Orbital data centers: SoftBank's Son questions Musk's plan

Masayoshi Son has publicly questioned the value of building data centers in space, telling SoftBank shareholders that such projects "won't do much to cut costs and will take too long" and that "in the battle for AI, the next few years will be far more important than what might happen a decade or so from now." The remark came up on an episode of the Equity podcast, where hosts Kirsten Korosec, Sean O'Kane and Anthony Ha discussed Son's comments alongside reporting on OpenAI's plans for custom chips and chipmaker Groq's new $650 million funding.

What did Masayoshi Son actually say about orbital data centers?

Son argued at a recent shareholder meeting that building data centers in space will not reduce costs meaningfully and will take too long to address urgent AI compute needs. He framed the timeline as critical, saying the next few years matter far more than developments "a decade or so from now," implying orbital projects are not a solution for the immediate compute bottleneck facing AI companies.

The hosts flagged the irony that SoftBank, a group known for big, risky investments, is voicing skepticism. Kirsten Korosec noted SoftBank's "long history of wild bets" and flagged that Son asking these questions carries weight given his profile.

How do other voices in the conversation view SpaceX's orbital claims?

Podcast discussion pointed at industry incentives and business alignment behind orbital data center talk. Sean O'Kane argued that building a satellite constellation to act as an "orbital data center," with satellites that need replacement every few years, would drive more launch demand and therefore more business for SpaceX. He also said SpaceX's position in launches is heavily tied to Starlink, arguing that Starlink helps push SpaceX to "80 or 90% of the launch market globally," and that removing Starlink would shrink that share to a much lower number, perhaps "20% or 30% or 40%."

Hosts also signaled broader skepticism. Anthony Ha warned that executives tend to "talk their own book," meaning predictions often align with what benefits their companies. The episode noted that Sam Altman has rolled his eyes at the orbital data center idea, and that Elon Musk's public support for space-based compute ties directly into SpaceX's commercial interests.

What else was discussed around the topic?

The podcast placed Son's remarks in a wider industry context, touching on OpenAI's plans for custom chips and Groq's funding round. The episode explicitly referenced Groq's new $650 million funding as part of the broader compute and chip conversation. Hosts debated which players might find a near-term market in leasing compute versus which bets are speculative and long term.

Kirsten suggested some founders and VCs have been swept up in the orbital data center idea in recent years, while Anthony cautioned that everyone in the debate brings financial and strategic baggage. Sean raised a practical point about replacement cycles for satellites, noting that the need to replace hardware every few years complicates any cost argument for orbit-based infrastructure.

Why it matters

Son's public skepticism matters because he is asking whether an expensive, long-range technical bet helps with immediate AI infrastructure shortages. His position highlights a tension between near-term needs for compute capacity and long-term engineering ambitions that may not deliver benefits quickly. The debate also exposes how company incentives, such as SpaceX's launch business and Starlink, shape which technical futures get promoted.

What to watch

Watch for concrete deals and timelines: whether SpaceX or other firms sign repeatable, commercial contracts to rent orbital compute and how they price replacement and launch costs. Also track chip- and data-center investments on Earth, including developments tied to OpenAI's custom chips and follow-on funding in chipmakers like Groq, which recently raised $650 million.