SoftBank will deploy the new AITRAS Converged AI-RAN Solution in its own network and offer it commercially to other operators from 2026
Amid a raft of announcements about the joint efforts of Softbank and NVIDIA to “accelerate Japan’s journey to being a global AI powerhouse, the AI-RAN stands out for telcos.
The press statement describes AI-RAN as new sort of network that runs AI and 5G workloads concurrently. In an outdoor trial, conducted in the Kanagawa prefecture, Softbank achieved “carrier-grade 5G performance” while running AI inference workloads on the network’s unused capacity.
SoftBank used NVIDIA’s AI Enterprise to build “real-world AI inference applications” for the trial, including autonomous, remote support for vehicles, robotics control and “multimodal retrieval-automated generation at the edge”.
Adding up the numbers
The premise is that, on average, mobile operators only use about a third of RAN capacity. Bringing AI into the mix would enable operators telcos to monetise the idle capacity for “AI inference services”.
The AI-RAN solution would turn compute power up or down dynamically to meet demand and supply without affecting the network’s performance. The idea is to build an ecosystem linking the demand and supply of AI technology using NVIDIA’s AI Enterprise serverless APIs and orchestrator.
The ecosystem could send external AI inferencing tasks to an AI-RAN server that has computing resources available to deliver “localised, low-latency, secure inferencing services”.
NVIDIA and SoftBank suggest operators could earn about $5 from AI-inference revenue from every $1 of capex they invest in new AI-RAN infrastructure. SoftBank estimates it have an ROI of up to 219% for every AI-RAN server it adds to its infrastructure.
Softbank plans to use this tech in its own commercial network and offer it to telcos globally from 2026 in the shape of AITRAS technology. It says this is a “convergence solution based on the AI-RAN concept that can host both AI and RANRAN…workloads on the same NVIDIA-accelerated computing platform”.
Source: Softbank – link here
Forget single-purpose RAN
“Shifting from single-purpose to multi-purpose AI-RAN networks can mean 5x the revenue for every dollar of capex invested,” said Ronnie Vasishta, SVP 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.”
SoftBank also announced it is building “Japan’s most powerful AI supercomputer” using the NVIDIA’s Blackwell platform, and will use the NVIDIA Grace Blackwell platform for its next supercomputer.
Further, SoftBank intends to use the DGX SuperPOD for its own GenAI development and other AI-related activities, as well as for those of universities, research institutions and businesses in Japan.