SpaceX's Million-Satellite Bet: The Case for Orbital AI Data Centers

On January 30, 2026, SpaceX filed an application with the Federal Communications Commission requesting authorization to launch up to one million satellites designed to function as orbital AI data centers. The FCC accepted the filing on February 4, opening a public comment period that closed on March 6. It may be the most audacious satellite constellation proposal ever submitted — and it is the primary reason Terafab exists.

The filing describes a constellation of 'AI Sat Mini' spacecraft deployed across multiple orbital shells between 500 km and 2,000 km altitude, each carrying approximately 100 kW of computing power. At full deployment, the constellation would aggregate hundreds of gigawatts of AI processing capacity in orbit — dwarfing the combined compute of every terrestrial data center on Earth.

The Satellite Specifications

Each AI Sat Mini is designed around massive solar arrays that dominate the spacecraft's structure, extending to approximately 170 meters in length — larger than SpaceX's own Starship V3 at 124 meters. The arrays provide near-constant power in orbit, where solar irradiance is roughly five times greater than at Earth's surface with no atmospheric losses or weather interruptions.

The satellites communicate via high-bandwidth optical intersatellite links — technology already proven on SpaceX's Starlink constellation — creating a petabit-scale laser mesh network across the constellation. Ka-band radio links provide backup communications and telemetry connections to ground stations.

The Economic Thesis

Musk's economic argument for orbital computing rests on three pillars. First, solar energy in orbit is roughly five times more intense than on Earth's surface and available nearly continuously — no night, no clouds, no atmospheric absorption. Second, heat rejection in orbit is passive: the vacuum of space provides a natural thermal sink, eliminating the massive cooling infrastructure that consumes 30–40% of terrestrial data center energy. Third, orbital data centers require no land, no power grid connections, and no water supply.

The critical variable in the economic equation is launch cost. At current Falcon 9 prices (roughly $2,700/kg to LEO), deploying a million satellites would cost hundreds of billions in launch fees alone. But SpaceX is not planning to use Falcon 9. Starship, with its fully reusable design and massive payload capacity, targets launch costs below $10/kg to LEO. At that price point, the economics of orbital computing shift dramatically.

Deployment Scale and Timeline

The sheer scale of the proposed constellation raises fundamental questions about deployment logistics. Even with Starship's planned capacity, launching one million satellites would require thousands of flights. At 60 satellites per Starship launch (the current Starlink V3 baseline), full deployment would require approximately 16,700 launches — roughly 46 launches per day for a year.

More realistically, deployment would occur over a decade or more, with initial capacity coming online in phases. SpaceX's Starlink constellation — currently at roughly 7,000 satellites after five years of deployment — provides a template, but the AI satellite constellation would be two orders of magnitude larger.

Regulatory and Space Environment Concerns

The FCC filing has already drawn scrutiny from astronomers, space debris researchers, and competing satellite operators. One million additional objects in LEO — even if carefully managed — would dramatically increase collision risk and further complicate an already congested orbital environment. The Kessler syndrome risk, where cascading collisions create runaway debris, becomes a more urgent concern with constellations of this scale.

The optical brightness of satellites with 170-meter solar arrays would also significantly impact ground-based astronomical observations, extending the concerns already raised by the Starlink constellation to a much larger scale. Environmental review requirements under the FCC's updated space debris rules could add years to the approval process.

Why This Matters Beyond SpaceX

Whether or not SpaceX deploys a million orbital AI satellites, the FCC filing signals something important about the direction of the space industry: computing is becoming a primary use case for orbital infrastructure, not just a support function. For decades, satellites have been communication relays, imaging platforms, or science instruments. The orbital data center concept reimagines satellites as compute nodes in a distributed AI network.

This shift has implications for every segment of the space economy. Launch providers would see massive new demand. Ground station networks would need to scale. Space situational awareness systems would require significant upgrades. And the demand for radiation-hardened semiconductors — the D3 chips that Terafab is being built to produce — would increase by orders of magnitude.

The orbital data center concept may prove to be ahead of its time, technically impractical at the proposed scale, or exactly right. But the filing itself has already changed the conversation about what satellite constellations are for — and that shift will influence space industry strategy for years to come.

Frequently Asked Questions