SoftBank Group is preparing the most audacious robotics gambit since the Vision Fund era. According to a Financial Times report on April 29, 2026, the Japanese conglomerate is creating a new artificial intelligence and robotics company called Roze that will deploy autonomous robots to build U.S. data centers – and is already targeting a $100 billion IPO valuation as soon as the second half of 2026. The move, confirmed in subsequent TechCrunch and Japan Times reports, would make Roze one of the largest pre-revenue listings in technology history if it lands at the targeted valuation.
The announcement collides with a brutal reality: the United States is short roughly 439,000 construction workers as of November 2025, according to the Information Technology and Innovation Foundation, and analytics group satellite imagery suggests up to 40% of AI data center construction sites face delays. Roze is Masayoshi Son’s bet that the labor wall blocking the global AI buildout can be broken by machines instead of people. Whether the math works at $100 billion – when the company has not yet shipped a single robot or completed a single data center – is the question now consuming Wall Street.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Roze AI Announcement: What SoftBank Just Revealed
The Financial Times broke the story on April 29, 2026, citing people familiar with the matter. Roze – sometimes referred to as Roze AI – will be a U.S.-incorporated, U.S.-listed company spun out of SoftBank Group, dedicated to automating the physical construction of large-scale data centers using autonomous robots. The Wall Street Journal reported the new entity will focus specifically on making American data center construction more efficient, while Reuters and Japan Times confirmed the planned U.S. listing target and the second-half 2026 timeline.
SoftBank, when contacted by TechCrunch on April 29, declined to comment. That silence is unusual for a story this large, but the lack of a denial – combined with multiple converging reports from FT, WSJ, Reuters, and Japan Times within a 24-hour window – indicates the underlying decision has been made internally even if the public structure has not yet been finalized. According to insiders cited by reporters on April 30, the IPO timing carries real risk: if regulatory or market conditions deteriorate, the listing could slip to 2027, and several SoftBank executives are described as skeptical of the $100 billion valuation target.
The strategic logic is clear even if the execution is unproven. SoftBank already controls roughly 11% of OpenAI after completing a $41 billion investment in December 2025, and on April 1, 2026, executed the first $10 billion tranche of an additional $30 billion follow-on. The Vision Fund 2 segment posted a $2.4 billion gain in the December 2025 quarter, with $4.2 billion of that increase coming from the rising mark on the OpenAI stake. By creating Roze, Son is attempting to capture not just the value of AI software through OpenAI, but the value of the physical infrastructure that AI software will run on – and to do it through a U.S.-listed vehicle that can tap American capital markets directly.
Inside Roze’s Robot-Built Data Center Vision
The technical core of Roze is the use of autonomous robots to perform tasks currently handled by skilled human trades – the electricians, pipe fitters, HVAC technicians, and structural workers whose absence is choking the industry. While SoftBank has not published a technical roadmap, the company’s reported focus on server farm assembly and the integration of robotics with large-scale data center infrastructure points to several near-term targets: cabinet installation, cable routing, busway assembly, hot/cold aisle containment, structural steel positioning, and pre-fabricated module integration.
The economics of automating these tasks have shifted dramatically over the past 24 months. A modern hyperscale data center site can employ between 1,500 and 5,000 workers at peak, according to industry sources cited by BRG and DataBank. Backlogs for skilled trades now stretch to roughly one year. The ITIF reports that the construction industry is short 439,000 workers, with electricians and pipe fitters identified as the most acute bottleneck. In one regional case, electricians’ union membership doubled to 14,700 since 2018 – yet contractor surveys show 82% still struggle to fill hourly craft roles.
Roze’s pitch, as described by people briefed on the plan, is to insert robotic systems into the construction sequence at points where the labor cost is highest and the variance is lowest – the repetitive, precision-critical tasks that consume the most worker hours per megawatt of capacity built. If Roze can deliver even a 20% reduction in peak headcount and a 90-day shortening of typical 18-to-24-month build cycles, the value proposition to hyperscalers ordering tens of gigawatts of capacity is enormous.
The $100 Billion Valuation: How Wall Street Is Reading It
A $100 billion IPO target for a company that has not shipped a product places Roze in extraordinary company. Among private companies that have actually listed, only a handful – Saudi Aramco, Alibaba, Visa, Facebook – debuted with valuations in that bracket, and all had massive existing revenue. Roze’s valuation, by contrast, would rest on a story: that automating data center construction is the next $1 trillion bottleneck and that SoftBank, through Roze, controls the platform to break it.
