A five-month-old Los Angeles startup called Orbital has asked American regulators for permission to do something that sounds almost implausible: launch up to 100,000 satellites into space, each one functioning as a data centre powering artificial intelligence. According to SpaceNews’ reporting on the filing, the company submitted its application to the Federal Communications Commission on June 24, 2026, aiming to bring 10 gigawatts of computing power from space to meet rising demand for artificial intelligence. The filing follows the Los Angeles-based venture emerging from stealth earlier the same month with 5 million dollars in pre-seed funding, ahead of a planned demonstration mission next year.
What Orbital has actually proposed building
According to the same SpaceNews report, Orbital’s filing outlines plans to deploy 100 kilowatt class satellites in low Earth orbit at altitudes of 500 to 850 kilometres, each fitted with solar arrays and radiators spanning around 100 metres and a dry mass of 1.5 to 2.5 metric tons. Similar to orbital data centre plans filed earlier in the year by other companies, including Starcloud and Cowboy Space, Orbital said the primary data path for its proposed Orbital Datacenter System would rely on optical intersatellite links with third party constellations, such as SpaceX’s Starlink.
Why computing power is moving into space at all
Orbital CEO and founder Euwyn Poon, who previously founded the electric scooter company Spin before selling it to Ford, described orbital data centres as relatively straightforward systems built around solar panels, radiators and electronics, though he noted they must also operate in the vacuum of space and withstand radiation. The broader appeal of moving computing into orbit lies in sidestepping some of the biggest constraints facing data centres built on the ground, since space offers continuous solar power and a natural, cooling environment without the land and water demands of a terrestrial facility.
Why launch, not construction, is the real bottleneck
Poon has said building satellites at scale will be one of the company’s biggest challenges, describing orbital data centres themselves as fairly simple systems while noting that the complexity is all launch, pointing to the importance of heavy lift launch vehicles for deploying constellations of this size. According to SpaceNews, Orbital is among several companies waiting for SpaceX’s Starship to become available before large orbital computing networks can realistically be deployed at scale.
How Orbital’s plans compare to the wider industry
Orbital’s proposal joins a rapidly crowded field of similar filings over the past year. Poon noted that Starcloud, a rival startup, is targeting 200 kilowatt satellites for its own proposed 88,000 spacecraft constellation, an indication that Orbital’s own satellite specifications could still be revised upward as the company finalises its design. This wave of activity reflects a broader shift in the industry, moving from early stage announcements toward actual regulatory filings, with several major companies now pursuing orbital computing constellations of their own.
What comes next for Orbital’s ambitious plan
Orbital is preparing a demonstration mission for next year, though Poon has said the first payload will be much smaller than the hardware planned for future operational satellites, describing it as maybe one one hundredth the size of a single GPU setup planned for later missions. The company plans to follow that demonstration with Orbital-1, its first dedicated compute satellite, currently slated for 2028, which Poon said is being designed to resemble the future production version as closely as possible, even though the full constellation is not expected to be deployed until well into the next decade. Poon has also said that some specifications could still change before the design is finalised, and that Orbital is looking for manufacturing partners and broader collaboration opportunities while the core satellite platform is designed in house from Los Angeles.
Why this remains a genuinely open question
Poon has compared the scale of the challenge to his experience building Spin, where early design choices around swappable components made a meaningful difference to efficiency at scale, suggesting Orbital could see similar gains across successive generations of its own satellites. Whether Orbital and its competitors can actually solve the launch costs and manufacturing scale needed to make a constellation of 100,000 satellites viable remains genuinely uncertain, and much of that answer will depend on developments outside Orbital’s own control, particularly how quickly SpaceX’s Starship becomes a reliable, frequently reused launch vehicle in the years ahead.
