SoftBank Pivots from Telco to AI Infra
SoftBank is rebranding its core business, transitioning from a traditional telecommunications carrier to an "AI infrastructure provider." The company's new focus is on orchestrating intelligence across distributed edge and cloud environments to power devices, robots, and other systems.
This strategic shift is underpinned by massive capital reallocation, including the divestment of its entire stake in Nvidia for approximately $5.83 billion to fund new AI ventures. SoftBank is channeling significant funds, reportedly $22.5 billion, into OpenAI to finance the immense capital expenditures required for next-generation AI systems and supercomputers, such as the rumored "Stargate" project. This is part of a larger $110 billion investment in OpenAI, alongside Amazon and Nvidia. A core component of this new direction is the "AI-RAN" (AI-Radio Access Network) initiative, which integrates AI into wireless network infrastructure. SoftBank is developing its own AI-RAN product called "AITRAS," which combines RAN control and AI server functions on a single GPU-based platform. This technology has already demonstrated a 30% improvement in 5G uplink speeds in network tests. The company's "Telco AI Cloud" vision aims to create a distributed AI infrastructure across Japan, combining large-scale AI data centers for training with low-latency edge servers, or "Regional Brains," for inference. This architecture is designed to support applications requiring real-time processing, such as autonomous systems and "Physical AI," by allowing devices like robots to offload heavy AI workloads to the network edge. To execute this vision, SoftBank is collaborating with several key technology partners. A five-year collaboration with NVIDIA is central to developing AI-RAN solutions and building Japan's most powerful AI supercomputer using NVIDIA's Blackwell platform. The company is also working with Ericsson to enable dynamic AI workload offloading and with Nokia on executing external AI workloads on the AI-RAN infrastructure. This pivot marks a significant change in investment strategy for SoftBank, moving from a broad venture capital model to highly concentrated, capital-intensive bets on foundational AI infrastructure. The firm is not just investing in AI companies but is actively building the core compute layer, a move that repositions it as a direct player in the AI race. This has led to a global restructuring of its Vision Fund team to focus resources on these large-scale AI initiatives. The strategy also involves fostering a domestic AI ecosystem in Japan. SoftBank has launched initiatives like the "AI Foundation for Startups" and the "AI Boost Program" in partnership with NVIDIA to provide startups with access to high-performance GPU systems and technical mentorship. The goal is to support the development and commercialization of new AI solutions on its emerging infrastructure.