Several factors make the number more defensible than it first appears. Big Tech is committing more than $650 billion to AI infrastructure capex in 2026 alone, with Microsoft, Google, Amazon, and Meta accounting for the bulk of the spend. The OpenAI-led Stargate Project, in which SoftBank is a key partner alongside Oracle and MGX, anticipates up to $500 billion of data center investment over four years. Roze’s addressable market is therefore not a niche – it is the largest single capital expenditure category in technology, and one where labor scarcity is a genuine ceiling on growth.
However, the skepticism inside SoftBank is also informed. The Vision Fund’s history is studded with mispriced bets: WeWork, Katerra (the construction tech startup that collapsed in 2021), and Zume (an AI pizza-delivery robotics startup that absorbed hundreds of millions before shutting down in 2023). The Roze valuation will be tested most severely on a single question: how much real revenue can the company show in the 12 months before listing? If the IPO comes in late 2026 with little more than pilot deployments and letters of intent, the $100 billion number will face the same fate as WeWork’s pre-IPO mark.
SoftBank’s 2026 AI Portfolio: How Roze Fits
To understand Roze, it helps to map SoftBank’s existing AI exposure. The conglomerate’s bet on artificial intelligence is now structured around three pillars: an equity position in OpenAI for the application layer, Arm Holdings for the chip layer, and – increasingly – a portfolio of compute and infrastructure investments at the physical layer. Roze is the keystone of that third pillar.
| SoftBank AI Position | Stake / Commitment | 2026 Status | Strategic Role |
|---|---|---|---|
| OpenAI equity | ~11% (post $41B Dec 2025 round) | $10B first tranche of $30B follow-on closed Apr 1, 2026 | Application / model layer |
| Arm Holdings | Majority stake (post 2023 IPO) | Positioned as Nvidia compute alternative | Chip / IP layer |
| Stargate Project | Partner alongside Oracle, MGX | $500B over 4 years targeted | Hyperscale data center capex |
| Vision Fund 2 OpenAI gain | $4.2B (Q4 2025) | Drove $2.4B segment gain | Mark-to-market upside |
| AI Computing segment | Arm, Graphcore, Ampere | $590.7M loss (9 months to Dec 2025) | Operational chip exposure |
| Roze AI (new) | 100% pre-IPO | $100B IPO target H2 2026 | Robotics / data center build |
The AI Computing segment containing Arm, Graphcore, and Ampere reported a $590.7 million loss over the nine months ending December 2025, demonstrating that owning the chip layer is not yet the cash machine SoftBank needs. The OpenAI stake is paper gain, mark-to-market and unrealized. Roze, if it lists at $100 billion, would be SoftBank’s first chance to monetize a portion of its AI thesis through a public market exit since Arm itself.
The U.S. Data Center Labor Crisis Roze Is Targeting
Roze’s commercial premise rests on the depth and durability of the U.S. construction labor shortage. The current data is unambiguous. The Information Technology and Innovation Foundation reports that, as of November 2025, the construction industry was short approximately 439,000 workers, the bulk of them in skilled trades. Total construction unemployment stood at roughly 481,000 in 2025, but BRG analysts note that more than 60% of data center providers report difficulty filling qualified roles, indicating the workers who exist do not match the trades the industry actually needs.
The downstream effect on AI infrastructure is severe. Tom’s Hardware, citing analytics group satellite imagery published April 17, 2026, reported that up to 40% of AI data center construction sites in the U.S. are at risk of delays. A Utility Dive analysis pegged contractor sentiment more bluntly: 60% of contractors reported projects postponed or canceled in the past six months, while 82% reported difficulty filling hourly craft positions. At the same time, 65% of contractors still expect the data center construction market to grow over the next 12 months, against just 8% who expect contraction. Demand and supply are diverging.
| U.S. Data Center Labor Metric | 2025-2026 Figure | Source | Implication for Roze |
|---|---|---|---|
| Construction worker shortage | 439,000 (Nov 2025) | ITIF | Direct addressable bottleneck |
| Total construction unemployment | ~481,000 (2025) | BRG / ABC | Skill mismatch, not raw supply |
| AI data center sites at risk of delay | Up to 40% | Tom’s Hardware / satellite analytics | Hyperscaler urgency |
| Contractors with postponed projects | 60% (past 6 months) | Utility Dive | Margin pressure on builds |
| Contractors with hiring difficulty | 82% (hourly craft roles) | ABC survey | Wage inflation continues |
| Contractors expecting market growth | 65% next 12 months | Utility Dive | Sustained demand for automation |
| Active U.S. data center projects | ~3,000 | BRG | Roze pipeline opportunity |
| Peak workers per mega-site | 1,500 – 5,000 | DataBank / industry | Each robot replaces multiple shifts |
This labor pinch translates into measurable cost. Wages, per diems, relocation, and on-site housing are all rising, and the backlog for trades like electricians and pipe fitters has stretched to nearly a year. For a hyperscaler trying to deliver a gigawatt of capacity in 2027, every quarter of delay is hundreds of millions in deferred AI training revenue. That is the wedge Roze is built to drive.
