SpaceX filed for FCC approval of a million-satellite orbital data center constellation in early February. The filing advanced Wednesday with FCC acceptance and a public comment schedule. Chairman Brendan Carr shared it on X, signaling regulatory momentum. The formal SpaceX-xAI merger closed Monday, combining Musk's space and AI operations.
Musk's pitch centers on power economics. Solar panels generate roughly five times more power in orbit than on Earth, he told Patrick Collison's "Cheeky Pint" podcast. He claims 2028 will be a tipping point when orbital data centers become "the most economically compelling place to put AI."
The logic has gaps. Power is one cost among many in data center operations. Solar is one option among several for power generation. Getting hardware to orbit, maintaining it there, and managing failures remotely add costs Musk hasn't detailed publicly.
The competitive field is broader than SpaceX. Google plans Suncatcher test satellites by early 2027. Blue Origin is developing TeraWave for space-based connectivity. Starcloud deployed a refrigerator-sized satellite with an Nvidia GPU in late 2025. Axiom Space targets ISS-based modules by 2027. China's ADA Space is building AI-enabled satellite constellations.
For scale context: global data center capacity is projected to reach 200 GW by 2030, representing roughly a trillion dollars in infrastructure. SpaceX's proposed million satellites would dwarf the 14,000 currently in orbit.
The technical challenges are substantial. Orbital computing faces power constraints that limit machine learning inference workloads. Latency between orbit and ground stations varies by satellite position. Heat dissipation in vacuum requires different architectures than terrestrial cooling. Launch costs per compute unit remain unproven at scale.
Two infrastructure paths will likely run parallel: terrestrial data centers will continue absorbing most near-term growth through grid connections and renewable procurement, while a smaller cohort tests orbital systems. Most APAC CIOs should plan around terrestrial renewables rather than wait for orbital viability.
FCC approval isn't guaranteed. The proposal exists somewhere between technically feasible and economically foolish. History suggests skepticism until someone ships at scale. We'll see.