SoftBank wants to build data centers with robots — and maybe IPO at $100B

SoftBank is doing that thing where they take a seemingly insane idea and treat it like it’s the most obvious next step. This time: a robotics company that builds data centers. And they’re already floating the idea of a $100 billion IPO.

Let me unpack that because it sounds like a buzzword generator went rogue. The new company, reportedly called “SoftBank Robotics Infrastructure” or something along those lines, will design and deploy robots to construct and maintain data centers. The pitch is that you can’t scale AI infrastructure fast enough with human labor alone — you need machines that can weld, wire, and rack servers at robot speed.

There’s a poetic irony here. AI needs data centers. Data centers need physical construction. So you use AI-powered robots to build the data centers that will run the AI that controls the robots. It’s a closed loop that feels like something out of a sci-fi novel, but SoftBank is betting tens of billions it’s real.

The $100 billion IPO figure is the part that made me blink. That would put it in the same league as Arm, SoftBank’s crown jewel, which went public at $54 billion and has since climbed. But a robotics construction company? At that valuation? They’d need to show revenue that matches the hype, not just a prototype that can carry a server rack across a warehouse.

SoftBank has form here. They threw money at Boston Dynamics, bought Arm, and dumped billions into WeWork. Their hit rate is mixed, but their willingness to swing big is consistent. Masayoshi Son clearly sees robotics and AI infrastructure as the next railroad boom, and he wants to own the tracks.

What’s less clear is whether there’s a market for a dedicated robotics construction firm. General contractors already use automation — drones for site surveys, autonomous vehicles for material transport. SoftBank’s bet is that they can integrate all of that into a single, proprietary system that’s faster and cheaper than what’s cobbled together today. Maybe they’re right. But the history of robotics in construction is littered with demos that never scaled.

The IPO timeline is vague — “in the coming years” — which is SoftBank-speak for “we’ll see how the market feels.” If AI infrastructure spending keeps accelerating, the timing could be perfect. If the hype cycle cools, that $100B number will look like a fever dream.

Either way, I’m watching this one closely. Not because I think it’ll definitely work, but because the circular logic is so audacious it might just break through. And if it does, we’ll look back and say “of course they did that.”