How Roze Compares to Boston Dynamics, Figure AI, and Agility Robotics
Roze enters a humanoid- and mobile-robotics market that is suddenly very crowded. Boston Dynamics – owned by Hyundai Motor Group – has the longest engineering pedigree, with Atlas now demonstrated in industrial use cases. Figure AI raised at a $39.5 billion valuation in 2025 backed by a Microsoft, Nvidia, and OpenAI consortium, and is deployed in a BMW manufacturing pilot. Agility Robotics has Digit working in Amazon and GXO logistics warehouses. Each has a head start on Roze in terms of fielded units.
What separates Roze is the vertical integration thesis. Boston Dynamics, Figure, and Agility sell or lease robots to customers; Roze, as currently described, will be the customer – building entire data centers as a turnkey service to hyperscalers. That positions Roze less as a robotics-product company and more as a robotics-enabled construction general contractor, competing with the likes of Turner, DPR, and AECOM rather than with Figure or Agility directly. The valuation comparison set is therefore mixed: peers like Figure ($39.5B private) on the robotics side, peers like AECOM ($16B market cap on roughly $16B revenue) on the construction side.
The Stargate connection is decisive. SoftBank’s role in the OpenAI/Oracle/MGX Stargate consortium gives Roze a structural advantage no competing humanoid or construction-tech firm can match: first refusal on tens of gigawatts of build capacity that are already capitalized but waiting on labor. If even a quarter of Stargate’s pipeline routes through Roze, the company will start with implied multi-year backlog before its IPO.
Expert Reaction: What Analysts Are Saying
“SoftBank is essentially trying to verticalize the AI buildout,” said an analyst at a New York-based hedge fund quoted by industry trackers covering the FT story. “If you control compute through Arm, models through OpenAI, and now the physical capacity through Roze, you become indispensable to every hyperscaler regardless of which model wins.”
Skepticism is equally well-sourced. “The lesson of Katerra and WeWork is that capital cannot fix construction by itself,” noted one infrastructure investor cited by reporters covering the announcement. “Roze will need to ship robots that work in unstructured job sites, in heat, in dust, around live electrical work. That is a much harder problem than warehouse logistics.”
Masayoshi Son, SoftBank’s founder, has not personally commented on Roze publicly, but his strategic posture across 2025 and 2026 has been consistent: maximum exposure to AI at every layer of the stack. The company’s December 2025 OpenAI investment, the April 2026 follow-on tranche, and the Roze announcement together represent more than $50 billion of fresh AI commitments in under five months. SoftBank’s own filings describe the OpenAI position as the “core” of its AI strategy, with Roze positioned as the operational arm.
The skeptical view inside SoftBank itself is the most telling. According to coverage cited on April 30, several SoftBank executives have privately questioned whether $100 billion is achievable in the second half of 2026, citing pre-revenue status, IPO market volatility, and the technical risk of robotic construction. The same insiders flagged that a 2027 listing remains a real possibility – a quiet caveat that suggests the public timeline is aspirational rather than committed.
Stock Market Impact: SoftBank, Robotics, and Construction Tech
The Roze announcement landed in a tech market already absorbing news that OpenAI missed internal revenue targets, briefly pressuring AI-linked names through April 28, 2026. The Nasdaq weakened on the day SoftBank’s plan first leaked, and Bloomberg’s Tech segment reported AI-linked stocks falling on the OpenAI report, which was published the same week as the Roze leak. SoftBank Group’s own stock will see its valuation increasingly bifurcated between the parent company and the implied value of a successful Roze listing.
The bullish read for SoftBank shareholders is straightforward: a $100 billion Roze IPO would crystalize a meaningful portion of the company’s AI thesis at public-market prices, providing both liquidity and a benchmark against which the OpenAI mark-to-market can be triangulated. The bearish read is that a failed IPO – either delayed to 2027 or priced significantly below $100 billion – would re-rate the entire SoftBank AI story downward, much as WeWork’s failed IPO did to Vision Fund 1.
