SoftBank's AITRAS Implements Centralized and Distributed AI-RAN Architectures Using Arm-based NVIDIA Platform

SoftBank Corp. (“SoftBank”), in collaboration with Fujitsu Limited (“Fujitsu”), completed the implementation of CU (Central Unit) functions on the NVIDIA Grace CPU Superchip platform as part of its efforts toward accelerating the commercialization of “AITRAS,”^* a converged AI-RAN solution. Additionally, the implementation of D-RAN (Distributed RAN) functionality, which deploys both the CU and DU (Distributed Unit) functions on a single NVIDIA GH200 Grace Hopper Superchip, was completed.

The NVIDIA Grace CPU Superchip leverages two Arm® Neoverse™ V2-based Grace CPUs.

Through its development efforts, SoftBank achieved two types of architectures in AITRAS: C-RAN (Centralized-RAN), which deploys CU and DU on separate servers to enhance base station resource efficiency, and D-RAN, which processes CU and DU on the same server to minimize hardware configuration.

This development enables reduced hardware requirements when commercially deploying AITRAS, which is expected to bring cost savings and improved efficiency in implementation.

AITRAS, which is currently undergoing verification in outdoor environments, is configured as a C-RAN architecture designed for large-scale data centers for telecom operators, with CU processing being executed on NVIDIA Grace Hopper servers. Through SoftBank's development initiative, CU processing can now be executed on NVIDIA Grace CPU servers, doubling the number of RUs (Radio Units) that can be accommodated by a single CU. This advancement reduces the number of servers required for CU processing, contributing to cost reduction in C-RAN architecture for data centers.

Additionally, the D-RAN architecture is designed for environments with small-scale RU deployments, where CU and DU processing servers are deployed at the same site as the RU. In conventional C-RAN architecture, such deployments required separate servers for CU and DU processing, leading to higher deployment costs. With the development of D-RAN architecture that enables both CU and DU processing on the same NVIDIA Grace Hopper server, hardware requirements for commercial deployment can be reduced.

SoftBank plans to progressively deploy C-RAN and D-RAN architectures in outdoor testing environments from fiscal year 2025 (April 1, 2025 - March 31, 2026), proceeding with performance verifications.

Hideyuki Tsukuda, Executive Vice President & CTO at SoftBank Corp. said: “This achievement in the development of ‘AITRAS’ represents an important step forward for our technological innovation. Through close collaboration with our partner companies, we succeeded in realizing both C-RAN and D-RAN architectures. We are confident this will significantly improve the efficiency and cost-performance of network deployment.”

Soma Velayutham, Vice President of Telecoms at NVIDIA, said: “AI-RAN requires a scalable and 100% software defined platform to enable extension to both centralized and distributed network architectures, SoftBank's continuous innovations on AITRAS, harnessing NVIDIA AI Aerial, are accelerating the path to commercial deployments of AI-RAN.”

Eddie Ramirez, vice president of go-to-market, Infrastructure Line of Business, Arm, said: “Today's AI-RAN solutions demand the highest levels of both performance and energy efficiency in order to continue to scale, by leveraging market-leading Neoverse technology, SoftBank's AITRAS solution pushes the boundaries of RAN performance while making network deployment more cost effective. We look forward to continuing our collaboration with SoftBank, NVIDIA, Fujitsu and telco partners in driving innovation in this space.”

Shingo Mizuno, Corporate Executive Officer & EVP, Vice Head of System Platform (in charge of Network Business) Fujitsu Limited, said: “The C-RAN/D-RAN configuration, achieved in part via Fujitsu's vRAN software and engineering, will reduce deployment costs and power consumption, and marks a significant achievement toward the commercialization of AI-RAN. Fujitsu will continue to provide value to customers and society through the development and provision of new technologies like vRAN software.”