Data Centers Are Running Out of Room. Companies Want to Put Them in Orbit

AI and cloud buildout is straining power, water, and land on Earth, and four orbital data center proposals, including one for up to a million satellites, are now under Federal Communications Commission review, moving the orbital option from thought experiment to filing.

The largest of the four would field up to one million satellites equipped with optical inter-satellite links, lasers that pass data between satellites without touching a ground station, and would use specific radio frequencies to talk to Earth. The other three proposed constellations would deploy roughly 88,000, 51,600, and 100,000 satellites respectively, all in sun-synchronous orbit, a polar, sun-synchronized path that gives continuous sunlight to solar panels and stable thermal conditions, useful for power-hungry hardware that cannot be repaired on a service call. Starcloud's 88,000-satellite filing is the most visible of the four; the largest proposal of one million satellites sits in its own docket and would dwarf the current global satellite fleet many times over if approved as filed.

All four applications have been accepted by the FCC and are undergoing agency review, according to a WilmerHale client alert summarizing the filings. The review stage is the institutional step that determines whether any of them move toward a license. It is the first time orbital data centers have reached this regulatory moment in public.

Proponents argue orbital data centers are a way around terrestrial constraints. Power-hungry AI training clusters and dense cloud regions are running into real limits: long grid interconnection queues, water-rights fights in drought-stricken counties, and the flat land a hyperscale campus needs. Space, in this view, gives continuous solar power, no water cooling, and no local permitting fight.

Filing, however, is not deployment. The FCC's acceptance puts the proposals into agency review, not into a construction queue. The review will weigh spectrum rights, orbital debris mitigation, and interference against each proposal. Each application sits in its own docket, and the agency's order on any one of them sets a precedent the others inherit. Given the size of the proposed constellations, end-of-mission disposal is a likely focus.

There is no single U.S. regulator for orbital data centers. The WilmerHale alert lays out the picture: a constellation in sun-synchronous orbit may need spectrum coordination through the FCC, launch and reentry approvals from the FAA, NOAA engagement if remote sensing is involved, and export-control clearance if the system touches certain technologies. Beyond U.S. borders, companies will need approvals from the International Telecommunication Union, the U.N. body that allocates radio spectrum globally, and from national regulators for any ground stations in their territories. The result is a multi-agency diligence exercise for any serious funder: regulatory, cybersecurity, data governance, insurance, operational risk, and end-of-mission obligations all sit on the table before construction.

The engineering test bed is another open piece. Space is hostile to the silicon that runs modern AI. Radiation gradually degrades compute and memory; thermal management is harder in vacuum than in a chilled-air data hall, with no easy convective cooling. Maintenance and replacement are real costs because no one services a server in orbit today; hardware that fails stays failed, and the constellation has to keep working around it. End-of-mission disposal is a regulatory and orbital-debris problem of its own. Each of these has been solved for individual spacecraft, and a few have been solved for satellite constellations, but none has been solved at the scale of a million satellites carrying compute payloads.

The economics are open as well. Launch, hardware, and maintenance costs at the proposed scale are still estimates rather than delivered prices, and ground buildout remains cheaper for most workloads today.

The four filings are early, the engineering, regulatory, and economic gaps are named and visible, and the path from application to operational constellation runs through more agencies and harder physics than the filings imply.

Which of the four applications actually clears the FCC's review step, and what the agency says about the open technical and policy questions, is the next concrete test.