SoftBank Group is building a new company called Roze with a singular, ambitious goal: use artificial intelligence and robotics to construct the data centers the world’s AI ambitions demand. The conglomerate is already targeting a US IPO valuation of around $100 billion, with a listing potentially as early as the second half of 2026.

The move is a signal that the AI infrastructure race has entered a new phase — one where even how you build the physical facilities is being automated.

What Roze Actually Is

Roze is not a software company. It is a physical infrastructure play, designed to tackle the bottleneck that is slowing AI expansion faster than model performance: the inability to build data centers quickly enough.

SoftBank founder Masayoshi Son is driving the effort, and the company is being assembled from a collection of strategic acquisitions. The planned assets include:

• ABB Robotics — SoftBank’s $5.4 billion purchase of the global robotics division, one of the world’s leading suppliers of industrial automation
• Ampere Computing — SoftBank’s $6.5 billion ARM-derived processor business for cloud infrastructure
• DigitalBridge assets — Infrastructure and land holdings from SoftBank’s $3 billion stake
• Energy, land, and existing infrastructure assets from SoftBank’s broader portfolio

The central idea is straightforward: if you control the robots, the chips, and the land, you can build AI data centers faster and more cheaply than anyone else. Roze is betting that robotics-led construction becomes the competitive differentiator in the data center arms race.

KPMG has been hired to prepare the financial documentation for the public offering, and SoftBank is planning an analyst day at a Texas data center site in July 2026 — a show-and-tell designed to build investor confidence before the IPO.

Why This Matters Now

Big Tech is on course to spend over $700 billion on AI infrastructure in 2026. That spending is straining power grids, supply chains, and the physical capacity to build facilities fast enough. Google, Microsoft, Meta, and Amazon are all hitting construction timelines as the limiting factor in AI expansion.

Roze’s pitch is that robotic construction can compress timelines and costs. Instead of a 3-5 year data center build, the target is something significantly shorter. ABB Robotics brings the automation hardware. Ampere brings the compute to run it. DigitalBridge provides the real estate footprint.

Masayoshi Son has staked SoftBank’s recent trajectory on AI — most visibly through a $30+ billion commitment to OpenAI and the $500 billion Stargate infrastructure fund co-announced with US government backing. Roze is the next logical step: rather than funding other people’s AI infrastructure, SoftBank wants to own the infrastructure layer outright.

The IPO Math (and the Skepticism)

A $100 billion valuation for a company being assembled from parts, without a commercial track record as a unified entity, is aggressive by most standards. Some SoftBank executives have reportedly described the targets as ambitious, and the IPO timeline could slip to 2027 depending on market conditions and how quickly the business units can be integrated.

The Financial Times first reported the plans on April 29, and the story was quickly confirmed by CNBC, TechCrunch, and Benzinga. The planned July analyst day in Texas suggests SoftBank wants to demonstrate the thesis — not just announce it — before approaching public markets.

What This Means for Business

For most companies, the SoftBank Roze story is not directly actionable — very few businesses will be building their own robotics-led data centers. But the broader signal matters for anyone planning around AI.

Infrastructure is the new constraint. The reason hyperscalers are spending $700 billion on AI infrastructure this year is not speculative — it is because AI demand is already outpacing supply. Data center availability, power access, and GPU allocation are already limiting factors for many mid-market AI deployments.

AI is becoming utility infrastructure. When companies with SoftBank’s resources start treating AI data center construction as a physical infrastructure problem requiring industrial robotics, it signals that AI is transitioning from an experimental technology to the foundational layer of the next economy. The same way electricity generation and broadband networks became preconditions for doing business in the 20th century, AI compute access is becoming a precondition in this one.

The window for building AI capability is open, but it narrows. As AI infrastructure scales, the advantage shifts to those who built organizational AI capability early. The businesses that will benefit most from cheaper, faster AI infrastructure are the ones already running AI-assisted workflows — not the ones waiting for the technology to mature.

For business leaders trying to figure out how to approach AI: the SoftBank-scale bet is a datapoint about conviction. The question is not whether AI will be pervasive — that is settled. The question is whether your business will have the internal capability to use it when the infrastructure catches up.

---

The practical next step is the free Working With Claude field guide. Thirty-two pages covering the ecosystem, Claude Code, and how to govern a rollout properly. Get your copy.