Sam Altman was sitting before a packed auditorium in New Delhi when the moderator asked him about Musk's plan to place data centers in Earth's orbit. Altman smiled and replied without hesitation: “Honestly, I think the idea of putting data centers in space is ridiculous in the current environment.” The audience burst into laughter. And the tech world immediately split into two camps.
Musk's Dream of 1 Million Satellites
Elon Musk is not someone who thinks small. Last December, at an internal employee meeting at xAI, he publicly unveiled his plan for the first time: a constellation of one million satellites operating as orbital data centers. In February 2026, SpaceX then officially confirmed the plan and began actively recruiting engineers to turn it into reality. Musk also suggested that SpaceX's planned acquisition of xAI would significantly accelerate the entire project.
It sounds like the plot of a Netflix show. But Musk is serious—and there is logic behind it. Ground-based data centers consume enormous amounts of energy, use water for cooling, and are facing growing opposition from residents in affected communities. In the US alone, more than 1,200 new data centers had been approved by the end of 2024—four times as many as in 2010. Meanwhile, opposition from residents in Texas, Oklahoma, and other states is rising like the tide before a storm.
Altman Hit the Brakes: The Math Simply Doesn't Work
Altman did not dismiss Musk's vision forever. He admitted that “one day it will make sense.” But he then presented specific arguments that brought space enthusiasts back down to Earth.
First: the cost of launching payloads into space. Despite the revolution SpaceX brought about with its Falcon 9 and Starship rockets, launches are still astronomically expensive compared with building a ground-based data center powered by cheap electricity. Second—and this may be an even more pressing issue—what do you do when a GPU breaks in orbit? And they do break. Altman himself confirmed it: “Unfortunately, they still break a lot.” Repairing a chip in orbit is not like calling a technician.
“Orbital data centers will not play a role at a scale that matters in this decade,” Altman concluded.
Interestingly, Altman and Musk are not the only ones looking to the stars. In November 2025, Google unveiled Project Suncatcher, an ambitious plan for solar-powered space data centers. CEO Sundar Pichai even suggested that the first such facilities could be in space as early as 2027. That is just a year away. Does Altman's skepticism now look a little like a challenge to the others, or like a sober voice of reason amid the space fever?
It might seem like an academic dispute between two billionaires over the future of technology. But behind it all lies a much bigger question: where and how will we build the infrastructure for the AI that will shape the coming decades of human civilization?
Ground-based data centers are running up against physical limits such as shortages of energy, water, space, and public support. In theory, space removes these limits. The sun shines continuously, the vacuum theoretically provides cooling, and no one will hand you a petition. But the reality is different for now. High launch costs, high hardware failure rates, and the absence of any service infrastructure in orbit make Musk's dream more of a visionary project than a business plan at this point.
Altman himself, meanwhile, is leading the Stargate project, an investment of more than $500 billion in ground-based AI infrastructure in the US. He therefore has a personal interest in ensuring that ground-based data centers remain relevant. That does not mean he is wrong. But the context is worth mentioning.
Sources: tomshardware.com and economictimes.indiatimes.com