Adjacent listed names benefit asymmetrically. Construction-tech and modular-build firms, electrical contractors with national footprints, and U.S.-listed robotics suppliers will all see incremental interest as Roze validates the category. Conversely, traditional general contractors with heavy reliance on craft labor face a potential long-term margin threat if Roze’s pricing succeeds in undercutting human-built data centers.
The Stargate Connection and OpenAI’s Build-Out
The strategic glue holding Roze together is Stargate. Announced in early 2025 and updated through 2026, the Stargate Project is a partnership between OpenAI, Oracle, SoftBank, and Abu Dhabi’s MGX targeting up to $500 billion of AI data center capex over four years. SoftBank is the lead financial partner, OpenAI the lead anchor tenant, Oracle a major cloud and infrastructure operator, and MGX a sovereign capital partner. Stargate’s first sites are already breaking ground in Texas and other states, but the labor crunch is hitting that pipeline as much as anyone’s.
Roze’s emergence resolves an internal tension in the Stargate plan. Without a robotics-enabled construction partner, Stargate’s $500 billion target collides directly with the 439,000-worker shortage and the 40% delay risk on AI sites. With Roze, SoftBank can credibly argue to OpenAI and Oracle that the build pace can be accelerated regardless of trade availability – turning a labor problem into a capital problem, which SoftBank is uniquely positioned to solve.
Whether OpenAI itself will be a Roze customer is the immediate strategic question. Coverage of the announcement strongly implies that Stargate sites are first-line candidates for Roze’s robot crews, but no formal commercial agreement has been disclosed. If OpenAI’s own delivery pressures – including the recent revenue miss – push it toward faster, cheaper builds, signing a multi-gigawatt construction contract with Roze could double as a strategic alignment with its largest equity holder.
Risks and Headwinds: Why Roze Could Miss
Three specific risks dominate the Roze story. First, technical risk: construction sites are vastly more unstructured than warehouses. Boston Dynamics’ Atlas, Figure 02, and Agility’s Digit have all demonstrated controlled-environment competence; none has been validated at scale on an active data center pour. Roze must build, ship, and field-prove an unprecedented system inside roughly 12 to 18 months to support the IPO timeline.
Second, market risk: the IPO window in late 2026 is uncertain. Comments from OpenAI’s CFO Sarah Friar earlier this year and Sam Altman’s reaction to revenue concerns have rattled AI-linked equities. If Nasdaq sentiment turns by Q3 2026, the $100 billion target collapses to a banker’s pitch rather than a market-clearing price. Insiders already concede 2027 is a realistic alternative.
Third, regulatory risk: data center construction is governed by state-level building codes, Occupational Safety and Health Administration rules, and union jurisdictions. Roze’s robots will need to clear inspection regimes that were not designed with autonomous systems in mind. Unionized construction trades – particularly the International Brotherhood of Electrical Workers – have institutional incentive to slow adoption. Tom’s Hardware’s April 17 reporting noted that regulatory friction is already a contributor to the 40% delay statistic; that same friction will apply to Roze’s deployment timeline.
Five Predictions for Roze and SoftBank’s AI Stack
Prediction 1: The IPO slips to 2027 in its current form. The Roze entity needs at least one full year of audited operating data and at least one large-scale anchor contract – most likely a Stargate carve-out – before any underwriter will accept a $100 billion mark. SoftBank insiders citing 2027 as a realistic fallback are signaling, not posturing.
Prediction 2: Roze announces a strategic anchor customer by end of Q3 2026. The most efficient path to validating the valuation is naming a hyperscaler – likely OpenAI via Stargate, possibly Oracle Cloud Infrastructure – as the first multi-gigawatt build partner. Expect this announcement to be timed to maximize the IPO roadshow.
Prediction 3: A robotics acquisition follows within six months. Roze cannot build its robot fleet from a standing start. Expect SoftBank to either acquire a humanoid or mobile-robotics platform outright or to take a major equity stake in one – Figure AI, Agility Robotics, and several lesser-known construction-robotics startups are obvious candidates – to compress the technology timeline.
Prediction 4: A competing offering emerges from a hyperscaler. Amazon, Google, or Microsoft – each with internal robotics programs and existing construction relationships – will likely respond with a vertically integrated alternative inside 12 months, either through partnership or direct investment. The first signs of this counter-move will appear in 2026 capex disclosures.
Prediction 5: SoftBank monetizes a portion of OpenAI to fund Roze. The follow-on $30 billion OpenAI commitment is large; the second and third tranches in July and October 2026 may be partially redirected if Roze’s capital needs accelerate. Watch for any modification to the announced tranche schedule as a leading indicator.
