SoftBank Corp. Announces Development of "AITRAS," a Converged AI-RAN Solution

In This Article:

(Graphic: Business Wire)
(Graphic: Business Wire)

Global deployment of AITRAS planned for telecom operators

TOKYO, November 13, 2024--(BUSINESS WIRE)--SoftBank Corp. (TOKYO:9434, "SoftBank") announced it launched full-fledged initiatives to develop "AITRAS," a convergence solution based on the AI-RAN concept that can host both AI and RAN (Radio Access Network) workloads on the same NVIDIA-accelerated computing platform. SoftBank will not only implement AITRAS – a carrier-grade convergence solution with high capacity, high performance and high quality – across its own commercial network, it also aims to expand the solution to telecom operators globally from 2026 and beyond.

AITRAS is the productization of AI-RAN, an architecture that is aimed at achieving high performance and efficiency in AI and RAN while leveraging the features of NVIDIA’s AI and accelerated computing platform. The solution not only provides a carrier-grade RAN with high capacity, high performance and high quality on an NVIDIA GH200 Grace Hopper Superchip platform — it also enables concurrent and efficient operations with a variety of AI applications, including generative AI.

In real-world wireless environments, various high-speed processing and algorithmic functions are applied to maximize performance while dealing with daily occurrences such as sudden traffic surges and multipath radio interference. However, as these processes become increasingly advanced and complex, maintaining system stability becomes a challenge.

AITRAS’s L1* software, which is SoftBank’s newly developed L1 software that is based on the NVIDIA AI Aerial platform, provides the high stability and high performance essential for a carrier-grade RAN through parallel processing of the signal and optimization of task initiation timing, while also contributing to the maximization of the RAN capacity and reduction in power consumption.

SoftBank also developed an orchestrator that coordinates with the virtualization platform, which is expected to improve cost-effectiveness through operational efficiency and reduced power consumption by efficiently and dynamically allocating computing resources based on the unique application characteristics of AI and RAN.

By implementing AITRAS, telecom operators can build an AI infrastructure while maintaining investments in its traditional RAN infrastructure. Additionally, by leveraging AI for network optimization, they can achieve efficient and innovative infrastructure investments.

AITRAS is implemented with NVIDIA AI Enterprise, a software platform that includes a set of functions that enables the easy development and deployment of large language models (LLMs). With this implementation, enterprise customers can also develop and deploy their own AI applications. Further details are available in the press release "SoftBank Corp. Implements NVIDIA AI Enterprise on Edge AI Server of AITRAS Converged AI-RAN Solution" issued on November 13, 2024.