SoftBank’s new robotics company isn’t about robots — it’s about building data centers faster

SoftBank is at it again. This time they’re starting a robotics company that builds data centers. And they’re already eyeing a $100 billion IPO.

Let that sink in for a second. A robotics company that doesn’t make cute delivery bots or warehouse pickers. It makes data centers. Or rather, it builds them.

The logic is simple enough: we need more data centers to run AI workloads, but building them is slow, expensive, and labor-intensive. So why not throw robots at the problem? SoftBank’s new venture, reportedly called “SoftBank Robotics Infrastructure” or something along those lines, aims to automate the physical construction of data centers — digging foundations, laying cables, stacking servers, the whole messy process.

I’ve seen this pattern before. Companies like Built Robotics and Dusty Robotics have been trying to automate construction for years. But they focus on general construction — roads, foundations, drywall layout. SoftBank is going vertical on data centers specifically. That’s a bet I actually respect. Data centers are standardized enough that you can design a robotic workflow that repeats across sites. You don’t need a general-purpose construction bot; you need a specialized one that does one thing really well.

What makes this interesting is the scale. SoftBank isn’t building a small pilot project. They’re reportedly targeting a $100 billion valuation for this spin-off. That’s bigger than most tech IPOs. It’s bigger than Arm’s IPO, which SoftBank already pulled off. They’re betting that data center construction is going to be one of the most capital-intensive industries of the next decade, and they want to own the automation layer.

But here’s where I get skeptical. Building data centers isn’t just about physical labor. It’s about permitting, supply chains, power procurement, and dealing with local utilities. Robots can dig trenches faster than humans, but they can’t negotiate with the county planning board or convince a utility company to upgrade a substation. SoftBank’s robotics play might speed up the construction phase by 20-30%, but the bottlenecks are often elsewhere.

Also, $100 billion IPO? That’s ambitious. Even optimistic. For context, the entire global robotics market is worth maybe $50 billion today. SoftBank is claiming their data center construction robot company alone could be worth twice that. Either they know something we don’t, or this is classic Masayoshi Son overhype. He’s done it before with WeWork and with some of his Vision Fund bets. The man swings big.

I’ll give them this: the timing is right. AI demand is exploding, data center capacity is constrained, and labor shortages in construction are real. If SoftBank can actually deliver a robotic system that cuts data center build times by even 30%, they’ll have no shortage of customers. Hyperscalers like Google, Amazon, and Microsoft are desperate for more compute capacity. They’d pay a premium for speed.

But the execution risk is enormous. Robotics in unstructured outdoor environments is hard. Weather, mud, uneven terrain, coordination with human crews — it’s not a clean factory floor. SoftBank has a mixed track record with robotics. They bought Boston Dynamics, then sold it. They invested in a bunch of robotics startups that went nowhere. This time feels different because they’re building the company themselves, not funding someone else.

I’ll be watching this closely. If they pull it off, it changes how we think about infrastructure. If they don’t, it’s another expensive lesson in how hard it is to automate the physical world. Either way, it’s a bet worth paying attention to.