SoftBank establishes SB Neo to launch 10GW U.S. neocloud business
SoftBank plans to launch SB Neo in March 2027, a new U.S.-based neocloud business providing GPU compute and inference-as-a-service to hyperscalers and enterprises. The venture aims to reach 10 gigawatts of capacity by 2030, targeting heavy AI training and streaming inference workloads.
Key Takeaways
- SB Neo will launch operations in March 2027 as a consolidated subsidiary owned 51% by SoftBank Corp. and 49% by SoftBank Group Corp.
- The venture targets 10 gigawatts of AI compute capacity by 2030, leveraging SoftBank's access to gas-fired power plants for energy security.
- Infrastructure will run on the proprietary Infrinia AI Cloud OS, which supports multi-tenant Kubernetes and API-based inference for large language models.
- SoftBank aims to triple or quadruple annual operating income to between $18.5 billion and $25 billion through this infrastructure expansion.
- The company is negotiating a $10 billion loan backed by its stake in OpenAI, providing a corporate guarantee to secure the financing.
Why It Matters
SoftBank’s entry into the U.S. neocloud market directly challenges specialized providers like CoreWeave and traditional hyperscalers by addressing the critical bottleneck of power availability. By integrating its proprietary Infrinia software stack with gigawatt-scale energy assets, SoftBank is positioning itself as an essential infrastructure layer for AI-heavy streaming and inference workloads. For the streaming ecosystem, this adds significant liquidity to the GPU rental market, potentially lowering the cost of real-time AI video processing and metadata generation. Watch for SB Neo's first major customer contract with OpenAI, which could validate the platform's suitability for massive-scale training and solidify SoftBank's influence over the generative AI value chain.
Additional Context
The neocloud sector has evolved from a niche category into a major infrastructure segment, with market revenues reaching $35.2 billion in early 2026, according to Mordor Intelligence. This growth is driven by the 'capacity crunch' faced by traditional hyperscalers, as companies like Microsoft and Google struggle to build data centers fast enough to meet AI demand. Per Synergy Research Group in April 2026, neocloud revenues grew by over 200% year-over-year, with the market projected to approach $400 billion by 2031. Specialized providers such as CoreWeave and Lambda have capitalized on this by offering faster access to the latest NVIDIA Blackwell and Hopper GPUs.
SoftBank’s expansion follows its broader 'Activate AI for Society' strategy, which includes a commercial GPU cloud launch in Japan scheduled for October 2026. Per Capacity, May 2026, this Japanese service utilizes NVIDIA GB200 NVL72 hardware to offer sovereign AI infrastructure, ensuring data remains within local jurisdiction—a model SoftBank appears ready to export to the U.S. via SB Neo. To fund these massive capital requirements, SoftBank has pursued aggressive liquidations and margin loans. Per Reuters and TechFundingNews in July 2026, the company added a corporate guarantee to its $10 billion OpenAI-backed loan to overcome lender hesitation regarding the valuation of private equity collateral, reflecting the high stakes of Masayoshi Son's infrastructure-first pivot.
Read full article at siliconangle.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source