04/30/2026 | Press release | Distributed by Public on 04/30/2026 18:01
SoftBank Group is preparing a new high-stakes wager on the artificial intelligence boom, with plans to create and list a standalone U.S.-based AI and robotics company that could become one of the year's most closely watched public offerings.
According to a report by the Financial Times, the proposed company, to be called "Roze," would focus on building data centers and deploying robotics systems to automate and accelerate AI infrastructure construction. The venture is reportedly being spearheaded by SoftBank founder Masayoshi Son, who is targeting a valuation of roughly $100 billion and aiming for a listing as early as this year.
If completed, the move would mark another dramatic escalation in Son's attempt to position SoftBank at the center of the global AI infrastructure race, where the battle is increasingly shifting from software models to the physical systems required to power them: data centers, energy networks, advanced chips, and automation.
Register for Tekedia Mini-MBA edition 20 (June 8 - Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab.
The proposed structure is also part of a major shift underway across the technology sector. As AI models grow larger and more compute-intensive, companies are no longer competing solely on algorithms or applications. Increasingly, advantage is being determined by access to electricity, land, cooling systems, semiconductors, and construction capacity. That has transformed AI infrastructure into one of the most capital-intensive industries in the world.
Roze appears designed to capitalize on that transition.
The company would reportedly combine existing SoftBank infrastructure assets with robotics capabilities, including ABB Robotics, which SoftBank agreed to acquire last year. The strategy points to an effort to industrialize AI infrastructure deployment itself, using automation to cut labor costs, speed construction timelines, and reduce bottlenecks that have emerged as hyperscalers race to build capacity.
The move comes when global demand for AI compute has surged far beyond expectations over the past two years, driven by enterprise adoption, generative AI services, and autonomous systems. Major technology firms are collectively expected to spend hundreds of billions of dollars annually on AI-related infrastructure, with shortages already emerging in power supply, networking equipment, and advanced chips.
Masayoshi Son has increasingly framed this moment as a once-in-a-generation technological transition, comparable to the rise of the internet or smartphones. His strategy has evolved accordingly. Once known primarily for venture capital-style investments through the Vision Fund, SoftBank is now concentrating more aggressively on foundational infrastructure tied directly to AI expansion.
The company's involvement in the Stargate initiative illustrates that shift. Earlier last year, SoftBank joined forces with OpenAI, Oracle, and other partners on the planned $500 billion Stargate project aimed at expanding U.S. data center capacity.
SoftBank has also accelerated work on its own facilities, including a large-scale development project in Ohio. The Roze venture would likely serve as an extension of that broader buildout strategy, consolidating infrastructure assets into a dedicated vehicle capable of attracting external capital.
At the same time, the proposed IPO could help address growing investor concerns about SoftBank's financing commitments.
The company has pledged tens of billions of dollars toward AI initiatives, including more than $30 billion tied to OpenAI. Those commitments have fueled questions about leverage, liquidity, and exposure to an industry still struggling to convert rapid adoption into sustainable profitability.
An IPO of Roze could partially ease those concerns by monetizing infrastructure assets while creating a separate publicly traded entity capable of raising its own capital. In effect, SoftBank may be attempting to replicate a familiar playbook from earlier technology cycles: spinning out capital-intensive operations into vehicles investors can value independently.
Still, the scale and timing of the plan underscore the risks involved.
The Financial Times noted that some SoftBank executives view the valuation target and timeline as highly ambitious, particularly given geopolitical uncertainty and market volatility linked to the Middle East conflict. Rising energy prices and supply chain disruptions have complicated infrastructure economics globally, especially for power-hungry data center projects.
Those pressures matter because AI infrastructure economics depend heavily on stable energy access and long-term financing conditions. Higher borrowing costs or prolonged geopolitical instability could materially alter projected returns on large-scale projects.
There is also a deeper question surrounding the broader AI investment cycle itself.
Investors have recently become more cautious about whether current infrastructure spending levels are sustainable, particularly as some leading AI firms struggle to meet aggressive revenue projections. Concerns have grown that the sector could face periods of overcapacity if enterprise monetization fails to keep pace with infrastructure expansion.
Yet Son appears determined to lean further into the cycle rather than retreat from it.
SoftBank's Vision Fund posted a $2.4 billion gain in the December quarter, helped significantly by appreciation tied to OpenAI. That performance has strengthened Son's conviction that AI infrastructure will become one of the defining investment themes of the coming decade.
The proposed Roze listing indicates that SoftBank no longer sees robotics and data centers as adjacent businesses to AI. Instead, they are increasingly being treated as core strategic assets in the race to dominate the next phase of computing.
If the deal scales, Roze could become one of the first major publicly traded companies built specifically around the physical backbone of the AI economy rather than the software layer sitting on top of it.