The Federal Communications Commission’s (FCC) Space Bureau has accepted an application from Space Exploration Holdings, known as SpaceX, and is now seeking comment on the plan, including a suite of requested regulatory waivers.
Filed on 30 January, the application seeks approval for a new non-geostationary orbit (NGSO) constellation dubbed the SpaceX Orbital Data Center system. The satellites would operate at altitudes between 500 and 2,000 kilometres, across 30-degree and sun-synchronous inclinations, organised into orbital shells up to 50 kilometres thick.
SpaceX said the system would rely primarily on high-bandwidth optical inter-satellite links, allowing satellites to communicate directly with one another and with the company’s existing first and second-generation Starlink constellations. Telemetry, tracking and command functions would also be handled within the network.
The company has applied to use parts of the Ka-band spectrum – 18.3–19.3 GHz for space-to-Earth links and 28.6–29.1 GHz for Earth-to-space on a non-interference, unprotected basis.
Alongside the application, SpaceX is seeking exemptions from several FCC rules, including standard NGSO processing rounds, deployment milestones and surety bond requirements. It has also asked for flexibility around the level of technical detail required in its filings, particularly for beam layouts and orbital plane configurations.
The FCC has designated the matter a “permit-but-disclose” proceeding, meaning parties may engage with regulators, but those interactions must be publicly recorded.
At the centre of SpaceX’s pitch is the claim that orbital data centres could become the most cost- and energy-efficient way to deliver the computing power demanded by artificial intelligence. The company argues that terrestrial data centres, typically vast warehouses filled with high-performance computers, are already struggling to keep pace with AI workloads.
In support of its case, SpaceX pointed to launch capacity as the historical bottleneck. Even in 2025, described by the company as the most prolific year in spaceflight history, only about 3,000 tonnes of payload were placed into orbit globally, much of it Starlink satellites launched by SpaceX’s Falcon rockets.
The need to deploy thousands of satellites, SpaceX said, drove the high-cadence, reusable design of the Falcon program. The company now argues that its next-generation Starship launch vehicle will be the enabling technology for orbital data centres at scale.
According to SpaceX, Starship will begin launching larger “V3” Starlink satellites this year, with each flight delivering more than 20 times the capacity of current Falcon launches carrying V2 satellites.
Starship is also slated to deploy the next generation of direct-to-mobile satellites, which SpaceX claims will provide full cellular coverage worldwide.
For space-based data centres, the company envisages an even more dramatic ramp-up. It says Starship could ultimately fly as often as once per hour, carrying around 200 tonnes per launch. At that cadence, SpaceX argues that millions of tonnes of payload could be delivered to orbit each year.
In its filing, the company sets out a simplified calculation: launching 1 million tonnes of satellites annually, each generating about 100 kilowatts of computing power per tonne, would add roughly 100 gigawatts of AI compute capacity per year, with minimal ongoing maintenance once deployed.
SpaceX claims there is a long-term pathway to delivering as much as one terawatt of compute capacity per year from Earth launches alone.
SpaceX estimates that within two to three years, space-based computing could become the lowest-cost option for large-scale AI workloads, potentially allowing companies to train models and process data at unprecedented speed and scale.
The application also argues that the proposed constellation would build on the space sustainability practices used for Starlink, including end-of-life disposal to limit long-term orbital debris.
Beyond Earth orbit, SpaceX links the proposal to its broader ambitions for the moon and Mars. The company said Starship’s ability to transfer propellant in space would allow it to land large amounts of cargo on the moon, supporting permanent scientific and industrial bases. Over time, it argues, lunar manufacturing and electromagnetic mass drivers could enable satellites to be produced off-Earth and deployed deeper into space.
In the application, SpaceX described the project as an early step towards “becoming a Kardashev Type II civilisation”, a reference to a theoretical scale of technological advancement based on a civilisation’s ability to harness energy, in this case, a significant fraction of the sun’s output.
Elon Musk echoed that theme on his social media platform X, saying the proposed satellites would be spaced so far apart that they would be difficult to see from one another. “Space is so vast as to be beyond comprehension,” he wrote.
The scale of the proposal is unprecedented and is expected to draw close scrutiny from regulators, astronomers and space safety experts concerned about orbital congestion, spectrum management and the long-term sustainability of near-Earth space.