Table of Contents
- The Announcement
- What SoftBank Roze Actually Is
- The Real Problem Roze Solves
- The ABB Robotics Foundation
- The Stargate Connection
- Why $100 Billion for a Pre-Revenue Company
- The Bull Case
- The Bear Case
- What This Means for the AI Industry
- FAQ
The AI industry has a bottleneck that nobody talks about enough. It is not chips. It is not power. It is the fact that there are not enough human beings on Earth willing to build data centers fast enough. SoftBank Roze, Masayoshi Son’s newest venture, is a direct bet that autonomous robots can solve that problem, and Son wants public markets to value that bet at $100 billion before the company earns a single dollar of revenue.
On April 29, 2026, the Financial Times reported that SoftBank Group is preparing to spin out and list a new AI and robotics company called Roze in the United States. The target valuation: up to $100 billion. The timeline: as early as the second half of 2026. The product: autonomous robots that construct AI data centers. If Roze reaches its target, it would be one of the largest pre-revenue technology listings in history.
This is either Masayoshi Son’s most brilliant move since his early bet on Alibaba or his most reckless since WeWork. The answer depends entirely on whether you believe the construction labor shortage is the true constraint on AI scaling.
What SoftBank Roze Actually Is
Roze is a U.S. incorporated, U.S. listed company spun out of SoftBank Group. Its core mission is automating the physical construction of large scale data centers using autonomous robots. The company will bundle several assets from SoftBank’s portfolio:
ABB Robotics. SoftBank agreed to acquire ABB’s robotics division in October 2025 for $5.375 billion. ABB Robotics generated $2.3 billion in revenue in 2024 with a workforce of approximately 7,000 employees. The deal is expected to close in mid to late 2026, pending regulatory approvals in the EU, China, and the United States.
Existing robotics investments. SoftBank’s portfolio includes SoftBank Robotics Group, Berkshire Grey, AutoStore Holdings, Agile Robots, and Skild AI. These assets contribute expertise in warehouse automation, industrial manipulation, and AI driven motion planning.
Infrastructure assets. Roze is expected to bundle existing energy, land, and data center infrastructure from SoftBank’s broader portfolio into the new entity.
KPMG has been hired to prepare the financials and documentation for the public offering. SoftBank is planning an analyst day at a Texas data center in July 2026 to generate investor interest.
The Real Problem Roze Solves
Here is the number that makes SoftBank Roze make sense: 439,000. That is the construction worker shortage in the United States as of November 2025, according to the Information Technology and Innovation Foundation. Most of these missing workers are skilled positions: electricians, pipefitters, HVAC technicians, high voltage commissioning engineers.
The scale of the crisis is staggering. Over 400 data centers are currently under development across the United States. Amazon, Google, Microsoft, and Meta collectively plan to spend $700 billion on AI infrastructure in 2026 alone. Peak crew sizes that once maxed out at 750 workers per site now require 4,000 to 5,000 workers. The result: 45% of construction firms report labor shortage related project delays, and Sightline Climate projects that 30% to 50% of the 2026 data center pipeline will not materialize on schedule.
The hardest roles to fill tell the story: journeyman electricians, mechanical pipefitters, project managers with hyperscale experience, and high voltage commissioning engineers. You cannot train these workers in six months. Many require years of apprenticeship. And every hyperscaler is competing for the same finite pool.
From where I sit as someone managing enterprise infrastructure at a telecom, this is the constraint I hear about in every vendor conversation. The chips ship. The power contracts get signed. Then the construction timeline slips by eight to fourteen months because the electrical contractor cannot staff the site.
The ABB Robotics Foundation
The ABB Robotics acquisition is what separates Roze from a PowerPoint pitch. ABB is not a startup. It is one of the world’s leading suppliers of industrial robotics and machine automation, with decades of deployment history across automotive manufacturing, electronics assembly, and logistics.
ABB Robotics brought in $2.3 billion in revenue in 2024 with a 12.1% operational EBITA margin. Its roughly 7,000 person workforce includes robotics engineers, AI researchers, and field deployment teams with experience installing and maintaining industrial automation systems at scale.
The thesis is straightforward: take ABB’s proven industrial robotics platform, integrate it with AI capabilities from SoftBank’s broader portfolio (including its OpenAI relationship), and deploy autonomous construction systems purpose built for data center buildouts. Cable routing, conduit installation, repetitive structural assembly, site inspection, and materials transport are all candidates for robotic automation.
This is not theoretical. Autonomous systems already handle many of these tasks in automotive factories. The challenge is adapting them to construction environments, which are far less structured than factory floors.
The Stargate Connection
Roze does not exist in isolation. SoftBank is the chairman entity of the Stargate project, the $500 billion joint venture with OpenAI and Oracle to build AI data center capacity across the United States. Five new Stargate sites were announced earlier in 2026.
The math connects directly. If Stargate alone requires hundreds of billions in construction spending, and 30% to 50% of that construction is at risk of delay due to labor shortages, then the entity that solves the labor problem captures enormous value. Son is effectively building both the demand side (Stargate) and the supply side (Roze) of the AI infrastructure equation.
SoftBank’s own filings describe the OpenAI position as the “core” of its AI strategy, with Roze positioned as the operational arm. This vertical integration is the real play. Son is not just investing in AI companies. He is building the physical supply chain that those companies depend on.
Why $100 Billion for a Pre-Revenue Company
The $100 billion target valuation deserves scrutiny. Roze has not shipped a product. It has not generated revenue under its own name. Even including ABB Robotics’ $2.3 billion in annual revenue, a $100 billion valuation implies a price to sales ratio north of 40x, comparable to the most aggressively valued software companies at their peaks.
