SoftBank Develops AI-RAN Orchestrator for Unified Network Infrastructure

Philip Lee -

Tokyo, Japan- SoftBank (9434. T) has developed an orchestrator capable of running artificial intelligence (AI) and virtualized Radio Access Network (vRAN) applications on the same virtualized infrastructure.

The new orchestrator enables carrier-grade vRAN deployment alongside various AI applications, including generative AI, on GPU-computing infrastructure such as the NVIDIA GH200 Grace Hopper Superchip.

The technology addresses previous challenges in running AI and vRAN workloads simultaneously, as these applications typically require different Service Level Agreements (SLA) and server configurations.

SoftBank built the orchestrator using Red Hat OpenShift, allowing efficient resource management across AI and vRAN applications. 

The system integrates with NVIDIA AI Enterprise's serverless APIs, enabling users to run custom AI applications on the AI-RAN virtualized infrastructure.

"This realization of the AI-RAN concept and orchestrator development represents a significant step forward," said Alex Jinsung Choi, Chairman of the AI-RAN Alliance and advisor to SoftBank's Advanced Technology Lab.

The orchestrator's key features include deployment request matching, dynamic infrastructure resource adjustment, and centralized management of multi-cluster environments across Japan.

NVIDIA's VP of AI & Communications predicted the development would create an AI marketplace for AI-RAN infrastructure, potentially generating "billions of dollars in revenue opportunities for telecom operators worldwide from underutilized RAN infrastructure."