Table of Contents

1. Introduction
2. SoftBank and NVIDIA Collaboration
3. AI-RAN Software Stack
4. Achieving AI-RAN Multi-Tenancy
5. Enabling an AI-RAN Marketplace
6. AI-RAN Applications Showcased
7. Accelerating the AI-RAN Business Case

1. Introduction

AI-RAN technology is expanding rapidly, bringing new opportunities for telcos. Traditional RAN systems are not equipped to handle AI traffic, necessitating the development of new infrastructure. SoftBank and NVIDIA have collaborated to bring the AI-RAN vision to life, with successful outdoor trials showcasing the capabilities of AI-RAN.

2. SoftBank and NVIDIA Collaboration

SoftBank and NVIDIA's collaboration resulted in a successful outdoor trial in Fujisawa City, Japan. Using NVIDIA-accelerated hardware and software, the trial integrated full-stack virtual 5G RAN software with 5G core. This end-to-end solution incorporates various components to enable AI processing in RAN infrastructure.

3. AI-RAN Software Stack

The AI-RAN software stack includes SoftBank's 5G RAN Layer 1 functions, Fujitsu software for Layer 2 functions, and Red Hat's OpenShift Container Platform for virtualization. A SoftBank-developed E2E AI and RAN orchestrator ensures seamless provisioning of RAN and AI workloads. The NVIDIA GH200 Grace Hopper Superchip serves as the underlying hardware for AI-RAN infrastructure.

4. Achieving AI-RAN Multi-Tenancy

Multi-tenancy is a key aspect of AI-RAN technology, allowing concurrent processing of RAN and AI workloads without compromising performance. The Fujisawa City trial demonstrated successful concurrent AI and RAN processing using static allocation of resources between the two workloads.

5. Enabling an AI-RAN Marketplace

With AI computing capacity available on distributed infrastructure, an AI-RAN marketplace can be established. The integration of the NVIDIA AI Enterprise serverless API with the SoftBank E2E AI-RAN orchestrator enables telcos to become an AI distribution grid and tap into new revenue opportunities.

6. AI-RAN Applications Showcased

During the outdoor trial, SoftBank showcased various edge AI applications over the live AI-RAN network. These applications included remote support for autonomous vehicles, factory multi-modal AI applications, and robotics applications. The demo highlighted the benefits of edge AI processing for real-time applications.

7. Accelerating the AI-RAN Business Case

The Aerial RAN Computer-1 plays a crucial role in accelerating the AI-RAN business case. Energy efficiency and economics of GPU-enabled infrastructure are key considerations, especially in comparison to traditional CPU- and ASIC-based RAN systems. SoftBank's advancements in AI-RAN technology are paving the way for future innovations in the telecommunications industry.