SoftBank’s pitch to investors will likely rest on three pillars:
Total addressable market. If hyperscalers spend $700 billion on infrastructure in 2026 and construction labor represents 25% to 35% of project costs, the addressable market for construction automation exceeds $175 billion annually.
Margin profile. Robotics companies that successfully automate manual processes often command gross margins of 60% or higher once deployed at scale, far above traditional construction services.
Strategic lock in. As the Stargate chairman entity, SoftBank can effectively guarantee Roze’s first major customer. If Roze proves viable on Stargate sites, every hyperscaler facing the same labor shortage becomes a potential buyer.
The Bull Case
The bull case for SoftBank Roze comes down to timing and necessity. The construction labor shortage is not a cyclical problem. It is structural. The average age of a skilled electrician in the United States is climbing. Apprenticeship programs are not scaling fast enough. And the demand curve for data centers shows no sign of flattening.
If Roze can demonstrate even a 20% to 30% reduction in construction timelines on pilot projects, the value proposition becomes compelling. A hyperscaler spending $10 billion on a data center campus that comes online three months earlier captures hundreds of millions in accelerated revenue. At that point, the robotics system pays for itself in weeks.
The McKinsey projection that global AI capex will require $6.7 trillion by 2030 makes the long term demand picture clear. Someone has to build all of that. If it cannot be done with humans alone, robotics is not optional. It is infrastructure.
The Bear Case
Several SoftBank executives have privately questioned whether the $100 billion valuation is achievable. Their concerns are legitimate.
Pre-revenue risk. No amount of addressable market analysis changes the fact that Roze has not demonstrated autonomous data center construction at commercial scale. The gap between industrial robotics in a factory and autonomous construction on an active site is enormous. Factories are controlled environments with fixed layouts. Construction sites are chaotic, weather dependent, and filled with unpredictable variables.
Regulatory uncertainty. Construction codes, union agreements, and safety regulations vary by jurisdiction. Deploying autonomous construction robots at scale will require navigating a complex regulatory landscape that does not yet have frameworks for this use case.
IPO market conditions. The broader geopolitical environment, including tensions related to U.S. tariff policy and Middle Eastern conflicts, creates uncertainty for large IPOs. A $100 billion listing requires massive institutional demand, and risk appetite can shift quickly.
Execution timeline. The ABB Robotics acquisition has not closed yet. Integrating a 7,000 person robotics division, adapting its products for construction use cases, and deploying commercially viable systems is a multi-year effort. The H2 2026 IPO timeline is ambitious.
What This Means for the AI Industry
Regardless of whether Roze hits its $100 billion valuation target, the venture signals an important shift in how the AI industry thinks about scaling. For the past three years, the conversation has centered on chips and models. Who has the best GPU? Which model has the highest benchmark scores? How many tokens per second?
Roze is a recognition that the AI scaling problem is becoming a physical construction problem. You can design the most efficient chip in the world, but if you cannot build the building to put it in, the chip sits in a warehouse.
This is why the Big Tech Q1 earnings were so revealing. Alphabet, Microsoft, Meta, and Amazon collectively reported over $130 billion in Q1 capital expenditures and projected up to $725 billion for the full year. That money has to become concrete, steel, copper wire, and cooling systems. And right now, there are not enough workers to make that conversion happen on schedule.
For enterprise buyers evaluating AI strategy, Roze is worth watching for a practical reason: if robotic construction accelerates data center buildouts, it accelerates the timeline for AI capacity expansion. That means faster access to inference compute, lower latency for real time applications, and more competitive pricing as supply catches up with demand.
For investors, Roze represents the latest test of how far public markets will go in valuing AI adjacency. A $100 billion IPO for a pre-revenue construction robotics company would set a new benchmark for what the market considers “AI infrastructure.”
Masayoshi Son has been both spectacularly right and spectacularly wrong before. With Roze, he is betting that the most valuable company in AI will not be the one that builds the best model. It will be the one that builds the buildings where those models run.
FAQ
What is SoftBank Roze?
SoftBank Roze is a new U.S. incorporated company being spun out of SoftBank Group. It will deploy autonomous robots to construct AI data centers, bundling ABB Robotics, existing infrastructure assets, and AI capabilities from SoftBank’s portfolio. The company is targeting a $100 billion IPO valuation as early as the second half of 2026.
Why is SoftBank creating a company to build data centers with robots?
The United States faces a 439,000 person construction worker shortage, and 30% to 50% of planned data center construction is at risk of delay. With hyperscalers planning to spend $700 billion on AI infrastructure in 2026, the labor bottleneck is the single biggest constraint on AI scaling. Roze aims to solve this with autonomous construction robots.
How does Roze connect to the Stargate project?
SoftBank is the chairman entity of Stargate, the $500 billion joint venture with OpenAI and Oracle to build AI data centers across the United States. Roze is positioned as the construction arm that could accelerate Stargate buildouts while also serving other hyperscaler clients facing the same labor shortages.
Is a $100 billion valuation realistic for a pre-revenue company?
The valuation is aggressive. Even including ABB Robotics’ $2.3 billion in annual revenue, the implied price to sales ratio exceeds 40x. Bulls point to a $175 billion plus addressable market and strategic lock in through Stargate. Bears note that autonomous construction at commercial scale remains unproven, and several SoftBank executives have privately questioned the timeline.
When will the SoftBank Roze IPO happen?
SoftBank executives are targeting the second half of 2026, with an analyst day planned at a Texas data center in July. However, the timeline could slip to 2027 depending on regulatory approvals for the ABB Robotics acquisition, market conditions, and geopolitical factors.