Musk Is Serious About Space-Based Data Centers!

On Monday, February 2, Elon Musk officially merged his space company SpaceX with the artificial intelligence company xAI. This move makes much more sense in light of the ambitious plan that SpaceX submitted to the Federal Communications Commission (FCC) – to build a network of one million satellite data centers in Earth's orbit.

On the Friday before the merger, SpaceX filed an application with the FCC for permission to launch this massive constellation. While many thought it was just another one of Musk's provocations, the following week showed that he was deadly serious. The FCC accepted the application and set a timetable for public comments. Commission Chairman Brendan Carr even took the unusual step of sharing the filing on the social network X, suggesting that the project has political support.

On Wednesday, Musk laid out his case for orbital data centers in a new episode of the "Cheeky Pint" podcast, hosted by John Collison. Dwarkesh Patel also appeared as a guest. Musk explained the basic economic logic: solar panels in space produce approximately five times more energy than on Earth, which should reduce one of the main operating costs of data centers. "It's harder to scale on Earth than in space," Musk said on the podcast. "Any solar panel will give you about five times more energy in space than on Earth, so it is actually much cheaper to do it in space."

Musk identified 2028 as the tipping point for orbital data centers. "Mark my words, in 36 months, but probably closer to 30 months, space will be the most economically advantageous place to put AI," he declared. He went even further with his prediction: "In five years, we will be launching and operating more AI in space every year than the cumulative total on Earth."

The technical obstacles are enormous

However, experts are expressing considerable doubts. Alexander Wyglinski, a professor of electrical engineering at WPI, points to the costs: transporting each kilogram of material into orbit costs approximately CZK 25,000 – which means more than CZK 22.5 million per metric ton. "The cost of transporting all the construction materials and components into orbit, followed by building the platform in space, would be quite high, even if it were transported in modules," he said.

Olivier de Weck of MIT emphasizes another fundamental problem: electronics must be radiation-hardened or heavily shielded to protect them from cosmic radiation, which can cause "bit flips" – literally changing ones to zeros in a chip and vice versa. No one yet knows how long GPUs can survive in space.

Another enormous problem is heat dissipation. There is no atmosphere in space, which means there is no air to carry heat away from sensitive electronic components. "All those megawatts of heat must be radiated back into deep space," de Weck explained. Despite what movies suggest, space is not cold – it cycles between extremes depending on whether there is direct sunlight.

According to Wyglinski and de Weck, other technical obstacles include avoiding space debris (to which another million satellites will contribute), cloud-free ground stations capable of receiving laser-transmitted data, and scaling up the capacity to generate energy in space by a factor of one hundred. For context: by 2030, global data center capacity is estimated to reach 200 GW, representing infrastructure worth approximately CZK 25 trillion when located solely on Earth.

Potential benefits do exist

If Musk managed to overcome these obstacles, placing a data center in space would bring significant benefits. "They are physically secure against intrusion and environmentally friendly once operational," de Weck said. "Essentially, the three primary resources needed on Earth – land, energy, and cooling – are available in space 'for free' once the initial launch and deployment costs have been covered."

Chad Anderson, founder and CEO of Space Capital, an investment company focused on the commercial space industry, considers orbital data centers to be overhyped. "The narrative is compelling, and in the long term it is probably inevitable, but the physics is unforgiving," he said. "What we will see in 2026 will be pilot projects and proofs of concept rather than industrial-scale deployments."

SpaceX makes money by launching things into orbit, so the entire project is highly advantageous for Musk – especially now that SpaceX has an AI company attached to it. With the planned IPO (initial public offering) of the new SpaceX-xAI conglomerate just a few months away, you can expect to hear much more about orbital data centers. With technology companies continuing to invest hundreds of billions of Czech crowns annually in data center spending, there is a real chance that not all of the money will remain on Earth.

Sources: techcrunch.com and forbes.com