- SoftBank launched a new company, SB Neo, to provide AI cloud services in the U.S.
- The company says its biggest competitive advantage is access to the SoftBank Group's planned 10GW AI and energy infrastructure
- The move positions SoftBank alongside a growing number of AI infrastructure providers seeking to capitalize on surging demand for GPU compute in the U.S. market
SoftBank entered the increasingly competitive U.S. AI infrastructure market with the launch of SB Neo, a neocloud company which aims to provide GPU compute services for enterprise AI training and inference. Rather than competing solely on access to GPUs, the Japanese technology group believes its biggest differentiator will be access to 10GW AI and energy infrastructure.
"A major advantage is the clear prospect of securing the SoftBank Group's 10-GW-scale energy infrastructure, which enables the business to roll out at an early stage," a SoftBank spokesperson told Fierce Network.
SoftBank Corp. and SoftBank Group Corp. announced earlier this month the formation of the new company, SB Neo, which will begin offering AI cloud services in fiscal year 2027, leveraging the SoftBank Group's planned 10-gigawatt AI and energy infrastructure currently under development in the country. The joint venture will be majority-owned (51%) by SoftBank Corp. and the rest (49%) by SoftBank Group Corp. and will target large U.S. enterprises, including hyperscalers.
The announcement comes as AI infrastructure providers increasingly compete not just for GPUs but also for other resources like power, land and high-capacity networking. Rising demand for AI infrastructure is driving the neocloud market, which is likely to touch $400 billion by 2031, a growth of 58% compound annual growth rate. CoreWeave, Nebius, Lambda, Vultr and Crusoe are some of the prominent players of the U.S. neocloud market.
With the U.S. home to several AI and technology giants, including OpenAI, Anthropic, Meta, Microsoft and Google, the market has emerged as the epicenter of this AI infrastructure race. These companies require massive amounts of GPU capacity for training and inference, creating demand for AI cloud providers.
Gartner expects worldwide AI spending to grow by 47% in 2026 to touch $2.59 trillion, with AI infrastructure, including AI-optimized cloud services, servers, networking and semiconductors, representing the largest segment of spending.
SB Neo to target enterprise AI
SoftBank said SB Neo's primary customers will be major U.S. enterprises, including hyperscalers, reflecting growing demand for dedicated AI compute outside traditional public cloud offerings.
The company declined to disclose details about the underlying infrastructure, including networking technologies and deployment timelines.
"SoftBank Group Corp. and its group companies are already advancing large-scale AI infrastructure projects in the United States, and we will work to fully utilize this infrastructure," said the spokesperson, who declined to reveal the specific details of the projects and milestones.
The company also declined to comment on the networking technologies underpinning the platform, saying details regarding infrastructure equipment remain confidential.
SB Neo enters a rapidly evolving market
SB Neo enters a market that has seen rapid growth over the past two years as enterprises seek alternatives to the traditional hyperscale cloud providers for AI workloads. Companies such as CoreWeave have built businesses around GPU-optimized cloud infrastructure, while hyperscalers continue investing billions of dollars to expand AI capacity.
SoftBank's strategy differs in one important respect. Rather than emphasizing GPU availability alone, the company is highlighting access to large-scale energy infrastructure as a competitive advantage.
That reflects a broader shift across the AI infrastructure industry, where electricity availability has become one of the primary constraints on expansion. Gartner estimates that worldwide data center power demand will rise from 104GW in 2025 to 132GW in 2026. The U.S. grid-based power consumption is likely to grow by 17% by 2030 compared to 2025 levels, says S&P Global.
“With data center power electricity consumption estimated to reach over 1,200TWh by 2030, grid supply will be insufficient to meet the demands of future data center construction, affecting all data center users,” said the Gartner press release.
The launch of SB Neo also builds on SoftBank's AI cloud initiatives in Japan. Since May, SoftBank Corp. has been operating a beta version of its GPU cloud service powered by its Infrinia AI Cloud OS, a software platform designed for AI data centers. The company said experience gained through that deployment will support its U.S. expansion.
While SoftBank has yet to disclose the scale of its initial deployment, the company said it plans to expand capacity in phases toward its longer-term 10GW AI infrastructure vision.
As AI adoption accelerates, access to power is rapidly becoming as crucial as access to GPUs. SoftBank's latest move suggests that in the next phase of the AI infrastructure race, securing electricity may prove to be the industry's most valuable competitive advantage.
Related articles about neoclouds
Everyone wants to be a neocloud now – It’s not that easy
SoftBank launches sovereign neocloud for Japan
Australian neocloud Sharon AI scales VAST Data as its infra OS