【New Update : Nov. 28】 Press Releases on AI-RAN Development

SoftBank announced the "AI-RAN" concept in 2023 as the foundation for the next-generation mobile network that will support an AI-driven society. AI-RAN is an innovative architecture that integrates AI applications and the Radio Access Network (RAN) on a unified computing platform.
Recently, SoftBank achieved a high-performance, carrier-grade virtual Radio Access Network (vRAN) on the NVIDIA GH200 Grace Hopper Superchip. Additionally, SoftBank developed an orchestrator capable of operating both vRAN and AI on the same virtualized platform, thereby realizing the AI-RAN concept by providing vRAN and diverse AI applications on a GPU-based computing infrastructure. Furthermore, SoftBank has introduced this AI-RAN solution as "AITRAS" with plans for deployment within its commercial network and expansion to telecommunications operators domestically and internationally.
In November 2024, SoftBank published seven press releases regarding the development of "AI-RAN," including the aforementioned initiative. This page will explain how each of these press releases is related.

Structure of AITRAS and Positioning of Each Press Release

1. SoftBank Corp. Announces Development of “AITRAS,” a Converged AI-RAN Solution

First, AI-RAN refers to the concept of running both RAN (Radio Access Network) processing and AI processing on the same GPU board. The term "AI-RAN" is not exclusive to SoftBank, but a technological concept that will spread globally through the activities of the AI-RAN Alliance. Based on this AI-RAN concept, SoftBank has developed a product named "AITRAS." This press release announced the name "AITRAS" for the first time and the commencement of its full-scale development.
　

2. SoftBank Corp. Develops Orchestrator to Operate AI and vRAN on the Same Virtualized Infrastructure

This press release explains the development of an orchestrator within the AI-RAN concept. Efficient utilization of GPU resources by RAN and AI requires a system that manages and controls how resources are allocated and used. This system is called an orchestrator. AI and RAN applications have different characteristics, and their simultaneous optimization is required. SoftBank has developed an orchestrator that meets these requirements, allowing "AITRAS" to provide large capacity, high performance, and high-quality vRAN at carrier grade. Additionally, it enables the integration of various AI applications, including generative AI.
　

3. SoftBank Corp. and Red Hat Initiate Joint Research and Development to Advance AI-RAN Technology

This press release announces that SoftBank has signed a Memorandum of Understanding (MoU) with Red Hat to collaborate on developing the necessary virtualization infrastructure for implementing vRAN and AI applications on the GH200. Additionally, the orchestrator mentioned in the previous section utilizes Red Hat OpenShift, a Kubernetes-based hybrid cloud application platform. With the signing of the MoU, SoftBank and Red Hat will form a joint task force to accelerate the development of new functionalities in the virtualization infrastructure and orchestrator domains.
　

4. SoftBank Corp. Develops 5G L1 Software, Achieving Carrier-grade High Performance and Quality vRAN on NVIDIA Grace Hopper Platform

This press release addresses the achievement related to the AI-RAN concept of running vRAN processing and AI processing on the same GPU board. SoftBank has developed high-quality 5G vRAN on the NVIDIA GH200 Grace Hopper Superchip platform using L1 software based on NVIDIA AI Aerial and has deployed it for outdoor proof-of-concept environments for the first time in the world. The announcement includes detailed information about the experimental environment in Fujisawa City, Kanagawa Prefecture.
　

5. SoftBank Corp. and Fujitsu Strengthen Partnership for Realization of AI-RAN Commercialization

This press release announces that SoftBank has signed an MoU with Fujitsu to strengthen their partnership for the commercialization of AI-RAN from 2026 onwards. Fujitsu is providing the vRAN (virtual Radio Access Network) software compatible with the L1 software mentioned in the previous section, as well as Radio Units (RU). Moving forward, the two companies will jointly establish an AI-RAN validation lab in Dallas, Texas, USA, and promote the research and development of software aimed at maximizing RAN performance through AI applications.
　

6. SoftBank Corp. Implements NVIDIA AI Enterprise on Edge AI Server of “AITRAS”, Converged AI-RAN Solution

This press release addresses the achievement related to the AI processing within the AI-RAN concept of running vRAN and AI processing on the same GPU board. SoftBank has implemented NVIDIA AI Enterprise software on edge AI servers for AI processing within "AITRAS." This enables the development and deployment of various AI applications on edge AI servers, leveraging Open Source Software (OSS) and high-performance AI foundational models from partner companies.
Using NVIDIA AI Enterprise implemented on "AITRAS," SoftBank has developed AI applications such as cloud robots utilizing ultra-low latency LLM, RAG Menu @Edge, and "Traffic Understanding Multimodal AI" for autonomous driving. Detailed descriptions of each application are also provided.
　

7. SoftBank Corp. Develops Traffic Understanding Multimodal AI for Autonomous Driving that Operates on Low-latency Edge AI Servers

Development of "Traffic Understanding Multimodal AI" for Autonomous Driving on Low-Latency Edge AI Servers This press release announces the development of "Traffic Understanding Multimodal AI" for autonomous driving, one of the AI applications mentioned in the previous section. This application involves training a general AI foundational model with Japanese traffic knowledge, such as traffic manuals and regulations, as well as general driving scenes and handling unpredictable driving situations. This has built a multimodal AI that acquires the broad knowledge necessary for safely driving autonomous vehicles and can highly understand traffic conditions and driving risks.
Moreover, since this application operates on low-latency edge AI servers within AI-RAN, it can support autonomous driving remotely in real-time, where significant delays are unacceptable.

Through these seven press releases, SoftBank has announced significant achievements in establishing and commercializing AI-RAN technological foundations focused on the orchestrator, RAN infrastructure, and AI infrastructure. Moving forward, SoftBank aims to advance the implementation of this foundation into commercial networks and expand it to telecommunications operators domestically and internationally.
　

Related Press Releases
・NVIDIA Press Release
NVIDIA and SoftBank Corp. Accelerate Japan’s Journey to Global AI Powerhouse
・Red Hat Press Release
Red Hat and SoftBank Corp. Collaborate on Research and Development of AI-RAN Technologies

Related Sites
・NVIDIA Tech Blog
AI-RAN Goes Live and Unlocks a New AI Opportunity for Telcos

・AI-RAN Introduction Video

・AI-RAN Summary
https://www.softbank.jp/en/corp/technology/research/story-event/057/