Nvidia, the chipmaker giant, and SoftBank Corp, the telecom unit of Softbank Group, have announced the trial of the world’s first AI-powered 5G telecommunications network, according to joint statement issued by the companies. The system can simultaneously handle AI and 5G workloads, a technology known as AI-RAN (artificial intelligence radio access network). This network can support various applications, including remote support for autonomous vehicles, robotics control, and multimodal retrieval at the edge.
This represents a major shift for the telecom sector by introducing a more efficient infrastructure. This innovation enables telecom companies to turn existing network infrastructure, typically seen as a cost centre, into a potential revenue source by offering AI-driven services such as edge computing, AI inference, and enhanced data processing. Additionally, the ability to leverage AI in managing network traffic, maintenance, and optimisation can lead to greater scalability and adaptability.
The AI-RAN infrastructure is supported by the telecom industry and offers operators a chance to transform traditional base stations into revenue-generating AI assets. While traditional telecom networks typically operate at one-third of their capacity, AI-RAN is expected to enable telcos to monetise the remaining two-third capacity through AI inference services, it adds.
According to estimates from Nvidia and SoftBank, telecom operators could generate $5 in AI inference revenue for every $1 invested in AI-RAN infrastructure. Returns of up to 219% are anticipated by SoftBank on the AI-RAN servers it integrates into its infrastructure.
Jensen Huang, CEO of Nvidia, also revealed that SoftBank is the first company to receive Nvidia's new Blackwell chip designs, which will be used in its AI supercomputer to develop its own generative AI model.
The two companies also unveiled plans for SoftBank to use Nvidia’s AI Enterprise software to create a local, secure AI marketplace to meet the demand for AI computing. This initiative aims to position SoftBank as Japan's central AI grid, facilitating the development, distribution, and utilisation of AI services across industries and businesses.
Ronnie Vasishta, Nvidia's senior vice president of telecom, stated that shifting to multi-purpose AI-RAN networks could result in five times more revenue for every dollar invested. SoftBank’s field trial has proven the feasibility, performance, and economic potential of AI-RAN, he adds.
SoftBank's AI-RAN solution, “AITRAS,” was developed through a five-year collaboration with Nvidia. AITRAS integrates AI and RAN workloads through an orchestrator designed by SoftBank, improving communication efficiency by using a single Nvidia-accelerated GPU server for dense cells.
In 2020, Rakuten Mobile, a Japanese telecom operator, reportedly launched the world’s first commercial Open RAN (O-RAN) network. This deployment introduced a fully virtualized, cloud-native 4G and 5G network based on O-RAN architecture, prioritising cost-efficiency, flexibility, and innovation.