SoftBank’s New Robotics Arm Wants to Build Data Centers — and Maybe the Biggest IPO Ever

SoftBank is at it again. The Japanese conglomerate that brought us the Vision Fund, the ARM acquisition, and a whole lot of WeWork drama is spinning up a new robotics company — one that doesn’t just build robots for fun, but to build data centers. And if the rumors are true, they’re already eyeing an IPO that could hit $100 billion.

Let that sink in for a second. A hundred billion dollars. That’s not just big — that’s “bigger than most countries’ GDP” territory. But knowing Masayoshi Son, he’s not thinking small.

Here’s the pitch: You need massive data centers to train and run AI models. Those data centers are complex, expensive, and take forever to build. So why not use robots to build them faster, cheaper, and with fewer humans? It’s a meta play that’s been tried before — Google and Amazon have dabbled in robotics for logistics — but nobody has gone all-in on data center construction robots like this.

The timing makes sense. AI infrastructure is the new oil, and everyone from Microsoft to Meta is scrambling to build more compute capacity. The bottleneck isn’t chips anymore — it’s physical space, power, and construction speed. If SoftBank can automate a chunk of that, they could own a critical piece of the AI supply chain.

But here’s where I get skeptical. Robotics companies are notoriously hard to scale. Boston Dynamics has been around for decades and still hasn’t turned a real profit. And SoftBank’s track record with hardware bets is mixed — they sold Boston Dynamics to Hyundai in 2020 after failing to make it work. So why would this be different?

Maybe because the use case is narrower and more practical. Building data centers is a repetitive, high-value task that doesn’t require human-like dexterity. You can design robots specifically for lifting servers, laying cables, and assembling racks. That’s a much easier problem than building a general-purpose robot that can walk a dog.

And then there’s the IPO. A $100 billion valuation would make this one of the largest tech IPOs in history, rivaling Alibaba and ARM. But that kind of number assumes the company will actually deliver on its promise. Given SoftBank’s history of overhyping and underdelivering (WeWork, anyone?), I’d take that figure with a grain of salt.

Still, I can’t help but be intrigued. If SoftBank pulls this off, they’ll have created a vertically integrated monster — they own ARM chips, they own data center robotics, and they have the capital to build the infrastructure themselves. That’s a scary amount of control over the AI stack.

For now, it’s just a rumor. But knowing Son, he’s probably already planning the next $100 billion bet before this one even launches.