SoftBank CEO Questions Musk Orbital Data Center Hype

So when a high-profile figure stands up and asks the obvious question, it carries weight. Korosec also flagged the irony directly. SoftBank has a long history of making wild bets, and Son pitch decks have thrown a lot of money at bold ideas. This is the investor behind WeWork now urging caution.

Even if every piece of the orbital data center plan works, the costs to make it work will be very serious, and it is not happening for years and years. It is not a solution to any immediate compute problem.

How orbital data centers are supposed to work

The mechanical pitch is straightforward, and that is part of the appeal. Companies are severely compute constrained. Building enough data centers on Earth is proving challenging for all kinds of reasons, so the argument goes that space might be the release valve.

The execution is anything but simple. As Sean O Kane described it on the podcast, Musk approach involves making a constellation of satellites to form an orbital data center. Those satellites would need to be replaced every few years. That replacement cycle is the catch that turns an engineering project into a permanent logistics operation.

And that is where the incentives get interesting. A constellation that constantly needs replenishing is, in O Kane words, guaranteeing that much more business for the company that launches the satellites. The same firm proposing the orbital data center also happens to own the rockets.

SpaceX wins before the orbital dream arrives

Here is the part Son skepticism does not actually slow down. SpaceX does not need orbital data centers to pay off in order to profit from the AI boom right now.

The longer-term idea, as O Kane laid it out, is audacious: build an AI platform with an addressable market the size of U.S. GDP. But before getting anywhere near that, the company is simply renting out its compute. SpaceX has already struck deals with Google and Anthropic, and it recently signed another, described as its first post-IPO deal, to rent compute to a smaller player. The road is clear and the company keeps driving down it.

That is the broader pattern reshaping the sector. Neo-clouds, as the hosts joked, are the new oil. Anyone with a shot at leasing out spare compute is taking it.

The neo-cloud gold rush

The list of converts is genuinely strange. Groq, the chipmaker that was semi-hollowed out by Nvidia, just raised $650 million and is leaning into renting compute. Allbirds, the shoe company, went into bankruptcy and emerged not selling shoes but as a neo-cloud provider. When a footwear brand pivots to leasing GPUs, you know the demand signal is loud.

What happens over the next 24 to 72 hours

Expect Son comments to keep circulating as the rare high-profile dissent against orbital hype, and watch for whether other large investors echo or push back on him. His shareholder-meeting framing gives skeptics cover to speak up, so more pointed questions about timelines and unit economics are likely in the immediate aftermath.

The near-term tell to watch is deal flow, not satellites. If SpaceX or its rivals announce more compute-rental agreements in the coming days, it confirms O Kane read that the real money is in leasing today, regardless of what orbits tomorrow. Any movement around Groq newly funded expansion or fresh neo-cloud entrants would reinforce that the compute crunch, not the space race, is driving decisions right now.

The durability question is the one nobody has answered. Renting out spare compute is a clear business in the near term. Whether it holds up over the long term, especially for a constellation that has to be rebuilt every few years, is the interesting economic challenge the industry will be arguing about for the next couple of years.

Source: TechCrunch