SoftBank says its AI-RAN will make base stations ‘AI revenue producing assets’

AI-RAN is described as a new kind of telecommunications network that can run AI and 5G workloads at the same time. Softbank’s outdoor trial, conducted in the Kanagawa prefecture, achieved carrier-grade 5G performance and was able to do so while using the network’s excess capacity to run AI inference workloads concurrently, says the release.

For the trial, SoftBank used Nvidia AI Enterprise to build ‘real-world AI inference applications’ such as autonomous vehicle remote support, robotics control and multimodal retrieval-automated generation at the edge.

The pitch is that ‘traditional’ telco networks are designed to handle peak loads an on average use only a third of that capacity, while bringing AI-RAN into the picture means telcos could monetise the remaining capacity for ‘AI inference services.’

In terms of what this could mean economically, Nvidia and SoftBank reckon that operators can earn roughly $5 in AI inference revenue from every $1 of capex they invest in new AI-RAN infrastructure. SoftBank estimates it can achieve a return of up to 219% for every AI-RAN server it adds to its infrastructure.

AI-RAN solutions need to spin compute up or down dynamically based on demand and supply without compromising performance, we’re told, and so it wants to build an ecosystem that connects the demand and supply of AI technology by using Nvidia AI Enterprise serverless APIs and its in-house developed orchestrator.

The point of this is so that it can dispatch external AI inferencing jobs to an AI-RAN server when computing resources are available to deliver ‘localised, low-latency, secure inferencing services’, we’re told.

As well as using all this for its own commercial network, it also aims to expand the solution to telecom operators globally from 2026 via its AITRAS tech, described as 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”.

“Shifting from single-purpose to multi-purpose AI-RAN networks can mean 5x the revenue for every dollar of capex invested,” said Ronnie Vasishta, senior vice president of telecom at Nvidia. “SoftBank’s live field trial marks a huge step toward AI-RAN commercialization with the validation of technology feasibility, performance and economics.”

Ryuji Wakikawa, vice president and head of the Research Institute of Advanced Technology at SoftBank added: “SoftBank’s ‘AITRAS’ is the first AI-RAN solution developed through a five-year collaboration with Nvidia. It integrates and coordinates AI and RAN workloads through the SoftBank-developed orchestrator, enhancing communication efficiency by running dense cells on a single Nvidia-accelerated GPU server. We are confident this AI-driven innovation, AITRAS, will pave the way for new business models in telecommunications, serving as a crucial factor in the transformation of mobile operators.”

Meanwhile it was also announced that SoftBank is building ‘Japan’s most powerful AI supercomputer’ using the Nvidia Blackwell platform, and plans to use the Nvidia Grace Blackwell platform for its next supercomputer.

SoftBank plans to use the DGX SuperPOD for its own generative AI development and other AI-related pursuits, as well as that of universities, research institutions and businesses across Japan, we’re told.