SoftBank is turning AI infrastructure into a robotics play | FOMO Daily
11 min read
SoftBank is turning AI infrastructure into a robotics play
SoftBank is reportedly planning a new AI and robotics company called Roze that would focus on building data centres and could seek a U.S. listing at a valuation of up to $100 billion. The plan fits SoftBank’s wider push into physical AI, robotics, data centres and large-scale infrastructure for the next phase of the AI boom.
For the last few years, most people have talked about AI like it lives inside a laptop. The public sees chatbots, image generators, coding tools and agents doing clever things on a screen. But behind every AI product sits a physical machine room somewhere, full of servers, chips, cables, cooling systems and power infrastructure. That is the part most people do not see. The problem is that demand for AI compute has grown so fast that the world now needs more data centres, more energy and more construction capacity than the old system was built to deliver. This is where things change. If SoftBank is really building Roze around robotics and data centre construction, it is making a bet that the next big AI winner may not only be the company with the best model, but the company that can help build the factories of intelligence.
Roze looks like a new kind of infrastructure company
The reported Roze plan is not just another software startup idea. It sounds more like a new kind of industrial AI company, one that sits between robotics, data centres, land, energy and automation. The idea appears to be that robots and AI systems can help build the infrastructure that other AI systems need to run. That may sound circular, but it makes sense. AI needs data centres. Data centres are expensive and slow to build. Robotics could reduce some labour bottlenecks, improve repeatability, speed up parts of construction and help manage highly standardised infrastructure projects. What this really means is that SoftBank may be trying to turn the physical buildout of AI into a scalable business of its own, rather than simply funding other people’s data centre projects from the outside.
A possible $100 billion valuation is the number that grabs attention. It sounds huge because it is huge. A valuation like that would not be based on a small robotics services company. It would require investors to believe Roze could become a core platform for the AI infrastructure boom. That means the market would need to see more than a few robots on building sites. It would need to believe in a full pipeline of projects, partnerships, data centre demand, construction automation, access to capital and some kind of durable advantage. The problem is that the AI infrastructure market is filled with enormous numbers already. Once people start talking about hundreds of billions in data centre spending, a $100 billion company begins to sound possible. But possible is not the same as proven.
SoftBank is connecting the pieces
Latest
Top Picks
The latest industry news, interviews, technologies, and resources.
IDC says many EMEA organisations are struggling to move AI from pilot projects into measurable business value. The next phase will depend on CIOs building stronger ROI models, cleaner data foundations, better governance, cost discipline, compliance readiness and workforce adoption.
This reported move does not come out of nowhere. SoftBank has already been building around AI infrastructure, robotics and compute. It agreed to acquire ABB’s robotics business for $5.375 billion, giving it a serious industrial robotics asset if the deal closes. It also announced a $4 billion deal to acquire DigitalBridge, a digital infrastructure investment firm with exposure to data centres and connectivity. On top of that, SoftBank is part of the Stargate project, which was announced as a massive AI infrastructure effort for OpenAI in the United States. These pieces start to look connected when placed beside the Roze report. SoftBank is not simply buying AI hype. It is trying to own parts of the physical stack that make AI possible.
The Stargate project showed how large the AI infrastructure race has become. The project was announced as a new company intending to invest $500 billion over four years building AI infrastructure for OpenAI in the United States, with $100 billion to begin deploying immediately. Later updates described new data centre sites and a path toward a full 10-gigawatt commitment. Those are not normal technology startup numbers. They are industrial policy numbers. They are energy-grid numbers. They are national-infrastructure numbers. This is why Roze matters. If the world is going to spend at this scale, then the companies that can build, finance, automate and operate that infrastructure may become as important as the companies writing the AI models themselves.
Robots building data centres is not as strange as it sounds
At first, the idea of robots building data centres sounds like science fiction. But a data centre is actually one of the more logical places for construction automation to grow. These buildings are highly engineered, repeatable and built around systems that must be installed with precision. They involve racks, cables, power systems, cooling systems, prefabricated components, concrete, steel and strict quality control. The more standardised the design becomes, the more useful automation can be. The problem is that construction is still one of the harder industries to automate fully. Sites are messy. Weather changes. Local rules differ. Human trades are still essential. So the likely future is not robots replacing all workers overnight. It is robots helping with repeatable, dangerous, heavy or precision tasks while humans manage, supervise and handle the messy parts.
Physical AI is becoming SoftBank’s big theme
SoftBank’s chairman Masayoshi Son has been speaking more about physical AI, which is the idea that artificial intelligence will not stay trapped in screens. It will move into robots, factories, vehicles, logistics, infrastructure and machines that act in the real world. The ABB robotics acquisition fits that theme because ABB’s robotics business gives SoftBank industrial capability rather than just digital exposure. Roze would take the idea one step further by applying robotics directly to the AI infrastructure problem. What this really means is that SoftBank may be trying to join two trends at once. AI needs more physical infrastructure, and robotics needs bigger commercial markets. Data centre construction could become the bridge between them.
SoftBank has never been shy about big bets. That is part of its identity. Sometimes those bets look visionary. Sometimes they look reckless. The company has backed world-changing technology businesses, but it has also had painful misfires. That history matters because Roze, if it moves ahead, would be another big conviction play. A $100 billion IPO target would ask public investors to buy into the next chapter of Son’s AI vision. The risk is that the vision runs ahead of execution. Data centre demand is real, but valuations can still get stretched. Robotics is promising, but deployment can be slow. Construction automation is attractive, but hard. AI infrastructure is booming, but booms can create excess capacity if expectations go too far.
