SpaceX Aims for Orbital Data Centers: Heat is the Hurdle

SpaceX unveils AI1, an orbital AI computing satellite. Physics highlights heat, radiation, and repairs as major business challenges.

English · Original discussion in Spanish · Published

Sending Servers to Space: Physics vs. Hype

SpaceX has introduced AI1, a satellite whose sole stated mission is to perform artificial intelligence computing. Not internet, not GPS, not observation. Just processors in orbit. On paper, the idea solves two major problems of the data business on Earth: energy and heat. In practice, the second half of that equation is precisely what space doesn't fix, but exacerbates. And that's where the narrative starts to falter.

Space Isn't Cold, and That Changes Everything

It's the most common misconception: space is very cold, so cooling is free. False. Temperature is a property of matter, and there's very little of it in orbit. An object in sunlight heats up; in shadow, it cools slowly because it can only emit radiation. There's no air to convect heat away.

The consequence is inconvenient: dissipating the heat from an intensive computing rack requires enormous radiators, exposed and vulnerable. The International Space Station suffered leaks in radiator panels in 2009 and 2010 that necessitated repairs. Any orbital debris puncturing one of these panels cripples the system. Meanwhile, solar panels yield just over 1,300 W per square meter, a modest figure when translated into computing watts.

The Economic Equation: 11,000 Satellites and Sky-High Valuations

The skeptical view has a solid argument: SpaceX is unprofitable, and its valuation reflects the narrative, not the numbers. Tesla sells fewer cars than Toyota but is capitalized at triple the amount. With AI1, the suspicion is the usual one: an announcement at the opportune moment to sustain interest and stock price.

The opposing view also has data. SpaceX launches about 90% of the world's space cargo at costs that seemed impossible a decade ago, and operates a constellation of 11,000 satellites with nearly triple that projected. When they proposed reusable rockets, people also laughed. And it's not just Musk: China has invested public funds in orbital computing, shifting the matter from a novelty to a technological race.

The full breakdown of costs, detailing energy and maintenance line by line, yields a figure that is surprising for how much it depends on unresolved variables.

Is an Orbital Data Center Profitable?

There's no consensus here. 24/7 solar power without an electricity bill and background temperatures of about -150 ºC in shadow sound appealing until translated into engineering: cells must be oriented towards the sun, radiators profiled, liquid cooling added, and electronics shielded against high-energy solar radiation, storms, and cosmic rays. And any malfunction isn't fixed by an on-call technician: it requires an astronaut or writing off the equipment.

The projected date is late 2027. Given the sector's track record, that's an optimistic starting point.

The question isn't whether physics allows it. It's whether anyone will pay the bill before the competition does.

Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (60 replies).

More summaries

All summaries in English →

Back