SpaceX’s “One Million Satellites” AI Data-Center Plan Is Real — but the FCC Has Only Opened the File, Not Approved It

SpaceX has filed for permission to build what may be the most audacious communications-and-compute infrastructure proposal in spaceflight history: a new non-geostationary satellite system of “up to one million satellites” described as an “Orbital Data Center” network designed to support AI computing in orbit. 

But one central detail in early summaries needs the same kind of correction that matters in governance stories: this is an application, not an authorization — and the timeline is slightly off in many retellings.

The correction: the filing date — and what “accepted for filing” actually means

SpaceX filed the application on January 30, 2026 (a Friday), according to the FCC’s Space Bureau public notice. 

That matters because some versions of the story describe a “late Friday, January 31” filing — January 31, 2026 was a Saturday.

On February 4, 2026, the FCC’s Space Bureau accepted the application for filing and opened it for public comment. 

“Accepted for filing” is procedural: it means the FCC will consider the request and invites comment; it is not approval.

What SpaceX is actually asking the FCC to authorize

The FCC notice summarizes SpaceX’s request as a new NGSO system of up to one million satellites, operating as the “SpaceX Orbital Data Center system.”  Key technical points disclosed in the FCC public notice include:

• Altitudes: roughly 500 km to 2,000 km 

• Inclinations: 30-degree and sun-synchronous orbits 

• Orbital shells: spanning up to 50 km each 

• Architecture: heavy reliance on high-bandwidth optical inter-satellite links (laser links), with connectivity that may integrate with existing Starlink generations 

• Spectrum request: Ka-band segments 18.3–19.3 GHz (space-to-Earth) and 28.6–29.1 GHz (Earth-to-space) on a non-interference, unprotected basis

The FCC notice also indicates SpaceX is seeking waivers of certain FCC rules as part of the application package.

The “orbital data center” pitch — why SpaceX is pushing compute off Earth

Independent reporting based on the filing describes the concept as distributed in-orbit compute nodes powered by solar energy, designed to address terrestrial AI constraints such as electricity supply, cooling, and siting limitations. 

This is the strategic bet: the next AI bottleneck may be energy and infrastructure, not only chips — and space offers near-continuous sunlight and a heat-rejection environment that proponents argue could change the economics of large-scale compute. 

The uncomfortable math: orbit is not empty — and scale changes everything

Even before a million-satellite ceiling, low Earth orbit is already under strain.

• ESA’s 2025 Space Environment Report says about 40,000 objects are tracked, including about 11,000 active payloads. 

• In early February 2026, expert satellite tracking cited by Space.com put Starlink alone at 9,628 active satellites. 

A filing ceiling of “up to one million” doesn’t mean one million will fly — operators often file high to preserve flexibility, and regulators frequently force narrowing. 

But the ceiling is still consequential because it sets the starting point for the regulatory negotiation: orbital shells, spectrum rights, interference protections, collision-avoidance commitments, deorbit reliability, and transparency obligations.

The FCC’s dilemma: not a traffic authority — but a gatekeeper anyway

The FCC is not the world’s space traffic controller. Yet it is a critical gatekeeper because it licenses NGSO systems and spectrum use — decisions that shape what becomes economically viable and operationally normal.

This application forces a set of historic questions onto a process built for a smaller era:

• How much orbital “infrastructure” can be licensed before debris risk becomes a public-interest veto?

• What standards of deorbit reliability and collision avoidance should be required at megaconstellation scale?

• How should the FCC treat “non-interference, unprotected” spectrum claims when the applicant is proposing a civilization-scale network? 

The geopolitical subtext: space-based compute is becoming a sovereignty contest

Space-based data centers are not only a Musk idea — they are emerging as a strategic track globally.

• Reuters reported China’s major space contractor (CASC) has outlined a five-year push toward space-based AI data centers and a “Space Cloud,” including “gigawatt-class” digital-intelligence infrastructure ambitions. 

• Recent coverage also points to China’s “space computing” efforts led by ADA Space, which has been publicly associated with a planned constellation. 

So the FCC’s decision path on SpaceX will be read internationally as more than paperwork: it signals how the U.S. intends to regulate (or accelerate) a new category of strategic infrastructure.

What we can say with confidence — and what remains uncertain

Verified from FCC documentation:

• SpaceX filed on January 30, 2026 for an NGSO system of up to one million satellites. 

• The FCC Space Bureau accepted the application for filing and opened comment on February 4, 2026. 

• The system parameters include 500–2,000 km, 30° and sun-synchronous inclinations, and shells up to 50 km. 

• The spectrum request includes 18.3–19.3 GHz and 28.6–29.1 GHz on a non-interference, unprotected basis. 

Not yet determined (because the FCC has not ruled):

• Whether the FCC will approve anything close to the ceiling, or force staged caps/phasing.

• The final safety, reporting, and deorbit conditions — typically the core battleground in NGSO licensing at scale.