SpaceX Proposes Million-Satellite Orbital Data Center Network After xAI Merger

SpaceX has submitted an ambitious proposal to the U.S. Federal Communications Commission (FCC) to establish a sprawling network of up to one million satellites, each functioning as a data center in orbit. The filing, accepted by the FCC on Wednesday, signals a significant evolution of satellite infrastructure beyond communications, targeting large-scale computational operations for artificial intelligence (AI) applications.

This move follows SpaceX’s formal merger with xAI, Elon Musk’s artificial intelligence venture, completed the same week. By combining SpaceX’s expertise in satellite deployment with xAI’s AI development capabilities, the unified company aims to bridge terrestrial and space-based computing, opening new possibilities for distributed AI workloads.

FCC Chairman Brendan Carr publicized the filing on X (formerly Twitter), inviting public comment on the application. According to the submitted documents, the proposed network would not only support computational needs but is also described by SpaceX as “a first step towards becoming a Kardashev II-level civilization,” referencing a hypothetical stage of advanced planetary energy control.

Elon Musk expanded on the technological and economic motivations for such a network in a recent interview on the "Cheeky Pint" podcast with Stripe co-founder Patrick Collison and Dwarkesh Patel. Musk emphasized the unique advantages of building data centers in orbit, particularly in terms of power generation. Solar panels deployed in space can receive uninterrupted sunlight, without interference from the Earth’s atmosphere or daily cycles, allowing for greater energy efficiency. "Any given solar panel is going to give you about five times more power in space than on the ground," Musk claimed, noting that this could dramatically reduce electricity costs — a major expense for AI data centers, especially during model training processes.

However, Patel pointed out that running orbital data centers presents other logistical challenges, such as cooling, hardware maintenance, and the complexity of servicing malfunctioning components like GPUs (graphics processing units) in orbit. Unlike terrestrial data centers, where failed equipment can be replaced quickly, servicing hardware in space would require complicated retrieval missions and adds to operational risk.

Musk projected that by 2028, orbital data centers would become the most economically attractive option for deploying AI infrastructure, citing lower energy costs and scalability. He further predicted that, within five years, annual launches of orbital AI capacity could exceed the historical total of terrestrial installations.

Currently, global terrestrial data center capacity is expected to reach approximately 200 gigawatts by 2030, representing investment of roughly $1 trillion. The prospect of shifting some of this capacity into orbit reflects anticipated demand for AI workloads, energy intensity, and the integration of satellite infrastructure directly into the compute supply chain.

Following its merger with xAI, SpaceX can internalize rocket launch costs — one of the main barriers to deploying orbital infrastructure — thereby potentially accelerating the pace of satellite data center deployments. The newly combined entity reportedly intends to pursue an initial public offering in the near future to raise additional capital for these projects.

The FCC’s acceptance of the application has opened a formal public comment period, allowing industry stakeholders and the public to respond to the proposal. If approved and implemented, the network could reshape how and where advanced AI computations are performed, with significant implications for both the satellite and data center industries.

Reference: dataconomy.com