SoftBank’s Masayoshi Son is challenging one of Elon Musk’s most audacious AI ideas: building orbital data centers, questioning whether an expensive leap to space makes sense while the race for artificial intelligence is unfolding on Earth.

Early vision: Musk’s orbital data centers

Earlier this year, SpaceX outlined a plan to create “a constellation of a million satellites that operate as orbital data centers” and began hiring engineers to pursue the concept, pitching it as a way to meet AI’s soaring energy and compute demands.¹ Tech leaders such as Amazon founder Jeff Bezos and Google CEO Sundar Pichai have also shown interest in space-based data centers as a path to scaling AI infrastructure.¹

June 23: Son’s shareholder-meeting pushback

At SoftBank’s annual shareholder meeting in late June, Son directly questioned Musk’s vision. “What’s the point? What’s the benefit of building AI data center in space?” he asked, arguing that lower orbital electricity costs would not offset the added complexity and expense.¹

Citing an internal breakdown, Son said electricity is only about 7% of AI infrastructure operating costs, while chips and other expenses make up the remaining 93%.¹ Any saving on power, he warned, would likely be erased by “additional maintenance, networking and latency-related costs” in orbit.¹

Son stressed timing over spectacle: “In the battle for AI, the next few years will be far more important than what might happen a decade or so from now,” he said, adding that SoftBank wants to be a “first-comer in any businesses related to AI” on the ground rather than bet on uncertain space infrastructure.²

Wider skepticism and Musk’s incentives

Commentators note it is “very ironic” that Son is the skeptic, given SoftBank’s “long history of wild bets,” yet they share his doubts about Musk’s orbital hype.² TechCrunch’s Sean O’Kane argued that when Musk talks about replacing satellites every few years to form an “orbital data center,” he is also “guaranteeing that much more business” for SpaceX’s launch and space services.²

The debate underscores a broader tension in AI infrastructure: whether to prioritize immediate, terrestrial capacity or pursue futuristic, capital-intensive space architectures that may take a decade to prove out.