Barely five months after emerging from stealth, space infrastructure startup Orbital Compute Inc. has filed an ambitious proposal with the Federal Communications Commission (FCC) to deploy a colossal network of up to 100,000 data center satellites.
The massive constellation aims to beam 10 gigawatts of space-based computing power directly to Earth, attempting to bypass the crippling power, land, and water constraints facing terrestrial artificial intelligence infrastructure.
The June 24 filings shed light on the technical architecture envisioned by the Los Angeles-based venture, which recently secured $5 million in pre-seed funding backed by Andreessen Horowitz’s a16z Speedrun program.
Heavy-Lift Micro-Data Centers in LEO
Orbital’s regulatory proposal outlines a fleet of heavy-duty, 100-kilowatt-class computing platforms operating in low Earth orbit (LEO) at altitudes between 500 and 850 kilometers. Unlike traditional, compact shoebox-sized cubesats, each individual satellite will be a massive orbital asset featuring a dry mass of 1.5 to 2.5 metric tons, flanked by expansive solar arrays and thermal radiators stretching nearly 100 meters across.
To transmit processed information back to users on the ground, the Orbital Datacenter System will bypass traditional radio frequency bottlenecks entirely. The network will rely heavily on high-speed optical intersatellite links (ISLs) to piggyback data across third-party communications constellations, explicitly naming SpaceX’s Starlink as a primary transmission route.
The Orbital AI Market Intensifies
The startup’s aggressive filing arrives amid a broader, high-stakes “space data center race” as tech giants look to orbit to house energy-hungry GPUs. Orbital’s proposal enters a crowded field of hyper-scale filings currently sitting before the FCC:
| Company / Project | Proposed Constellation Size | Satellite Power Class | Core Operational Target |
| Starcloud | 88,000 Satellites | 200 Kilowatts (kW) | High-Density Cloud Compute |
| Orbital Compute | 100,000 Satellites | 100 Kilowatts (kW) | Dedicated AI Inference (Tokens) |
| SpaceX | 1,000,000 Satellites | 150 Kilowatts (kW) | Integrated xAI / Starlink Network |
“The demand for AI compute is outrunning what we can reasonably build on the ground — we’re short on power, land, and water all at once…Space solves all three. Sunlight is constant, cooling is free, and there’s no neighborhood to disrupt. We think the next generation of data centers won’t be built in the desert — they’ll be built in orbit.” – Euwyn Poon, Founder and CEO of Orbital
The Multi-Year Deployment Roadmap
Recognizing the extreme capital and manufacturing hurdles ahead, Poon stated that the full 100,000-satellite swarm will not realistically be fully deployed until well into the next decade. Instead, the company is following an iterative, risk-reduction development curve:
- 2027 Demonstration Mission: Slated to fly on a SpaceX Falcon 9, this mission features a heavily scaled-down test payload designed to benchmark a single graphics processing unit (GPU) at roughly 1/100th the size of an operational node. It will validate sustained space-based GPU workflows, radiation hardening, and thermal dissipation in a vacuum.
- 2028 “Orbital-1” Flight: The startup’s first purpose-built compute satellite, engineered to mirror the final 100-kilowatt production design as closely as possible.
Applying the Micromobility Playbook to Orbit
Poon, a Cornell-educated engineer and lawyer, brings a unique hardware background to the space sector. He previously founded the micromobility company Spin, scaling an internal fleet of hundreds of thousands of electric scooters across 100 cities before orchestrating a $100 million sale to Ford.
He maintains that mass-producing orbital data centers is fundamentally a supply-chain and manufacturing challenge rather than an exotic science problem. Because these satellites do not require complex, steerable communication antennas, they are simpler to assemble than a standard telecom satellite.
“The complexity is all launch…The rest of it is first-principles physics and manufacturing. I mean solar panels, radiators, and electronics.”
– Euwyn Poon, Founder and CEO of Orbital
Orbital and its competitors will be heavily reliant on mega-heavy launch platforms like SpaceX’s Starship to make these multi-ton constellations economically viable.
Spin achieved profitability only after moving from permanent batteries to swappable power packs; a lesson in iterative design he plans to apply to successive generations of computing spacecraft. With a core engineering team assembled from SpaceX, Amazon, and Northrop Grumman, Orbital is actively scouting manufacturing partnerships to shift satellite production away from bespoke, one-off builds and into hyper-scaled, automated assembly lines.