A public spat between tech billionaires Sam Altman and Elon Musk over the weekend has brought renewed attention to the ambitious yet challenging prospect of space-based data centres. The exchange highlighted a significant disconnect between the grand visions presented to public market investors and the more grounded realities articulated by industry experts.
The contention arose when Musk accused Altman of being a 'scammer', to which Altman retorted, "homeboy you’re the one sellling public market investors on short-term space datacenters." This pointed remark by Altman reflects a sentiment shared by many in the space and tech sectors: the widespread deployment of orbital data centres, particularly for tasks like AI inference, is not a near-term reality. Despite bullish analyst reports contributing to SpaceX's multi-trillion-dollar valuation, largely predicated on the potential for such processing power to fuel advanced AI models or act as an 'orbital neocloud', practical hurdles remain substantial.
Conversations with subject-matter experts, ranging from entrepreneurs leading other space data centre start-ups to engineers involved in major tech companies' orbital compute projects, consistently yield the same conclusion. They argue that significant progress in this area is contingent on two major advancements: vastly cheaper rocket launches and the ability to produce high-powered satellites at low cost and in large quantities. Without these foundational shifts, the economic viability of space data centres at scale remains questionable.
Elon Musk's response often centres on the promise of Starship, SpaceX’s colossal new rocket, which is anticipated to undertake its thirteenth test flight as early as 16 July. The theory is that if Starship can achieve frequent, fully reusable flights, the business case for orbital data centres could become much stronger. However, even if SpaceX successfully recovers both stages of the rocket in upcoming tests, achieving truly operational and repeatedly reusable flight is still likely several years away. Furthermore, any capacity for space data centre launches would probably be secondary to SpaceX’s existing commitments to NASA and the ongoing expansion of its Starlink network.
Adding to the scepticism, SpaceX itself acknowledged during its IPO roadshow that Starship might not be fully reusable in the short term, requiring the jettisoning of its second stage during each launch. Such an operational constraint would severely undermine the economic feasibility of frequent, large-scale space data centre deployments. Therefore, Musk's assertion that 'we start flying them next year' is viewed by many as an oversimplification. While launching a single satellite equipped for high-speed data processing next year is plausible, the critical question of launching and manufacturing these at a commercial scale is widely considered a challenge for the 2030s.