Starcloud raised $170 million at a $1.1 billion valuation today, 17 months after Y Combinator demo day. The company filed with the FCC for an 88,000-satellite constellation to function as a distributed data center in orbit. They launched their first satellite with an Nvidia H100 in November and trained a small language model on it in December. Benchmark and EQT led the round. SpaceX filed separately for up to one million data center satellites in January. Orbital AI compute is no longer a talking point. It is a funded infrastructure category.
The FCC appears to understand the ground is shifting. On March 2, the Space Bureau opened a proceeding on satellite market-access reciprocity, asking whether the long-standing assumption that foreign markets are sufficiently open to U.S. satellite operators still holds. On March 26, the Commission advanced its "Weird Space Stuff" NPRM to expand spectrum access for next-generation orbital missions that do not fit neatly into old satellite categories: in-orbit servicing, orbital labs, lunar surface operations. In plain terms, the FCC is beginning to regulate for a world where satellites are not just bent-pipe relays. They are becoming network nodes, platform ecosystems, and potentially compute infrastructure in orbit. Give it a few years and posts like this one might be served from a sun-synchronous orbit at 800 kilometers.
The numbers
Eight companies have filed plans, launched hardware, or committed funding to orbital data centers in the last 90 days. The market is projected to grow from roughly $1.8 billion in 2029 to $39 billion by 2035 at a 67% CAGR. The physics argument is straightforward: continuous solar power, passive radiative cooling in vacuum, and no land or water constraints. The economic argument is less settled. Varda Space Industries estimates orbital compute at roughly 3x the per-watt cost of terrestrial equivalents today. The bet is that launch costs keep falling (SpaceX's Starship is the key variable) and terrestrial energy constraints keep tightening.
What the FCC is actually doing
The reciprocity proceeding (comments due April 1, replies April 16) examines whether the WTO-era presumption of open markets still works when the satellite industry is no longer just bent-pipe communications. The "Weird Space Stuff" NPRM (SB Docket No. 26-54) proposes two concrete paths: codifying "frequency piggybacking" so emergent operators can share spectrum with consenting host spacecraft, and opening a secondary allocation in the 2320-2345 MHz band for space operations telemetry. The FCC is also asking whether inter-satellite laser links should be authorized by rule for TT&C and data relay to emergent spacecraft. These are not abstract policy questions. They determine which companies get first-mover spectrum access for the next generation of orbital infrastructure.