intelligence infrastructure across its national telecommunications network. 

The company stated that this approach is intended to position it as an AI infrastructure provider rather than a conventional telecommunications carrier.

The initiative, referred to as "Telco AI Cloud," integrates large-scale GPU data centers for AI training with an AI-RAN-based edge computing platform for low-latency inference and a proprietary software stack labeled "Infrinia AI Cloud OS."

This architecture differs from centralized hyperscale cloud models by distributing AI processing across existing telecommunications networks.

A key component is the AITRAS Orchestrator, which monitors demand and allocates computing resources between AI applications and radio access network controls based on availability and power consumption.

SoftBank announced several commercial partnerships related to the platform at the Mobile World Congress in Barcelona.

The company stated that it completed a proof-of-concept with Ericsson to test low-latency AI-RAN networks for industrial robots, as part of an existing arrangement with Yaskawa Electric.

The company also announced an agreement with Mitsubishi Heavy Industries to deploy edge data centers for on-premises AI inference in manufacturing operations.

SoftBank additionally confirmed integration agreements with Ericsson's O-RAN-compliant management systems and Nokia's AI platforms.

The company stated that it open-sourced a resource-allocation component of its orchestrator and joined the Linux Foundation's OCUDU initiative to support development of open-source AI-RAN technologies.