Data centres are becoming the new oil fields
There is a reason investors are obsessed with data centres. They are becoming one of the most important assets of the AI economy. Models need training. Products need inference. Enterprises need secure compute. Governments want national AI capacity. Cloud companies need more space and power. Every part of the AI stack eventually hits the same physical question: where does the computing happen? Data centres answer that question. They are not glamorous to the average user, but they are becoming central to economic power. The old internet needed servers. The AI internet needs far more. This is why SoftBank’s reported move into a data-centre-building robotics company deserves attention. It is not chasing the app layer. It is chasing the engine room.
The biggest challenge in AI infrastructure is not just building the buildings. It is powering them. Data centres need enormous amounts of electricity, and AI data centres are especially hungry because of the dense computing hardware inside them. That means land, grid connections, generation capacity, cooling and local political approval all become part of the business. A robotics company can help build faster, but it cannot magically remove energy constraints. This is where the story gets serious. The future of AI is tied to the future of energy infrastructure. Any company trying to scale data centre construction will have to deal with power access, environmental concerns and local resistance. The AI boom may be digital at the front end, but it is deeply physical at the back end.
Speed is becoming a competitive advantage
In the AI race, time matters. A model company that cannot get enough compute loses ground. A cloud company that cannot bring data centres online fast enough misses demand. A country that cannot build AI infrastructure quickly enough may depend on others. That is why construction speed has become strategic. If Roze can use robotics and automation to shorten build times, reduce errors or improve delivery certainty, it could become valuable very quickly. The problem is that construction timelines are not only slowed by manual labour. They are slowed by permits, power connections, supply chains, chips, transformers, cooling equipment and financing. Robots may help, but they are only one piece of the bottleneck.
A reported U.S. listing as early as 2026 would not only be about publicity. It may also be about financing. AI infrastructure is brutally expensive, and SoftBank has many large commitments across OpenAI, robotics, data centres and related assets. Turning Roze into a public company could give SoftBank a way to raise capital, crystallise value and create a listed vehicle for investors who want direct exposure to AI infrastructure automation. That is a clever financial move if the market is hungry for the theme. It also shifts some of the risk into public markets. What this really means is that Roze could be both an industrial plan and a capital-markets plan.
The Roze story is powerful, but investors will eventually ask hard questions. What exactly will the company own? Which assets will be folded into it? How much revenue does it have? What contracts support the valuation? What margins can robotics-driven construction produce? How dependent is it on SoftBank-related projects? How quickly can it scale? How much debt will sit underneath the structure? These questions matter because hype can get a company to the starting line, but proof keeps it alive after listing. A $100 billion valuation needs more than a good theme. It needs visible demand, strong execution and a believable path to profits.
The construction workforce question is coming
If robots begin playing a larger role in data centre construction, labour questions will follow. Some tasks may become safer and faster. Some jobs may shift toward supervision, programming, maintenance and robotic operations. Some traditional construction roles may face pressure. This is not unique to data centres. It is part of the wider automation story. The important question is whether automation improves productivity while creating new skilled roles, or whether it mainly becomes another way to squeeze labour. Governments and communities will care about this, especially because many AI infrastructure projects are sold with promises of jobs and local economic benefit. If robots build more of the infrastructure, the jobs story may need to be told honestly.
AI infrastructure is becoming a strategic asset. The United States, Japan, Europe, China and the Gulf are all thinking about compute capacity, chips, data sovereignty and energy. If SoftBank creates a U.S.-listed company focused on AI infrastructure robotics, it sits right in the middle of that global competition. Data centres are no longer just private buildings full of servers. They are part of national AI capability. The companies that build them may become strategically important suppliers. This is where Roze could become more than a financial vehicle. It could become part of the supply chain for national AI ambition.
This is bigger than SoftBank
The reported Roze plan is really a sign of where the AI market is heading. The first phase was model hype. The second phase was product rollout. The third phase is infrastructure. Chips, data centres, energy, cooling, networking and robotics are now becoming the centre of the conversation. That does not mean software stops mattering. It means software is no longer enough. The companies that win the next phase may be the ones that connect intelligence to physical capacity. This is why a robotics company building data centres can suddenly become a headline story. The market is starting to understand that AI has a body, not just a brain.
The next step is confirmation and detail. SoftBank has not formally laid out the full Roze plan publicly, and the current story is based on reporting. Investors will want to know what assets are included, who runs the company, how much capital it needs, what contracts it has, and whether the IPO timing is realistic. They will also watch how the ABB robotics acquisition, DigitalBridge deal and Stargate buildout fit into the bigger structure. If the pieces come together, Roze could become one of the clearest examples of the next AI infrastructure wave. If the pieces do not come together, it may become another grand SoftBank idea that sounded bigger than it was.
The real story is the factory behind the model
SoftBank’s reported Roze plan matters because it points to the next layer of the AI boom. The public sees models. Businesses see productivity tools. Investors are now looking at the factories behind it all. Data centres are the factories. Power is the fuel. Chips are the machinery. Robotics may become the construction force. That is the bigger picture. AI is not floating in the cloud. It is being built in steel, concrete, silicon and electricity. If SoftBank can turn that physical bottleneck into a scalable company, Roze could become a serious part of the AI economy. But the promise is massive, and so is the execution risk. The next AI race may be won not only by who thinks faster, but by who builds faster.
Passive investing has already changed the stock market by pushing steady flows into the assets held inside major funds and indexes. Bitcoin ETFs may now be creating a similar structure for BTC, giving it a powerful institutional bid while also making large-scale selling faster and cleaner when macro conditions turn.