A weekend social-media clash between two of the most prominent figures in the tech industry highlighted something experts have been saying for a long time. Orbital data centers will not become a profitable business tomorrow, or even a year from now, no matter what is promised to public investors.
A conflict of financial stakes
Responding to Musk’s accusations of fraud, Sam Altman wrote on X that it is Musk who is “selling public investors the idea of short-term space data centers.” Behind the harsh tone lies not merely personal hostility, but a clash between two business logics, since SpaceX’s two-trillion-dollar valuation is largely based on the promise of launching an entire fleet of orbital computing platforms for artificial-intelligence tasks.
Optimistic analysts call this potential unprecedented, but industry engineers, entrepreneurs from other space startups, and even the Google team developing its own orbital-computing project take a different view. They all arrive at the same conclusion, TechCrunch reports: there will be no serious impact on the market until cheaper rockets appear and it becomes possible to mass-produce high-power satellites at low cost.
The 13th flight and reusability in question
Musk’s response is predictable. He is betting on Starship, whose 13th test flight could take place as early as July 16, and he claims, without specifics, that data-center launches will begin next year. If the team manages to bring the vehicle to the point of repeatable flights, the business model could indeed begin to work.
However, even the successful return of both stages in this flight would not mean an immediate transition to operations. Full reusability is likely still a matter of several years, while Starship’s primary tasks will be obligations to NASA and the deployment of SpaceX’s own Starlink network.
Moreover, during the IPO, the company acknowledged that the second stage will most likely have to be expended on each launch in the near future, which makes the economics of orbital computing openly unprofitable.
Vacuum does not forgive excess heat
Beyond the debate over the cost of putting a kilogram into orbit, there is another, purely physical barrier that calls into question the concept itself. In a vacuum, heat is removed only through radiation, while processors and graphics chips convert 80 to 95 percent of consumed energy into heat.
For a hypothetical one-gigawatt data center, the radiators would have to cover an area ranging from hundreds of thousands of square meters to over one square kilometer, depending on their operating temperature and design. This is orders of magnitude larger than the area occupied by a ground-based data center cooling system. While various concepts for space cooling systems exist, none have yet demonstrated the ability to economically scale such a solution to the scale required for giant data centers.
Altman, it seems, merely voiced what most specialists already know. Orbital computing has a future in the form of narrowly specialized tasks, such as preliminary processing of satellite images or defense needs. Axiom Space test modules for such tasks are already operating on Kepler satellites. Large data centers in orbit remain a question for the 2030s, not for the next launches.
