Questions over Elon Musk’s push for orbital AI data centers have sharpened after SoftBank CEO Masayoshi Son publicly challenged the idea, saying at a recent shareholder meeting that space-based facilities would do little to reduce costs and would arrive too late to matter in the near-term race for artificial intelligence. The criticism comes as SpaceX’s recent IPO and expanding compute-rental business have drawn fresh attention to the company’s broader strategy for meeting surging demand for AI infrastructure.
Backers of the concept argue that shifting computing capacity into orbit has become more plausible as launch technology improves and terrestrial data centers face rising opposition. CNBC reported that SpaceX’s IPO raised $85.7 billion and gave the company more capital to pursue an orbital network tied to its Starlink satellite business and xAI’s computing needs. In January, SpaceX filed with the Federal Communications Commission for a constellation of up to one million satellites intended to underpin an orbital AI data center system. Musk has said solar-powered data centers in space could become more cost-effective than Earth-based facilities within two to three years, while investors such as Bullpen Capital partner Duncan Davidson have argued that the long-term business case could strengthen if launch costs fall and the cost of land, power and water on Earth keeps rising.
Other companies are also exploring similar ideas, suggesting the concept extends beyond SpaceX. Jeff Bezos said in a CNBC interview that building data centers in space is "very realistic," though he questioned whether some of the more aggressive timelines were achievable. Blue Origin has submitted FCC plans for a 51,600-satellite constellation under Project Sunrise, while Google is collaborating with Planet Labs on an orbital initiative known as Project Suncatcher. Startups are also testing pieces of the model: Starcloud has already launched an Nvidia H100 GPU into space on a test satellite, and Rendezvous Robotics says it is developing modular systems to help assemble large-scale orbital infrastructure.
Critics, however, say the central problem remains economic rather than conceptual. Son argued that the next few years matter more than what might happen a decade from now, making orbital data centers a poor answer to current AI capacity shortages. TechCrunch commentators echoed that skepticism, saying the engineering challenge may be significant and the cost burden serious even if the technology works. Sean O’Kane argued that a satellite-based data center would also create more demand for SpaceX launches, giving the company a direct commercial incentive to promote the model. Anthony Ha added that major figures in the debate all have financial interests at stake, noting that SoftBank is heavily invested in terrestrial data center projects while Musk’s vision would benefit SpaceX’s business.
Even some analysts open to the idea stop short of calling it cost-competitive today. Harvard economist Mark Weinzierl said he has not seen evidence that orbital data centers currently beat Earth-based alternatives on price, though he said it is reasonable to imagine a future in which falling launch and satellite costs intersect with rising costs and regulatory pressures on the ground. More than 100 proposed moratoriums on data centers have emerged across local, state and national jurisdictions, CNBC reported, and a Heatmap News poll found seven in 10 Americans would oppose a data center near where they live. For now, the debate centers less on whether space-based computing is technically imaginable than on whether it can become economically viable soon enough to matter in the fast-moving AI market.