Just a decade ago it seemed like an exaggeration to say that the Internet consumed as much electricity as some countries. Today the situation is even more surprising. Artificial intelligence is multiplying energy demand at an unprecedented speed and large technology companies are competing to build increasingly larger, more powerful and more voracious data centers. Some of them already They consume as much electricity as entire cities. Now an idea has emerged that seems straight out of a science fiction novel: taking those data centers off the Earth and into orbit.
Among the companies interested in exploring this possibility is SpaceX, Elon Musk’s space company, which is involved in various studies and proposals aimed at using Starship’s future mass launch capacity to deploy computing infrastructure in space. This was recently stated by Musk himself on the social network X.
Watch @ElonMusk provide a technical update on SpaceX’s capability to manufacture, launch, and operate AI satellites at scale → https://t.co/PSCyWrNsOg pic.twitter.com/vhtr46uax7
— SpaceX (@SpaceX) June 8, 2026
The question is inevitable: Does it make sense to build a data center in orbit when it is still difficult to do so on Earth? The answer is more complex than it seems. The problem with artificial intelligence is not the algorithms. When we think about artificial intelligence we usually imagine software.
However, behind every chatbot, every image generator or every model capable of answering questions, there are gigantic buildings full of processors. Thousands of chips run day and night performing calculations and each calculation consumes energy. The situation has become so demanding that some companies are negotiating directly with nuclear power plants to guarantee their electricity supply. Experts estimate that data centers could become one of the largest sources of energy demand in the coming decades.
That’s why some engineers have begun to wonder if part of that infrastructure could be moved off-planet. On paper, the idea has several attractions. The first is energy. In orbit, the solar panels receive light practically continuously and without the losses caused by the Earth’s atmosphere. A space data center could have an abundant energy source without competing for electrical resources with homes or industries.
The second benefit is cooling. Terrestrial data centers generate enormous amounts of heat. A significant part of their energy consumption, almost half, is spent simply on preventing equipment from overheating. In it space there is no air to transport the heat, which poses its own challenges, but also allows it to be dissipated through large radiators designed specifically for it.
There is also a potential environmental advantage. Moving part of these facilities to space could reduce pressure on electrical networks, water resources and increasingly scarce land. The problem is the size and here it is where science fiction begins to collide with engineering. Modern data centers are gigantic, the largest is in China and measures as much as 145 football fields.
Even with the advances promised by Starship, launching such a quantity of equipment into space would have an enormous cost. And that’s not counting support structures, energy systems, communications, radiation protection or the mechanisms necessary to carry out repairs. On Earth, when a server fails, a technician can replace it in a matter of minutes. In orbit, the operation becomes considerably more complicated.
There is also another less obvious problem. The Earth’s surface is protected by the planet’s atmosphere and magnetic field. Space computers do not enjoy such protection. Energetic particles from the Sun and deep space can alter data, cause calculation errors, or even damage electronic components.
That is why the systems used in satellites and space probes are usually more robust, but also more expensive and less powerful than their terrestrial equivalents. Build a data center with millions of processors running continuously under these conditions represents an enormous technical challenge. So, does it really pay off?
According to an essay by Sven Bilén and Wangda Zuo from Pennsylvania State University, they consider the idea to be technically interesting, but they warn that we are still a long way from being economically competitive. For both, it is still much cheaper to build a data center on Earth than to launch it into space, however, the equation could change in the coming decades.
If launch costs continue to fall, if the energy demand for artificial intelligence continues to grow, and if future generations of reusable rockets live up to their promises, some specialized applications could end up making economic sense off-planet. So, The interesting thing is that it no longer seems completely impossible. In fact, it’s close to being necessary.