SoftBank Group is preparing to spin out a new AI and robotics company called Roze, targeting a valuation of up to $100 billion in what would rank among the largest AI-related IPOs to date. The Financial Times reported the venture will focus on the physical buildout of AI infrastructure, using robotics to help construct data centers and bundling together SoftBank’s existing investments in energy, land, and digital infrastructure. Chief Executive Masayoshi Son is doubling down on what he calls physical AI as the next frontier, moving beyond software and model investment into the actual machinery of building and operating the facilities that AI requires.
The IPO could come as early as the second half of 2026, part of a broader effort by SoftBank to capitalize on surging investor demand for AI infrastructure while managing its massive financial commitments. SoftBank has already committed tens of billions to OpenAI and to the Stargate AI infrastructure initiative in the United States. Roze would represent a strategic pivot from pure capital deployment toward operational control of the physical infrastructure layer, giving SoftBank a direct role in constructing the facilities its portfolio companies depend on.
What Roze Would Actually Do
Roze is structured to address one of the most acute bottlenecks in AI infrastructure development. Specifically, it targets the shortage of skilled construction labor capable of building hyperscale data centers at the speed and scale the industry requires. By deploying robotics across data center construction sites, Roze aims to compress build timelines that currently stretch to three years or more from site selection to operation. Additionally, the company would bundle SoftBank’s existing land holdings and energy investments into a single platform. As a result, it creates a vertically integrated AI infrastructure developer that controls inputs from land acquisition and energy procurement through to facility construction and operation.
Why It Matters for the Infrastructure Market
The Roze announcement reflects a broader shift in how the most sophisticated AI infrastructure investors are thinking about competitive advantage. As our analysis of the GPU-as-a-service neocloud model has shown, the operators who control the full stack from physical infrastructure to compute delivery are building moats that pure capital cannot replicate. Furthermore, SoftBank’s move to create a robotics-powered construction company specifically for data centers signals that the physical construction bottleneck is now significant enough to warrant a dedicated billion-dollar solution. Consequently, the Roze spinout positions SoftBank not just as an investor in AI infrastructure but as an operator at every layer of it.
