en.Wedoany.com Reported - On July 2, Japan's SoftBank Corp. and SoftBank Group Corp. announced that they will establish SB Neo Inc. (tentative name) in the United States in July 2026 to operate a new cloud business targeting the U.S. market. The new company will be 51% owned by SoftBank Corp. and 49% by SoftBank Group Corp., making it a consolidated subsidiary of SoftBank Corp. It plans to launch AI computing cloud services for large U.S. enterprises within the fiscal year ending March 2028.
SB Neo's primary target customers are U.S. enterprises requiring large-scale AI model training and inference resources. The company will leverage the 10-gigawatt-class energy and AI infrastructure being developed by SoftBank Group Corp. to provide computing resources for large model training, inference, data processing, and high-performance computing tasks. According to current disclosures, service capacity will be expanded in phases, with plans to subsequently deploy 10-gigawatt-scale AI infrastructure.
The key to the new cloud business is not just "renting GPUs," but combining data centers, power supply, GPU clusters, orchestration software, and enterprise-grade cloud services. AI training and inference impose high demands on computing density, power stability, network interconnection, storage throughput, cooling systems, and operations platforms. This is especially true for large U.S. enterprises training foundation models, deploying industry-specific models, running AI agents, and conducting multimodal inference, which require continuous, scalable, and manageable computing resources. If SB Neo proceeds on a 10-gigawatt scale, its business scope will resemble a combination of an AI data center operator, a GPU cloud service provider, and an AI infrastructure platform, rather than the standard virtual machine services found in traditional public clouds.
SoftBank Corp. has already been advancing its neocloud business in Japan. On May 25, the company announced that it would launch the "AI Data Center GPU Cloud" equipped with the "Infrinia AI Cloud OS" in October 2026, and has already provided a beta version. This service combines AI computing infrastructure in Japanese data centers with an AI data center software stack, supporting Kubernetes as a Service and Inference as a Service, and providing GPU cloud capabilities for model training, inference, and data processing. The infrastructure includes advanced GPU platforms such as the NVIDIA GB200 NVL72.
The experience gained in the Japanese market will serve as a technical reference for SB Neo in the U.S. For AI cloud services to truly serve large enterprises, the platform must address issues such as multi-tenant isolation, GPU resource scheduling, containerized deployment, inference API management, automatic fault recovery, and cost control. The Infrinia AI Cloud OS is positioned as an AI data center software stack designed to centrally manage GPU resources, the Kubernetes runtime environment, and AI workload execution, thereby reducing the operational burden on enterprises for building their own development environments, managing computing resources, and deploying inference services.
The new U.S. cloud business also faces energy constraints. The construction of AI data centers increasingly relies on the synchronized planning of power and site resources; simply procuring GPUs does not ensure stable service capabilities. SoftBank Group Corp. has already been advancing the development of 10-gigawatt-class data centers and supporting power generation facilities in the U.S. Related plans involve 10 gigawatts of data center capacity, 10 gigawatts of new power generation capacity, and arrangements to strengthen the transmission network.
Market sources also indicate that after SB Neo launches its new cloud services in the U.S., it could significantly boost SoftBank Corp.'s profitability, with annual operating profit potentially reaching the 3 to 4 trillion yen range. This estimate is currently closer to external expectations, and the actual business contribution will depend on data center construction progress, GPU supply, customer contracts, power costs, cloud service pricing, and equipment utilization. What is clear at this point is that SoftBank Corp. will hold a controlling stake in the operating entity, SoftBank Group Corp. will provide U.S. AI infrastructure resources for support, and SB Neo will serve as the platform for launching the new cloud business in the U.S.
