The loudest voice questioning orbital data centers right now isn’t a short-seller or a rival rocket company. It’s Masayoshi Son, the SoftBank founder who has bankrolled some of the wildest bets in tech history. According to TechCrunch AI’s Equity podcast, Son told a recent shareholder meeting that building data centers in space won’t meaningfully cut costs and will take far too long, because “in the battle for AI, the next few years will be far more important than what might happen a decade or so from now.”
That’s significant, and not just because the idea is bold. It’s significant because of who’s saying it.
When the king of wild bets calls something a stretch
There’s a popular assumption that orbital data centers are the obvious next frontier. A couple of years ago, the concept would’ve been laughed out of the room. Now VCs and founders are nodding along. So when Son, a man whose pitch deck includes WeWork, steps up to ask “what is the point of data centers in space?”, it tells you the consensus has drifted ahead of the math.
TechCrunch AI’s Kirsten Korosec called the irony out directly. Son has “a long history of making wild bets,” she noted, so him playing skeptic “says something.” Her read: it’s healthy that someone with a high profile is finally asking the question a lot of people have quietly been asking.
Son’s actual objection is practical, not philosophical. Even if the engineering works, and even if the economics eventually pencil out, this isn’t happening for years. The compute crunch is now. Space solves a problem that exists in 2035, not 2026.
Follow the launch business
Here’s where TechCrunch AI’s Sean O’Kane adds the sharpest context. When Elon Musk pitches a “constellation of satellites” that need replacing every few years to form an orbital data center, he’s also “guaranteeing that much more business” for SpaceX’s launch arm.
That matters because SpaceX’s launch dominance leans heavily on its own demand. O’Kane’s estimate: strip out Starlink, and SpaceX drops from roughly 90% of the global launch market to maybe 20-40%. An orbital data center is, conveniently, another reason to launch a lot more hardware.
This is the “talking your own book” problem, and it runs through the whole AI moment. Executives predict the future that happens to be good for their business. That’s not a scandal. It’s just worth remembering every time a founder describes the inevitable next platform.
The neo-cloud gold rush underneath it all
Strip away the orbital drama and you find the real trend: everyone with spare compute is rushing to rent it out. O’Kane joked that “neo-clouds are the new oil,” and the examples are almost absurd. Groq, the chipmaker that just raised $650 million. Allbirds, the shoe company that went bankrupt and reemerged as a neo-cloud provider. SpaceX, which signed its first post-IPO compute-rental deal with a smaller player.
The demand is real because the industry is genuinely compute constrained. The open question, as O’Kane put it, is durability: “how durable is it in the long term.” Renting out spare capacity is a great near-term play. It’s a much shakier foundation for a decade-long business.
What to take from this
- Separate the timeline from the technology. Orbital compute may eventually make sense. It does nothing for the capacity shortage facing AI builders over the next few years.
- Read every infrastructure prediction through the predictor’s incentives. Musk’s vision drives launches. Son’s caution protects SoftBank’s near-term AI positioning. Both can be partly right.
- If you’re betting on the neo-cloud wave, ask whether the provider’s economics survive once the compute shortage eases. Spare-capacity rentals are cyclical.
- Watch who breaks ranks. When a serial mega-bettor like Son turns cautious, it’s often an early signal that a narrative has outrun its fundamentals.
The useful takeaway isn’t that orbital data centers are doomed. It’s that the smartest money in the room just reminded everyone to check the timeline and the incentives before chasing the next frontier. You can find the full conversation on TechCrunch’s Equity podcast.