SoftBank Corp. and Ericsson Achieve Interworking
Between the AITRAS Orchestrator and
Ericsson Intelligent Automation Platform

Background and challenges

AI-and-RAN is a concept within AI-RAN that aims to efficiently utilize server computing resources by running workloads with different characteristics—AI workloads and RAN workloads—on a single virtualized infrastructure.

As the use of AI expands and demand for communication networks continues to grow, there is an increasing need for management and operational technologies that can monitor demand for computing resources used for AI processing and RAN control in real time and dynamically and flexibly allocate those resources. In particular, interworking and role sharing between an orchestrator that manages and controls the overall AI-RAN platform and infrastructure, and SMO that manages and controls the RAN domain, represent key challenges in the implementation of AI-RAN.

Overview of the initiative

Against this backdrop, SoftBank and Ericsson defined and implemented a method to enable interworking between the AITRAS Orchestrator and the Ericsson Intelligent Controller (EIC) and Ericsson Orchestrator Lifecycle Manager (EO-LM), which are components of EIAP, and built a system in which the two functions operate in coordination. This project also demonstrates that, through the interworking between the AITRAS Orchestrator and EIAP, coordinated operation across AI workloads and the RAN domain is achievable.

In this initiative, when the EIAP requests the use of computing resources based on RAN requirements, the AITRAS Orchestrator allocates the required computing resources by taking into account overall utilization of the AI-RAN platform. This configuration enables computing resources that are temporarily required for RAN use to be dynamically allocated from the AI-RAN platform to the EIAP and promptly returned once processing is complete.

Specifically, the AITRAS Orchestrator identifies Kubernetes2 clusters that can be allocated for RAN use and generates temporary access tokens, which are provided to the EIAP. The EIAP then utilizes the relevant clusters as extended resources of the O-Cloud3 and executes the rApp⁴ workloads. After the workloads complete execution, the EIAP returns the computing resources to the AITRAS Orchestrator.

Through this coordination between two functions responsible for managing different domains, the dynamic adjustment and allocation of computing resources required for both RAN and AI use cases has been achieved.

[Notes]

1. *2
   Kubernetes is an open-source system for automating deployment, scaling, and management of containerized applications.
2. *3
   O-Cloud is an open specification defined by the O-RAN ALLIANCE for platforms that host RAN applications.
3. *4
   rApp is a cloud-based application that operates on the Non-Real-Time RIC (RAN Intelligent Controller) within the O-RAN architecture to enhance and optimize radio access network using AI and other advanced technologies.

This initiative realizes the coordinated operation of AI and RAN and dynamic allocation of computing resources, one of the use cases defined in the AI-RAN Platform & Infrastructure Orchestrator by the AI-and-RAN Working Group of the AI-RAN Alliance.

Next steps

A demonstration of this technology is scheduled to be showcased at the Ericsson booth at MWC Barcelona 2026.

Building on the insights gained through this initiative, SoftBank and Ericsson will continue research, development, and demonstrations aimed at further advancing AI-RAN. Going forward, the two companies will continue to explore interfaces that support coordinated operation of AI and RAN, with the goal of realizing more flexible and efficient network operations.

Executive comments

Ryuji Wakikawa, Vice President and Head of the Research Institute of Advanced Technology at SoftBank Corp., said:
"Within AI-RAN, resources are no longer dedicated solely to RAN. GPU resources across SoftBank's network are shared between RAN and AI, and centrally managed by the AI-RAN Orchestrator. When required, resources are dynamically allocated to RAN and seamlessly handed over to the SMO, which takes responsibility for RAN deployment and control. Establishing this integrated architecture between the AI-RAN Orchestrator and the SMO represents a significant step toward realizing AI-RAN."

Jawad Manssour, President and Representative Director at Ericsson Japan K.K., said:
"AI-and-RAN enables intelligence to be delivered deep into the network while enabling flexible and distributed orchestration. Through our collaboration with SoftBank, we are demonstrating how rApps (non-real-time RIC applications) and AI workloads can be harmonized and orchestrated together, unlocking new efficiencies and automation."