The Bigger Picture: Robotics as the Next $1 Trillion AI Layer
Roze is the first true financial-market expression of an idea that has been circulating in AI strategy circles for two years: that the bottleneck to AI growth is no longer chips, no longer power, but physical capacity to translate capital into operational gigawatts. With labor and tariff costs compounding, the value of any robotic system that can build infrastructure faster than humans accelerates with every quarter.
If Roze lists at anywhere close to $100 billion – even at $50 billion – it will reset the valuation of every adjacent robotics company. Figure AI’s $39.5 billion private mark suddenly looks conservative; Boston Dynamics, embedded inside Hyundai, becomes a sum-of-the-parts trade for shareholders; Agility Robotics’ next round will be priced against a public comparable. The category will not be the same once Roze trades.
The deeper bet, as Masayoshi Son has structured it, is that the AI buildout has at least a decade left to run, and that whoever owns the construction layer captures recurring economics that the chip and model layers cannot. Whether that bet is worth $100 billion before a single robot ships – or whether Roze becomes the next WeWork – will be decided in the next 18 months by ground-truth: by what gets built, by whom, and how fast.
Frequently Asked Questions
What is SoftBank’s Roze AI?
Roze, also referred to as Roze AI, is a new artificial intelligence and robotics company that SoftBank Group plans to create and list in the United States. According to Financial Times, TechCrunch, Reuters, and Japan Times reports dated April 29-30, 2026, Roze will deploy autonomous robots to make U.S. data center construction more efficient, automating tasks currently performed by skilled trades.
Why is Roze targeting a $100 billion IPO valuation?
The $100 billion target reflects SoftBank’s view that automating data center construction addresses a structural bottleneck inside the global AI buildout, where Big Tech is committing more than $650 billion to AI infrastructure capex in 2026 and the construction industry is short approximately 439,000 workers. SoftBank insiders cited by reporters acknowledge the valuation is ambitious and that a 2027 IPO timeline remains possible.
When will Roze go public?
SoftBank executives are eyeing a U.S. IPO as early as the second half of 2026, according to multiple Financial Times-sourced reports. Internal skepticism and standard pre-IPO requirements – audited financials, a multi-quarter operating record, and a major anchor customer – make a slip to 2027 a realistic alternative scenario.
How does Roze relate to OpenAI and Stargate?
SoftBank holds approximately 11% of OpenAI following its $41 billion investment completed in December 2025 and is the lead financial partner in the Stargate Project, the $500 billion data center initiative with OpenAI, Oracle, and MGX. Roze is positioned to accelerate Stargate construction by deploying robots on sites that would otherwise be exposed to the U.S. construction labor shortage.
Who are Roze’s competitors in robotics?
Roze enters a market that includes Boston Dynamics (owned by Hyundai), Figure AI (last valued at $39.5 billion in 2025 with backing from Microsoft, Nvidia, and OpenAI), and Agility Robotics (deployed at Amazon and GXO). Roze is differentiated by its turnkey-construction business model rather than selling robots to third parties, putting it in partial competition with construction giants like AECOM and Turner as well as humanoid robotics firms.
What are the risks to Roze’s plan?
The three main risks are technical (autonomous robots have not yet been validated at scale on active data center construction sites), market (the IPO window in late 2026 is sensitive to AI-stock volatility), and regulatory (state-level building codes, OSHA rules, and union jurisdiction add deployment friction). SoftBank’s own track record – including the failed Katerra construction tech bet and the Zume robotics shutdown in 2023 – adds historical caution.
How big is the U.S. data center labor shortage?
According to the Information Technology and Innovation Foundation, the U.S. construction industry was short approximately 439,000 workers as of November 2025, with skilled trades – electricians, pipe fitters, HVAC technicians – representing the most acute gap. Tom’s Hardware reported on April 17, 2026, citing satellite imagery analysis, that up to 40% of AI data center construction sites are at risk of delay. Industry surveys show 82% of contractors struggle to fill hourly craft roles and 60% have had projects postponed in the past six months.
What is Masayoshi Son’s role in Roze?
Masayoshi Son, SoftBank Group’s founder, is the strategic driver behind the Roze plan, which fits a broader pattern of aggressive AI commitments including the $30 billion follow-on OpenAI investment whose first $10 billion tranche closed April 1, 2026. Son has not personally commented on Roze publicly, and SoftBank declined to comment when contacted by TechCrunch on April 29.
